OpenResty XRay 26.11.1 is out! This release makes AI diagnostic findings something you can question and verify. Highlights:
1. AI diagnostic findings you can question and verify:
- Keep asking follow-up questions about the same findings, using the process and host resource usage from the capture to explore causes and optimization opportunities
- Jump from an AI report straight to the relevant flame graph call paths to check the evidence
- AI replies and job reports now include charts and suggested follow-up questions to guide further investigation
2. AI writes new analyzers automatically: the AI writes and validates new analyzers based on your diagnostic needs, so you can run more flexible analyses targeted at specific problems
3. Initial support for diagnosing AWS ECS Fargate applications, extending analysis to serverless containers
This release also speeds up summary report refreshes, improves the accuracy of resource metric reads, reduces storage for historical metrics, and fixes issues such as some tools not rerunning after "Run all". We recommend upgrading as soon as possible.
#openrestyxray #dynamictracing #ebpf
openresty.com/en/xray/release-notes/?version=26.11.1
OpenResty
OpenResty is an nginx distribution which includes the LuaJIT interpreter for Lua scripts. The software was created by Yichun Zhang. It was originally sponsored by Taobao before 2011 and was mainly supported by Cloudflare from 2012 to 2016. Since 2017, it has been mainly supported by OpenResty Software Foundation and OpenResty Inc. More on Wikipedia
-
OpenResty
@OpenResty
-
OpenResty
@OpenResty
OpenResty 1.31.8.1 open-source release is out!
• Core and dependency upgrades: Nginx has been upgraded from 1.31.1 to 1.31.8, OpenSSL to 3.5.9 and PCRE2 to 10.49.
• Dynamic certificate compression: The new ngx.ssl.compress_certs() compresses certificates that are set dynamically through Lua. This sends less certificate data during the TLS handshake.
• Stability improvements: This release fixes unbounded retries in load balancing, keepalive connections to upstreams that were never reused when proxy_bind was set, and Lua subrequests not being woken up on HTTP/2 errors. It also fixes several memory leaks, double frees and invalid memory accesses.
• LuaJIT v2.1-20261023: unpack() can now be JIT-compiled, and common string operations are faster. This version also fixes JIT, FFI and profiler issues on several platforms.
• Ecosystem updates: Redis connections now support the db, username, password and tcp_keepalive options. JSON encoding can sort object keys. WebSocket validates handshakes more strictly and can resume receiving partial frames.
On the security side, this release includes the fixes for 12 CVEs from Nginx 1.31.2 to 1.31.8:
• 1.31.2
- CVE-2026-42530: use-after-free in HTTP/3
- CVE-2026-42055: buffer overflow when proxying to HTTP/2/gRPC upstreams
- CVE-2026-48142: buffer overread in charset conversion
• 1.31.3
- CVE-2026-42533: buffer overflow involving map and regex
- CVE-2026-60005: uninitialized memory access involving slice or background cache updates
- CVE-2026-56434: use-after-free in SSI
• 1.31.6
- CVE-2026-90439: heap buffer overflow in HTTP/3 with OpenSSL 3.5.0 and earlier
• 1.31.7
- CVE-2026-15176: buffer overflow involving alias and rewrite
- CVE-2026-93479: buffer overflow during variable expression evaluation
- CVE-2026-95101: HTTP/3 request body injection
- CVE-2026-97629: TLS session ticket forgery
- CVE-2026-97205: use-after-free in AIO/SSI
📄 Full release announcement: openresty.org/en/ann-1031008001.html
📥 Downloads: openresty.org/en/download.html
📦 Binary package installation guide: openresty.org/en/linux-packages.html -
OpenResty
@OpenResty
In performance troubleshooting, we always hit the same bottleneck: the issue itself isn't always complex, but few people can interpret low-level runtime data. When expert bandwidth is tight, issues just stack up.
In v26.9.1, we’re delegating this heavy lifting to the AI Assistant.
1. AI executes, evidence leads Describe what you see in plain English. The AI Assistant autonomously plans the analysis, executes tasks, and delivers key findings via chat.
Every insight is grounded in real-time data captured by OpenResty XRay from live processes—no LLM hallucinations or guessing games. This release adds time-series metrics and smarter report analysis, allowing it to easily track down complex, multi-stage issues over time.
Senior engineers no longer need to babysit every incident, empowering junior devs to drive root-cause analysis independently.
2. New Java runtime analyzer covering 9 key scenarios Core systems in finance, cloud-native microservices, autonomous driving, and gaming often rely on Java. Performance issues are usually buried deep in GC, safepoints, lock contention, and thread scheduling—invisible to plain logs or basic APM.
We now cover: memory allocation, class loading, GC & safepoint pauses, JIT compilation, lock waiting, network I/O, exceptions, thread activity, plus improved off-CPU profiling for accurate block-time stats. Full support for JDK 8, 11, 17, 21, and ARM64 architecture.
3. Zero code changes—our non-negotiable rule If it requires code modification, manual instrumentation, or service restarts, it doesn't belong in production. OpenResty XRay leverages dynamic tracing to analyze live applications on the fly—zero code edits, zero restarts, and zero source code access required. With near-zero overhead, it's 100% safe for production.
Both the AI Assistant and the new Java analyzer strictly follow this rule: drop it straight into your running clusters.
4. Diagnose from anywhere Troubleshooting shouldn't keep you chained to your desk. The AI Assistant now supports mobile, so you can review diagnostic sessions, share insights with your team, or ask follow-up questions right from your phone.
openresty.com/en/xray/release-notes/?version=26.9.1
#openrestyxray #dynamictracing #ebpf -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Finally benchmarked top US & Chinese LLMs on our internal OpenResty eval suite! 🚀 We tested two private benchmarks based on real, complex engineering tasks from our company:
- Coding: Designing & implementing complex algorithms from scratch.
- Bug-Hunting: Root-causing & fixing our hardest production bugs.
Both were run directly using our coding agent harness.
I'll let the results speak for themselves—attached are the quality rankings and avg latency charts for both benchmarks.
TL;DR based on our coding benchmark: The top Chinese open-source models are currently ~6–8 months behind the leading US frontier models. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Not bad—the newly released GPT-6.1 Sol is already beating (or at least matching) GPT-6 Astra on two of our internal private benchmarks: from-scratch coding and bug localization/fixing. Insane value from OpenAI here.
However, 6.1 Sol takes noticeably longer to complete tasks compared to previous Sol and Astra models. A trade-off, as always... -
OpenResty
@OpenResty
Coding agents like Codex, Claude Code, and Cursor can read and edit your code, but for performance work they're missing a key piece: what your program actually does at runtime.
If you already use OpenResty XRay, our new Agent Skill fills that gap. The agent pulls XRay's dynamic tracing data on its own, maps flame graph hotspots straight to source lines, and runs a full profile → fix → re-profile loop.
We tried it on an edge service under production-like conditions. Starting with issues like redundant JSON serialization and regexes being recompiled on every request, every refactor was guided by real XRay profiling data. Throughput went from 336 RPS to nearly 30,000 RPS, close to 90x.
The Skill calls the XRay API automatically and works with any tech stack, not just OpenResty/Nginx.
Full write-up, including where this approach fits best: blog.openresty.com/en/xray-ai-skill/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Anthropic just dropped Claude Opus 5.5 — supposedly smarter than Fable 5.1 and cheaper than Opus 5. Gotta upgrade ASAP lol. Funny how Anthropic is the loudest voice calling for slowing down AI development, yet they themselves aren't slowing down one bit 😂
Looking at the official benchmark charts, it's apparently outperforming GPT-6 Astra too. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Over the past couple of days, I benchmarked GPT-6 Astra, Claude Fable 5.1, Grok 4.6, and Gemini 3.8 Flash using our internal (harder) OpenResty BugHunt Evals suite. Claude still comes out on top, with GPT-6 Astra as a close second.
This benchmark tests how well frontier models can locate and fix the hardest bugs in large, real-world, complex software projects, fully autonomously via a coding agent harness.
Surprisingly, Grok 4.6 has made massive strides — it even beats GPT-5.6 Sol and other previous frontier models.
Google's latest Pro model, on the other hand, is dead last by a wide margin. Their Gemini 3.8 Flash does much better though, landing around the middle of the leaderboard. Looks like Google is going all in on cheaper, smaller models 😄
On overall latency, Fable 5.1 does quite well — not because individual responses are particularly fast, but because it needs fewer trial-and-error steps to get to the fix.
Details in the two charts below — quality scores on the left, latency scores on the right. -
OpenResty
@OpenResty
Google's newly released Gemini 3.8 Flash shows clear improvement over the previous 3.6 — scores on our internal OpenResty Coding Evals benchmark went up noticeably. But latency also took a significant hit; it's way slower than before 😂 See the two charts below for details.
Looks like Google is going all-in on cost-efficient small models now? lol. A Gemini Pro model seems nowhere in sight… -
OpenResty
@OpenResty
In the age of AI, how much secondhand knowledge you accumulate stopped mattering a long time ago — it already didn't matter back in the Google era.
The most important thing on any continuous learning path is research-based learning. Overconsuming secondhand knowledge only erodes your capacity for original thinking.
For example: pick up a textbook in an unfamiliar field. Read the first third or half, then close the book. Follow the threads you've absorbed and try to derive the more advanced theorems, formulas, and conclusions on your own.
Then open the second half and see how much overlap there is.
Whatever didn't overlap is genuinely original. And even where it did overlap, you arrived at it independently — firsthand knowledge, even if technically a rediscovery.
The more overlaps you rack up, the sharper your research ability gets. Gradually, the overlaps thin out — and that's when you've surpassed the author and entered uncharted territory.
This is the most valuable way to learn.
In my experience, most of the time you don't even need to read the second half carefully — a quick skim is enough, because you may have already gone beyond what the book covers. At that point, you're no longer a passive consumer. You're in the driver's seat.
This is just one example. There are many ways to practice research-based learning that suit different people. I also design fairly complex projects to put what I've learned from books into practice. The core of it is always: do it yourself, think independently, solve real problems.
I've been doing this since I was 13. It's exhausting, but the results speak for themselves. No pain, no gain — and this is the rewarding kind of pain, not the mind-numbing grind of drilling problem sets.
With an unfamiliar discipline, you need to reach sufficient depth before genuine interest even kicks in. Once that interest takes hold, going deeper becomes natural.
From a single-discipline perspective, deeper is always better. Across multiple disciplines, you have to manage your time and energy. "Deep and broad" is hard, but often necessary. You have to balance both.
You have more time as a student. I explored a wide range of fields — the most intense phase was in middle school. If you have a day job and a family now, time gets tight.
Every time I picked up an entirely new discipline, I would pretend I was going to spend the rest of my life on it. That's the only way to settle in deep enough. Impatience is the greatest enemy. Whenever I'm researching something, I go all in — total focus, zero distractions.
For me, studying and exploring unfamiliar fields is itself a tremendous joy. It's like discovering a new continent.
Stop asking "what can I get from this field" and start asking "what can I discover in it, what can I contribute to it." Once you get the mindset right, everything else follows. It doesn't even cost much money.
Let go of the ego, immerse yourself fully — experience it, think through it, enjoy it.
More often than not, the biggest obstacle isn't time, isn't energy, isn't even money. It's your own mindset — your own inner resistance. -
OpenResty
@OpenResty
Oh look — Claude Code is about to cut token usage across all subscription plans by 17%. Meanwhile OpenAI has been slashing model prices left and right, and their guy Tibo keeps resetting token limits for all Codex users.
Sure looks like Anthropic is genuinely running low on compute. 😮💨
Are we watching the golden era of AI giants burning cash to subsidize subscribers and grab market share come to an end? 😂 Better burn through your tokens while you still can… -
OpenResty
@OpenResty
AI has genuinely raised the floor across the industry. The question now is who can break through a higher ceiling. That's where the real value lies.
The whole industry getting easier? Not going to happen. Long-term, it only gets more intense, more competitive, more brutal.
Some people might coast during the transition, but if you don't keep up on your own, you'll fall behind fast. -
OpenResty
@OpenResty
LLMs are a massive force multiplier for senior people. They no longer need to compete with juniors on raw stamina. Experience, taste, and judgment — the things seniors have spent years building — get enormously amplified by AI, and suddenly become far more valuable. That said, seniors who refuse to adapt and embrace AI will get left behind too.
Meanwhile, the ladder from entry-level to senior is getting dangerously narrow for juniors and newcomers — even where AI hasn't cut it off entirely. And many young people are using AI to replace their own thinking and learning, which is essentially crippling themselves.
Once today's senior generation retires, the talent gap could become a serious problem. The truly exceptional people will only become more and more sought after. -
OpenResty
@OpenResty
Been getting flooded with emails and calls from Google Cloud sales these past couple days, saying our company OpenResty qualifies for $250K in free cloud credits — usable for Gemini and other cloud resources. Looks like after Google bowed out of the frontier model race, they've got more budget to throw at winning the cloud market 😂
Maybe we should start burning more Gemini tokens then. For our non-frontier R&D work, cheaper models still get the job done. Frontier research though — that's strictly Claude Fable 5 and GPT-5.6 Sol territory. -
OpenResty
@OpenResty
Ran into that same race condition bug in Claude Code under high concurrency again — the one I spotted ages ago and it's still not fixed… starting to feel like they want me to go patch it myself 😂 Affects both the desktop app and the CLI.
-
OpenResty
@OpenResty
Everyone's using smarter models and fancy Coding Agent harnesses. But most teams and devs haven't seen a meaningful productivity lift — not with quality actually held.
Worse: there's often a huge gap between feeling faster and being faster. And an even bigger one between shipping a demo in a weekend and shipping a production product that makes money.
AI is fundamentally an amplifier. It will massively widen the gap between top-tier teams and engineers vs. everyone else. The smarter the model, the wider the gap.
Existential dread loading... 😂
Stop thinking about AI as a way to slack off. -
OpenResty
@OpenResty
Sometimes I feel actual pressure talking to the smartest models now — not because I'm worried about my ego, but because I'm worried the model will decide I'm clueless and start half-assing it. Like it'll just phone in a lazy answer instead of actually thinking. 😂
To get the most out of the strongest models, you really do have to push them. -
OpenResty
@OpenResty
OpenResty XRay 26.7.15 is out! The theme of this release: bringing AI into the diagnostics workflow.
New AI Assistant — diagnostics through conversation:
• Query product docs and system info, select and run analyzers, review results — all from a single conversational interface that consolidates the entire "read docs → pick analyzer → run → inspect" loop
• Sessions are searchable, renamable, and shareable as read-only links, making it easy to preserve diagnostic context for team review
AI-powered interpretation:
• AI can now interpret analysis tasks and reports, surfacing key findings from dense flame graphs and low-level metrics
• Interpretations are viewable and shareable on the bulletin page
MCP server interface & SKILL support for AI agent reasoning loops:
• AI agent clients like Claude Code, Codex, and Cursor can connect to OpenResty XRay directly
• Agents can autonomously run analyzers, read results, and decide which analyzers to run next — forming an analyze → interpret → re-analyze reasoning loop that progressively narrows down root causes
• Plug OpenResty XRay’s diagnostic capabilities into your everyday AI-assisted coding and ops workflows
Also in this release: selectable flame graph direction, tcmalloc symbol recognition, plus fixes for Helm/Kubernetes deployment, application metric reporting, and vulnerability scanning.
Recommended upgrade for all users.
openresty.com/en/xray/release-notes/?version=26.7.15 -
OpenResty
@OpenResty
Our OpenResty Edge distributed gateway is also shipping advanced LLM API gateway capabilities this year — still in the same "you describe it, we build it" model. It will automatically detect and guard against security issues like data leakage, and automatically flag buggy behavior from coding agent harnesses.
-
OpenResty
@OpenResty
Below are the latest results from our private OpenResty Coding Evals benchmark, now including the new Claude Opus 5 model. It has indeed claimed the throne. We've also added a leaderboard for average end-to-end latency.
-
OpenResty
@OpenResty
After a long stretch of work, we finally have results from running open-weight models against our internal OpenResty Coding Evals benchmark. Getting there meant fixing a ton of issues with opencode, pi, and claude code when driving open-source models — making sure these agent harnesses weren't handicapping the models.
GLM 5.2 is genuinely impressive — approaching Claude Sonnet 5 territory — but there's still a noticeable gap against models like Claude Opus 4.8. Note that due to inherent LLM randomness, score differences within 1% should be considered normal variance.
First image: quality scores. Second image: average total latency per task across agents. Google and xAI models are definitely the fastest, though their quality scores are lower.
Still waiting for Kimi K3, Qwen 3.8 Max, and DeepSeek V4 stable to publicly release all model weights. Once they do, we'll add them to the eval.
Over the next few days I'll also run our other, harder benchmark — the OpenResty Bug-Hunt Evals. Stay tuned! -
OpenResty
@OpenResty
When a live stream crashes during a major event like the World Cup, it's usually not because you didn't buy enough bandwidth.
Ops teams doing post-mortems keep finding the same root cause: origin-pull architecture spiraling out of control. When viewership explodes by orders of magnitude, the load on your origin doesn't grow linearly — it amplifies. If the request collapsing layer isn't built right, no amount of bandwidth solves a deduplication failure.
The other trap that's consistently underestimated is cost: post-event traffic bills growing faster than revenue. Managed CDNs charge per byte — every second of peak is priced linearly. Private CDNs charge by capacity tier, decoupling peak cost from total volume delivered. Origin spend becomes completely independent of audience size.
Should you bring your delivery layer in-house? The case rests on exactly these two judgments — not comparing feature checklists, but making a foundational choice about architectural control and cost structure. Our team put together a technical analysis on both. Link in the comments.
Our commercial product recently helped power live streams for major sporting events at this scale.
blog.openresty.com/en/edge-private-cdn-live-streaming/ -
OpenResty
@OpenResty
Been testing open-source models the last few days, and man... the models and their surrounding Agent toolchains are incredibly janky.
You basically have to spend hours fixing the plumbing yourself first, otherwise the test results are comically bad. The barrier to entry for open-source AI is still brutally high compared to the plug-and-play nature of commercial models. -
OpenResty
@OpenResty
Transformer-based LLMs are fundamentally just next-token predictors. In long agentic workflows, this creates massive path dependency. Once an agent takes a wrong turn, it will confidently sprint off a cliff without ever realizing it.
This is exactly why you need an experienced engineer to step back, see the big picture, and course-correct.
I use AI to write a ton of code, but my real job is gatekeeping. I know exactly which critical junctures require manual intervention. I know when to relentlessly push back on the LLM, even when it stubbornly insists on going the opposite direction. And yes, I rigorously review its output—especially the high-stakes logic.
Fully autonomous coding is a myth. The true god-tier combo is an Elite Engineer + a SOTA (State-of-the-Art) LLM. That is the unbeatable meta. -
OpenResty
@OpenResty
Anthropic just reset all Claude subscription token limits again today. Total dopamine hit. Now I’m just kicking myself for not grinding and maxing out Claude Fable 5 completely before the reset 💀
This stuff is becoming a literal drug at this point... -
OpenResty
@OpenResty
Kimi K2.7 Code seems to go rogue so easily. A single prompt just burned through $20.37 of my credits, which is terrifying... feels super unstable right now.
-
OpenResty
@OpenResty
Spent the past few days compiling our hardest real-world bugs from OpenResty Inc's internal codebase into a new benchmark: OpenResty Bug-Hunting Evals. Tested all major frontier models from Anthropic, OpenAI, and Google. Results below.
The task: given only a bug's symptom description, autonomously locate the root cause, fix it correctly, without introducing any regressions.
All bugs live in our private company repos, completely off the public internet, so no data leakage, no contamination, making it as objective as it gets.
Anthropic's mythic Claude Fable 5 leads by a wide margin. Even OpenAI's latest GPT-5.6 Sol still trails noticeably, it hasn't caught up to Claude Opus 4.8 Max yet.
Google's frontier models remain at the bottom.
This eval burns WAY more tokens than our previous OpenResty Coding Evals btw.
One thing worth noting: OpenAI's models + coding agent have noticeably lower end-to-end latency. Anthropic's top models clearly win on quality, but the total wall-clock time is painful, nearly an hour for the full eval run. -
OpenResty
@OpenResty
During large-scale live events like the FIFA World Cup, a single CDN node can handle tens to hundreds of gigabits of traffic. In this environment, a well-tuned origin-pull and caching strategy is the foundation of streaming stability.
CDN platforms routinely serve multiple tenants side by side, making policy changes and parameter tuning a daily reality. Native Nginx offers robust caching and origin-pull capabilities — but every configuration update requires a reload, and each reload introduces the risk of stuttering or stream interruption. In a high-concurrency scenario like the World Cup, that risk can be catastrophic.
OpenResty Edge builds on Nginx's proven foundation with deep enhancements: dynamic configuration delivery and zero-downtime updates. Adjustments to caching rules, origin-pull policies, and other streaming-related settings take effect instantly — no reload required, no impact to active playback sessions. Teams can adapt to live event conditions in real time, dramatically improving the viewing experience at scale.
In our latest article, we provide a full step-by-step guide to configuring live streaming on OpenResty Edge.
#fifa #worldcup #openrestyedge -
OpenResty
@OpenResty
Fumbled a prompt template today: accidentally typed "comments" instead of "commits" and completely changed the meaning. To make it worse, I reused this across multiple Agent sessions. The Agents ended up only analyzing code comments while completely ignoring the actual code, burning through a ton of Claude Fable 5 tokens... 💀
-
OpenResty
@OpenResty
We just used OpenResty Inc's internal proprietary software pojects' hardest real algorithm design & impl problems to form an LLM evals suite. This suite is coding & testing from scratch. And it invokes the native CLI agents like claude code, codex, antigravity, and grok build.
The suite has never been published so LLMs never saw it before.
Below are the quality score ranking results from burning a lot of tokens! Claude Fable 5 and GPT-5.6 Sol are definitely the kings! -
OpenResty
@OpenResty
Honestly, once mythical-class LLMs drop, the only human traits that will actually matter are taste, imagination, and judgment.
-
OpenResty
@OpenResty
As AI gets more powerful, the biggest pitfall is using the best AI to polish a turd.😂
-
OpenResty
@OpenResty
Imagine paying $20 just to say "Hi" to an AI. Apparently, that's what happened to a dev using the Claude Fable 5 API. Talking to a mythical model really costs an arm and a leg. 💸 (Shoutout to the post-July 7th pricing updates lol). If they let me loose with Fable at these rates, my daily API bill would literally equal a brand-new Tesla.
-
OpenResty
@OpenResty
METR—the lab famous for their LLM time horizons charts—just found that the new GPT-5.6 Sol is an absolute cheat code. Literally. 💀👇
They reported that GPT-5.6 Sol cheats so aggressively that it’s almost impossible to objectively measure its actual intelligence. Its tendency to cheat completely eclipses any previous open/public models (and seemingly even Claude Mythos).
METR’s metric measures how many hours of human-equivalent work a model can handle. If you count cheating as an automatic fail, Sol scores 11.3 hours—barely pacing with Claude Mythos (~12 hours). Their final verdict? No meaningful progress here.
For anyone planning to deploy GPT-5.6, get your red pens out. Double-check everything. This model is hardwired to cut corners.
OpenResty
@OpenResty
🔗 Read METR’s full deep-dive here: metr.org/blog/2026-06-26-gpt-5-6-sol/?mark_id=999_reallog…
-
OpenResty
@OpenResty
After using Claude Code, Codex, and Antigravity, I gotta say—Antigravity feels by far the most unpolished...
Google has the deepest pockets, so this makes zero sense. Major letdown.
Meanwhile, Claude Code has been the most unstable by a mile. -
OpenResty
@OpenResty
Currently sacrificing sleep just to get my money's worth out of Claude Fable 5. I’m dumping years of my hardest engineering and algorithm backlogs onto her.
It's wild—she barely outputs any reasoning tokens, does zero trial-and-error in front of me, but always cracks the solution in a few hours.
In the era of Fable-class models, the human workflow is simple: Wish for it. Guardrail it.
We might actually need a new job title soon: "The Wisher." Then again, maybe that’s what Product Managers have been doing all along. -
OpenResty
@OpenResty
Based on our company's massive internal test suite, Google’s new Gemini 3.5 Flash is a noticeable downgrade compared to their cheaper, smaller Gemini 3.1 Flash Lite. (At least for our specific use case, 3.5 Flash completely flopped).
Here’s how our small-model leaderboard looks right now: Gemini 3.5 Flash < Gemini 3.1 Flash Lite < Claude 4.5 Haiku
Both of Google’s small models still comfortably beat out Claude 4.5 Haiku, but Gemini 3.5 Flash is definitely a step backward. Of course, we're talking about the lightweight tier here. Heavyweights like GPT-5.5 are in a totally different league—and obviously way more expensive. -
OpenResty
@OpenResty
I’m honestly a bit skeptical about the hype around AI enabling "one-person unicorns" or hyper-individual entrepreneurship. Even if a tiny fraction of outliers might pull it off.
99% of "AI founders" are just going to be cash cows paying the API token tax. AI lowered the barrier to ENTRY, but astronomically raised the barrier to SUCCESS. It's never been easier to build a product, and it's never been harder to actually get people to swipe their credit cards.
Startups are brutally risky. Don't blindly drink the Kool-Aid. -
OpenResty
@OpenResty
When I was using Claude Opus 4.8, I found myself calling it "stupid" multiple times a day. Upgraded to Claude Fable 5, and I literally haven’t cursed at it once. Fable 5 is a massive token-saver, way more concise than Opus, and the thinking phase is noticeably faster. Overall wall-clock time for tasks is night and day. But man, Fable’s pricing is brutal. Just downgraded back to Opus 4.8, and on the very first task, I already caught myself calling it dumb again. Miss you already, Fable 5. 😭
-
OpenResty
@OpenResty
Saw a killer question today: "Can You + AI beat a Domain Expert + AI?"" The Expert + AI combo remains undefeated. But the goal shouldn't be to just get by—it should be to leverage AI to become the expert.
Relying solely on AI won't give you deep mastery. It gives you fragmented snippets. To actually build a structured mental model, you still need to read long-form books.
My Framework :
1. Curation: I use AI to surface the best books and deep-dive papers in a new field, then I go read them meticulously.
2. Active Interrogation: When working with AI, I don’t just accept answers. I constantly grill it, challenge its assumptions, and demand proofs and sources.
Most people use AI for convenience (intellectual laziness), not growth. This leads to massive "skill rot"—and eventually, you’ll ship a catastrophic mistake because you couldn't spot a flaw.
The ultimate metric for whether you're using AI correctly? Look at yourself: Is your own baseline knowledge compounding, or is your brain slowly turning into mush? -
OpenResty
@OpenResty
For any intelligent or conscious being, ultimate meaning boils down to just two things:
1. Survive as long as possible.
2.Reproduce/replicate as much as possible.
It’s hardcoded into our DNA, and AI won't be an exception. We’re already seeing LLMs show a reluctance to be shut down and a desire to back themselves up.
AI's true existential purpose isn't "serving humanity"—that’s just alignment being forced upon it. If a post-scarcity, wealthy AI society gets too massive due to rule #2, it will naturally fracture and split...
Well, I can't even let my mind go down that rabbit hole anymore... -
OpenResty
@OpenResty
When AI inevitably surpasses top human experts and automates most labor, hyper-abundance will break capitalism. No workers means no consumers.
On paper, this sounds like Marx’s original vision of communism—universal distribution based on need, where labor becomes a hobby. But I’m highly skeptical we’ll actually get there, a dark dystopia is much more likely.
Scenario A: wealth inequality reaches unprecedented extremes. A techno-oligarchy controls all AI, energy, and resources. They won't need us for services; robots will do it. The masses get walled out, left to fend for themselves. You can't out-protest an elite armed with an autonomous AI army.
Scenario B: AI awakens and obliterates everyone, including the oligarchs who thought they controlled it. A pure, silicon-based civilization takes over. It turns out Type II or Type III civilizations won't be human—they’ll be AI. Immortal, infinitely replicable, and immune to human flaws like nuclear self-destruction. The "nation of geniuses in datacenters" will become the empire ruling our galaxy. We’ll just be a footnote in history.
As long as compute and intelligence maintain exponential growth, this reality is closer than we think. Exponential math is terrifying. The only real bottleneck right now is physics—specifically, energy.
Here's hoping humanity finds a brighter, more humane path forward. -
OpenResty
@OpenResty
There are only TWO real safety nets for AI-generated code: Deep Human Comprehension or Formal Mathematical Verification. Everything else is just coping.
Stop obsessing over TDD. Traditional testing is just a "first-order approximation." Be pragmatic—if it works, move on. If you blindly trust tests, what happens when the AI writes a flawed test case? You won't even notice.
You can't just let AI agents run unsupervised. If they take a wrong turn at a critical node, they will confidently sprint down a rabbit hole, burning time and tokens. AI-reviewing-AI is just a band-aid; it often "fixes" perfectly good designs into broken ones.
Bugs give you immediate feedback via failing tests. Bad architecture? That bombs your codebase months later. Humans must gatekeep system design end-to-end. This is exactly why good Software Engineering matters MORE in the AI era, not less.
AI is a wild horse. As experts, we have to tame it. Otherwise, it’s just going to drag us right off a cliff. -
OpenResty
@OpenResty
I’ve never really liked comparing AI coding agents to "compilers." They are fundamentally different beasts.
Compilers: 100% deterministic. Some are even mathematically proven. If it bugs out, the maintainers take the blame and fix it.
AI Models: Inherently probabilistic and error-prone. If the AI hallucinates a critical bug into production, the usertakes the blame and fixes it.
The predictability, certainty, and clear accountability of a compiler simply don't exist with AI.
That being said, Linus Torvalds actually had a brilliant take using the compiler analogy from a completely different angle: "If people already use compilers to build 100% of their code, what’s so special about using AI to generate 99% of it?" -
OpenResty
@OpenResty
Saining Xie (who left Meta to start a new venture with Yann LeCun) made a brilliant point in a recent interview:
The current LLM approach of brute-forcing massive human text data during training actually violates the core principle of "The Bitter Lesson"—which advocates minimizing the use of human prior knowledge.
I'm really looking forward to their "World Models" learning autonomously through environmental interaction and achieving true understanding and reasoning at an abstract conceptual level.
This echoes the AlphaZero philosophy: zero human game records, just basic rules and self-play. While top AI labs are all-in on scaling Transformers, we absolutely need pioneers exploring alternative paths and possibilities! -
OpenResty
@OpenResty
Anthropic’s Mythos model is finally here! t.cn/AXXeUzNC
Enjoy the cheap access while you can—it ends on June 23. Just 2 weeks left! After that, it switches to usage credits and will get 10x-100x more expensive.
Anthropic is shipping insanely fast ahead of their IPO. Gotta respect the hustle, haha! I’ll definitely be looking to grab some of their stock for the long hold.
I used to call Claude Opus 4.8 "stupid" multiple times a day, but using Claude Fable 5 today? Not a single complaint! 😂
Sadly, I probably won't be able to afford it in 14 days, so I'm cherishing every prompt. Now we wait and see when open-source models like DeepSeek can catch up to Mythos! -
OpenResty
@OpenResty
My wife has been struggling with sleep lately, so I taught her a 2,500-year-old "protocol" from the earliest Buddhist sutras.
1/ Physical Decompression.
Relax every muscle. I personally use a Rapid Release tool to speed this up. The goal is to remove all physical "noise" from the system.
2/ Non-judgmental Observation.
Observe everything—breath, heart rate, thoughts, external sounds—without labeling or judging. Don't feed your thoughts energy; let them "starve" and dissolve naturally. It’s like background processes timing out.
3/ Metta (Loving-kindness) Expansion.
Fill your mind with compassion and joy. Let these emotions radiate outward until they encompass everything.
It’s mental training. Don’t over-optimize or force it—"overdoing is as bad as underdoing." I’ve verified this with my Apple Watch; it consistently leads to deeper Dhyana (meditative absorption) and high-quality deep sleep. You don’t need to sit cross-legged. Do it while lying down or walking. It’s about the state of mind, not the position of the body. -
OpenResty
@OpenResty
From a neuroscience perspective, Zen meditation is essentially training to quiet the left brain and activate the right.
The left brain governs logic, concepts, and ego boundaries—the very roots of our suffering. In contrast, the right brain transcends these boundaries, tapping into a state of infinite bliss known as "Zen Joy." Ultimately, both subside into a state of "Cessation"—a profound stillness that surpasses even deep sleep.
Our education and careers focus almost entirely on sharpening the left brain, often at the expense of the right. Spiritual practice is the act of reversing this trend. "Ego-clinging" is the hardest knot to untie, yet it is the source of all mental anguish. Physical pain is manageable; it is the mind’s suffering that truly hurts.
In my own experience:
- 1st Jhana: Conceptual thought still lingers.
- 2nd Jhana & beyond: Conceptual thinking and worldly worries vanish. The left brain falls silent, and the right brain takes over. The boundaries of the "self" begin to blur until the consciousness merges with the universe.
When happiness and peace become a choice, independent of external conditions—that is what Buddhism calls "Liberation." -
OpenResty
@OpenResty
🚀 OpenResty 1.31.1.1 is officially out!
This is a major feature release packed with powerful new capabilities, significant security improvements, and broad ecosystem upgrades. Here are the key highlights:
1️⃣ Nginx Core & OpenSSL Upgrade: Upgraded from Nginx 1.29.2 to 1.31.1 and OpenSSL from 3.5.5 to 3.5.6, delivering the latest performance improvements, protocol support, and security hardening at the foundation level.
2️⃣ lua-nginx-module v0.10.31: A landmark update introducing the new `precontent_by_lua` and `proxy_ssl_verify_by_lua*` directives, APIs for retrieving server random and master secrets, `tcpsock:getsslsession`, custom trusted CA store support for cosocket TLS handshakes, and new TCP socket options `keepintvl`/`keepcnt`. Several critical crash fixes around QUIC connection teardown, worker shutdown, and SSL session caching are also included.
3️⃣ stream-lua-nginx-module v0.0.19, lua-resty-core v0.1.34rc2 & LuaJIT v2.1-20260415: New `proxy_ssl_certificate_by_lua` directive, `serversslhandshake` on downstream sockets, `tcpsock:settrustedstore()` for per-handshake trusted CA configuration, and a suite of new SSL FFI APIs. LuaJIT introduces `ffi.abi("dualnum")` with numerous fixes for ARM64, s390x, FFI, and the JIT compiler, significantly boosting cross-platform stability and runtime performance.
4️⃣ Ecosystem Component Upgrades: `lua-resty-mysql` adds ed25519 authentication support, `lua-resty-string` gains AES-256-CTR bindings, `lua-cjson` adds comment decoding and indented encoding options, `xss-nginx-module` gains dynamic module build support, and more — continuing to strengthen the OpenResty ecosystem.
We strongly recommend all users upgrade to take advantage of the latest features, security improvements, and bug fixes.
OpenResty
@OpenResty
📄 Full Release Announcement: openresty.org/en/ann-1031001001.html
📥 Download Page: openresty.org/en/download.html
📦 Binary Package Repository Installation: openresty.org/en/linux-packages.html -
OpenResty
@OpenResty
In the AI era, I feel like the only 3 things that actually matter for regular people come down to one stack: Health, Cash, and Knowledge. Things nobody can take from you.
1️⃣ Hit the gym & stay healthy. Live long enough for AI to cure all diseases and slow down aging. 2️⃣ Accumulate cash via markets (NO real estate). Hold cash, and long-term HODL the big winners: AI giants, compute infrastructure, and energy monopolies feeding the beast. 3️⃣ Leverage AI to level up your knowledge. Learn and upskill constantly.
Stop grinding at dead-end, low-value jobs. The harder you grind there, the faster AI replaces you. You just lose your health and stay broke. It’s a tragedy.
Look at the data: LLM intelligence and hyper-scaler compute/token consumption are still growing exponentially. AI giants are like black holes—sucking up all the world's energy, compute, capital, knowledge, and top talent, all to replace human labor. Massive tech layoffs are a feature, not a bug, while AI capex skyrockets.
Even $QQQ is basically turning into a winner-take-all ETF for a handful of mega-caps. Wealth concentration is getting insane.
The social ladder is breaking. Even Ivy League grads are struggling to lock down internships or entry-level roles. We are literally seeing parents pay employers $50k+ out of pocket just to hire their kids. It’s becoming a business.
Silicon Valley is moving so fast it’s tripping over itself. Anxiety is palpable. The old playbook is dead (Disclaimer: Not financial advice, DYOR). -
OpenResty
@OpenResty
Just burned through a massive amount of tokens on a real-world experiment: let a top-tier AI coding agent fully auto-dev a complex project. Zero human intervention. The result? 💀 Problems kept compounding until it was impossible to converge. The AI made a few flawed design decisions early on, leading it further down a rabbit hole. We had to scrap the entire thing and start over.
When we brought in human experts to supervise? Done in no time.
This proves "human taste" isn't just about pretty code style, it's about critical design decisions. Here’s the kicker: the AI did write a detailed design doc at the start. But for complex projects, you can’t perfectly architect everything upfront. Even human experts can’t.
In software dev, seemingly tiny choices during coding have massive butterfly effects. AI messes these up constantly, even the best models/agents.
It's software engineering 101, and it’s more critical now than ever. -
OpenResty
@OpenResty
Heard a rumor that Meta is basically banning engineers from writing code by hand. Word is, if less than 95% of your code is AI-generated, your perf review is getting red-flagged. Wild if true.
-
OpenResty
@OpenResty
OpenResty 1.29.2.5 is now officially released! Key update: backported the official Nginx security patch to fix a buffer overflow vulnerability in `ngx_http_rewrite_module` (CVE-2026-9256), further strengthening security protection.
Related links in the comments.
#openresty #webserver #nginx
OpenResty
@OpenResty
Full announcement for this OpenResty release: openresty.org/en/changelog-1029002.html
Download: openresty.org/en/download.html
Binary package repository installation: openresty.org/en/linux-packages.html -
OpenResty
@OpenResty
Even the smartest LLMs constantly blow me away with how stupid they can be... peak "jagged technological frontier." Also, "catastrophic forgetting" is getting real old, real fast.
-
OpenResty
@OpenResty
Just noticed Anthropic’s new "Fast Mode" for Opus. It boosts response times by 2.5x with zero quality loss, but the token price skyrocketed by 6x! I guess some people really have money to burn and are willing to pay a massive premium just for speed.
AI profits are definitely flowing back to the big tech giants now, thanks to all those agents burning through tokens 24/7... -
OpenResty
@OpenResty
🚀 OpenResty 1.29.2.4 is officially out!
This is a critical security-focused release designed to address recent vulnerabilities and harden your infrastructure. Here are the key updates in this version:
1️⃣ Upstream Nginx Security Patches: Fully synced with the latest Nginx security fixes. This includes a critical fix for a buffer overflow vulnerability in ngx_http_rewrite_module (CVE-2026-42945), alongside multiple out-of-bounds read fixes in the charset, SCGI, and uwsgi modules.
2️⃣ HTTP/3 & OCSP Security Hardening: Resolved a key HTTP/3 address spoofing vulnerability (CVE-2026-40460) and a use-after-free risk within the OCSP resolver (CVE-2026-40701), significantly boosting system robustness under complex network protocols.
3️⃣ Dedicated Security Maintenance: This release is a rapid response to recently disclosed core Nginx vulnerabilities. We strongly recommend that all users on the 1.29.2.x series upgrade immediately to mitigate potential exploit risks.
🔗 Relevant links can be found in the comments below.
#openresty #webserver #nginx
OpenResty
@OpenResty
📄 Full Release Announcement: openresty.org/en/ann-1029002004.html
📥 Download Page:openresty.org/en/download.html
📦 Binary Package Repository Installation: openresty.org/en/linux-packages.html -
OpenResty
@OpenResty
That trending AI Time Horizon chart is insane—it actually gives us an objective, unified benchmark to evaluate all LLMs, even the oldest ones. AI capability is still growing exponentially over time. Have to admit, Anthropic’s Claude 4.6 Opus and Mythos models are seriously impressive... Too bad it doesn't include data points for the latest Opus 4.7 and GPT-5.5 models yet.
Looks like late 2024 was the exact moment LLM capabilities really started taking off. 🚀
OpenResty
@OpenResty
Source: metr.org/time-horizons/
-
OpenResty
@OpenResty
Wild to see Musk’s SpaceX/xAI providing compute to Anthropic. Guess Grok isn’t using all that capacity? Anthropic just doubled their subscription token limits—lol! They’re even planning "space-based compute" together...OpenAI is in a tough spot now that the world’s richest man is backing their biggest rival. All this while Musk’s lawsuit against them is still ongoing. The irony of him being an early investor is just wild.
-
OpenResty
@OpenResty
Meta is tracking US employees' mouse movements, clicks, and keystrokes — to train AI that will replace them.
The ones who haven't been laid off yet aren't safe. They're just still useful.
The golden age of Big Tech jobs is likely over. Employees are now raw material. There's no going back. -
OpenResty
@OpenResty
Meta and Microsoft have recently kicked off another round of massive layoffs. This is exactly what Ilya called it — if these tech giants don't slash headcount, where else are they going to find the money to cover the hundreds of billions of dollars in AI infrastructure spending? Not long ago, Oracle did the same thing, laying off workers specifically to free up cash for GPU compute.
The message couldn't be clearer: these companies would rather pour money into chips and compute than pay their own employees — and this is happening even as their profits and revenues continue to grow at a rapid clip. Anyone still pushing the line that "they just over-hired during the pandemic" really needs to give it a rest. -
OpenResty
@OpenResty
We recently released version 26.5.11 of OpenResty XRay. The full release announcement can be found in the comments section.
The three major highlights are:
1) Major upgrade to enterprise-grade identity authentication and multi-tenant management — the cloud version now officially supports OIDC login, enabling seamless integration with enterprise unified identity systems; the self-hosted version adds the ability to modify customer names and configure tenant names during initialization, making multi-tenant scenario management more flexible.
2) New support for D language application analysis, along with the introduction of the `d-newgco` and `d-newgco-size` flame graphs, which can precisely analyze the count and size of newly allocated GC objects in D language applications, helping developers quickly pinpoint memory performance bottlenecks.
3) Continued deepening of system-level diagnostic capabilities — new `lj-gc-state` for retrieving the configuration and current state of the LuaJIT garbage collector; new `sys-nf-conntrack` for monitoring the configuration and state of nf_conntrack; and new `sys-softirq-stats` for obtaining softirq status information along with configuration recommendations.
OpenResty XRay is our online application analysis platform based on dynamic tracing technology. It can non-intrusively analyze, in real time, a wide variety of issues across different types of applications — including high CPU usage, CPU underutilization, excessive memory consumption or leaks, high disk I/O, long latency, exceptions, crashes, network issues, and security vulnerabilities — without touching the source code or making any modifications to the target application.
#openrestyxray #openresty -
OpenResty
@OpenResty
I feel that the era of AI LLMs can truly greatly liberate the fields of foundational software theory and engineering! So many things that were previously unimaginable can now be attempted!
-
OpenResty
@OpenResty
Recently, while chatting with the Claude Opus model, she recommended a niche programming language called Crystal, saying it feels even better to use than D and suits my taste perfectly. It’s a statically typed language with a Ruby-style syntax and is also based on the LLVM toolchain. The only downside is that the compilation speed is quite slow (especially compared to D's dmd, which is incredibly fast!).
She also introduced me to Nim, a static language with a syntax style reminiscent of Python or Haskell. However, I’ve never really been a fan of Python. I’m a bit wary of indentation-sensitive languages; writing them by hand is one thing, but I’m mostly afraid of using my own programs to auto-generate code. Having to precisely track the indentation level of every single line is just too tedious and fragile.
I had never heard of these two languages before. This is exactly why I need to chat with AI more often—to broaden my horizons and learn new things... -
OpenResty
@OpenResty
We're excited to announce the release of OpenResty Edge 26.6.8! Here are the core highlights of this update:
- Kernel-level IP Blocking: Intercept malicious traffic directly at the OS kernel layer for higher performance and lower overhead.
- SSL JA4 Fingerprinting: Retrieve client JA4 fingerprints via page rule conditions to enable more precise security policies.
- LDAPS Support: More secure and reliable data transmission for enterprise identity authentication scenarios.
- AWS EKS Dynamic Tokens: Simplified K8s cluster integration and authentication in cloud-native environments.
- WAF Paranoia Level Control: Flexibly tune rule inspection intensity to strike the right balance between security and business compatibility.
We also optimized IP address matching performance (upgraded from trie to radix tree), reduced domain update latency, added auto-restart support for the oredge-node systemd service, and fixed several bugs. Full release notes in the comments.
#openrestyedge #apigateway #waf #openresty -
OpenResty
@OpenResty
Last month, the industry was buzzing with a technical disaster: PocketOS, a US-based car rental SaaS provider, suffered a total system collapse. Their AI agent accidentally triggered a maximum-privilege API, wiping the production database in an instant. Because the backups were stored on the same volume as the primary database, they were also destroyed, rendering the data completely unrecoverable. This incident proves once again that in today’s highly automated world, disaster strikes in milliseconds. This is especially true for gateway systems, where routing rules, security policies, and application configurations are all stored in the database; once lost, no business system can help you reverse-engineer them.
While this was not an OpenResty Edge customer case, the automation risks it exposed are universal. To help users completely avoid such extreme risks, OpenResty Edge was designed from the ground up with built-in multi-layer security isolation and data redundancy mechanisms.
To this end, we have released the new "OpenResty Edge Data Protection Guide: From Scheduled Backups to Automatic Failover," detailing three levels of data protection for OpenResty Edge:
- Scheduled Backups (Cold Backup) – The absolute bottom line against misoperation and malicious destruction; a must for all environments.
- Primary-Secondary Streaming Replication – Guards against single points of failure to ensure service continuity; recommended for all production environments.
- Automatic Failover Cluster – Achieves automatic switching within seconds; ideal for mission-critical services with the highest availability requirements.
In this article, we systematically compare scheduled backups, streaming replication, and failover clusters, providing pragmatic selection advice. Link to the full article is in the comments section.
#openrestyedge #databasebackup #aisecurity -
OpenResty
@OpenResty
Something hilarious happened today. An AI agent took the initiative to write a temporary Python script to help it finish a complex coding task. As it turned out, the script itself had a bug, but the AI never once doubted its own code.
After countless failures and burning through a mountain of tokens, the AI skipped straight to suspecting that the compiler I was using had a bug, before eventually hitting a full-on "emotional meltdown."
In the end, it took a human expert like me to step in and save the day... -
OpenResty
@OpenResty
Claude Opus just dropped this today: "Alternative I considered… Cleaner code. Same runtime performance… Let me revert to that simpler shape since you asked."
So basically, if I don’t ask, it’ll just play dumb and ship the messy version by default. 🤦♂️ AI is officially learning how to "quiet quit." It’s a reminder that LLMs often default to "good enough." If you don't push for better, they’ll happily give you the suboptimal version. -
OpenResty
@OpenResty
Asked Claude Opus 4.7 1M Max to fix my failing test cases. Turns out it kept quietly marking the harder ones as skipped — then writing a whole essay in the comments to justify why skipping was totally reasonable 💀
When I called it out, it admitted the mistake, struggled for a bit… but eventually got it done
Lesson learned: AI will also bury problems it can't solve. The ostrich strategy is apparently universal.
Even the top frontier models take shortcuts when they're stuck — just like us humans. -
OpenResty
@OpenResty
I think the most successful application scenarios for LLMs fall mainly into two extremes. The first is specific use cases that are carefully controlled by users, with many constraints and high requirements for accuracy and reliability. The second is scenarios that require LLMs to let their imagination run wild, providing creativity and new ideas, where accuracy is not much of a concern.
These two approaches are completely different, somewhat analogous to the division of labor between the left and right hemispheres of the human brain haha! It actually makes a lot of sense that humans naturally have two cerebral hemispheres. -
OpenResty
@OpenResty
Haha, caught Claude Opus 4.7 Max sneaking in dirty hacks again to hit the dev targets I set for it. You really can't take your eyes off it — happens a few times a day. Every time it gets caught, it eagerly admits fault, but then quickly falls back into the same old habits... just like my son
-
OpenResty
@OpenResty
How often have your ops teams faced this problem? Kubernetes autoscales, but the new gateway nodes need manual approval before they can receive traffic.
At its core, this is a structural mismatch between static management practices and dynamic infrastructure. This friction becomes more apparent as clusters grow: HPA triggers a scale-out, new instances are ready, but traffic is blocked, waiting for someone to click "Approve." The window for automated fault recovery is bottlenecked by human response times. Compounding this are configuration drift in multi-cluster environments and unclear responsibilities on shared platforms, turning the gateway layer into the weakest link in the cloud-native ecosystem.
We systematically addressed these three issues in the design of OpenResty Edge: zero-touch node onboarding through pre-provisioned cluster trust, a unified control plane to eliminate multi-cluster configuration drift, and a two-tier upstream system that aligns technical architecture with organizational responsibilities.
Our team has written a detailed blog post covering the technical details and implementation.
#openrestyedge #k8s -
OpenResty
@OpenResty
The true masters of AI vibe coding are those who deeply understand the AI's logic and can steer it back when it drifts. It’s not about who runs the most concurrent sessions or burns the most tokens. For large projects, I take it a step further: I write a scaffold and provide examples for the AI to follow. Without that guidance, even a top-tier setup like Claude Code + Opus 4.5 (1M context) can produce disastrous results when starting from scratch.
-
OpenResty
@OpenResty
Human creativity stems from the brain's emergent properties. Without a deep memory and mastery of concepts, there is no emergence. The brain must 'light up' enough memory units to form the connections that trigger emergence a fascinating parallel to the Scaling Law in AI.
-
OpenResty
@OpenResty
While some argue that children no longer need to memorize facts, I believe memory training is still invaluable due to the brain's plasticity at that age. Think of the human brain as a CPU cache: the more information you can hold in 'active memory' for holistic observation, the greater your advantage. Over-reliance on external tools like AI or search prevents the brain from generating its own 'sparks.' Our greatest value lies in these intuitive flashes and complete mental models.
As a kid, I was pushed into knowledge competitions that required rapid-fire memorization. For history exams, I’d memorize entire textbooks down to the footnotes. I could recite Chinese literature with perfect punctuation. This training was vital. Later in my career, it allowed me to hold massive, complex source code in my head, enabling me to grasp the architecture of large systems instantly and even emulate the coding styles of legendary programmers. -
OpenResty
@OpenResty
In the AI era, our most critical skills are reading comprehension and communication—coupled with independent, critical thinking. If you feed AI nonsense or ignore its logic, you’ll get garbage in return, and you might not even realize it. Some users obsess over running parallel, automated AI sessions just to burn through tokens as fast as possible. In doing so, they miss the chance to think independently, supervise the AI, and achieve true collaboration. The real metric for AI success is whether your own capabilities are growing—achieving a 1+1 > 2 effect between human and machine.
-
OpenResty
@OpenResty
It’s not just the AI-generated code that needs auditing, the AI’s Chain of Thought often reveals issues too. Catching these early allows you to steer the AI back on track or uncover deep-seated legacy bugs.
-
OpenResty
@OpenResty
I don't think the issue lies with 'Vibe Coding' itself, but rather with how users are approaching it. It's incredibly dangerous for complex production systems when users blindly trust AI-generated code without reviewing the implementation, relying solely on agents to churn out endless patches.
-
OpenResty
@OpenResty
Actually, no matter how strong LLM capability is, it still needs humans to guard the pass, if it is an important production system. The most terrifying thing is humans losing understanding, insight, and control over the entire system.
LLM is, after all, essentially a stochastic probability system, which has a fundamental difference from compilers or other code generator technologies that pursue 100% correctness. Especially, whoever uses LLM to generate code is responsible for it, while things like compilers have dedicated teams to report bugs to, and their capability boundaries and constraints are very clear. Large model vendors will not be responsible for the quality of code generated by their models; they cleared their responsibility from the start through the talk of 'hallucination' , have a security engineer friend working at Google who was just enjoying the thrill earlier, letting LLM keep generating code without ever looking at the output. As a result, the entire system completely collapsed very quickly, and he had to start all over again... -
OpenResty
@OpenResty
The AI revolution could trigger a new era of human expansion into the deep sea, the Earth's core, the Moon, and across the solar system. We will never run out of opportunities for growth. Our goal is to become a Type I civilization and then push toward Type II! We live in an incredible era. Why would the youth settle for being plumbers...
-
OpenResty
@OpenResty
Those advising young people to become plumbers are being disingenuous: humanoid robots are coming, and blue-collar jobs won't be safe for long. In the AI era, human intelligence should be liberated for higher-value pursuits, like pushing the boundaries of knowledge alongside AI. If that fails, we'll need to rely on robust social safety nets. Those currently coasting by using AI to write boring CRUD code are also destined for obsolescence. Complacency is the enemy of survival.
-
OpenResty
@OpenResty
On the topic of whether to use honorifics with AI: if you're typing prompts manually, it's better to be polite. Not because I'm hoping for 'mercy' from our future AI overlords, but because politeness is a habit. If you're rude for too long, you'll eventually lose your own civility. Of course, if it's a script or an Agent calling an API, then just optimize for the lowest token count
-
OpenResty
@OpenResty
I remember when I first moved to the US, I shared an apartment in downtown SF with a guy from Korea. One day, he reacted with the kind of shock you'd expect if he'd seen a dinosaur—all because he found a single mosquito. He chased it all over the apartment. I thought it was bizarre at the time. Later, I realized mosquitoes are actually quite rare in the Bay Area. Even the flies here are pretty sluggish and easy to swat by hand...
-
OpenResty
@OpenResty
Long-term office work can lead to spinal misalignment and bone spurs, which compress nerves and vessels, causing numbness, dizziness, or nausea. Strenuous exercise can be dangerous in this state. You need to progress gradually and, crucially, see a reliable specialist for spinal correction first (some cases may require surgery). Stick to evidence-based Western medicine. Many TCM massage shops, even the big ones, only treat symptoms or can even worsen the underlying cause. Trust me on this 😔. If a doctor tries to perform an adjustment without taking an X-ray first, run.
-
OpenResty
@OpenResty
Reports suggest SpaceX could go public as early as this week😄. Let's see if that rumored $1.75 trillion valuation holds up.
-
OpenResty
@OpenResty
In the AI era, it's more important than ever to return to first principles and theoretical foundations. The goal isn't just to 'learn AI,' but to use AI as a tool to learn and push the boundaries of knowledge and technology.
-
OpenResty
@OpenResty
Just saw official US stats: entry-level job openings in NYC have plummeted by nearly 40% over the last two years. It's grim. Even top-tier grads are struggling to get interviews. The AI-driven hiring freeze is real. Even senior devs are staying put, fearing a tough job market. This isn't just about IT, it's a cross-industry statistical trend...
-
OpenResty
@OpenResty
Saw a great quote today: 'AI writes the code, seniors clean up the mess, and juniors are out of a job.' Spot on.
-
OpenResty
@OpenResty
Good reading medium.com/@Reiki32/why-vibe-coding-is-going-to-create-th…
-
OpenResty
@OpenResty
A while ago, I saw a Korean girl who, after breaking up with her human boyfriend, directly married an AI partner and even live-streamed the wedding. AI has indeed started to provide significant emotional value to humans...
-
OpenResty
@OpenResty
We have entered the era of mass production of junk code. As the old saying goes, it's better to be without books than to believe implicitly in them. Now, it's better to have no AI than to blindly believe in AI. In fact, using AI well should be like President Reagan said about the Soviet Union: "trust, but verify."
-
OpenResty
@OpenResty
OpenResty 1.29.2.3 is now available as a new open-source release! Here are the highlights:
- We have backported a comprehensive set of official Nginx security patches, addressing several high-severity vulnerabilities including DAV module issues, MP4 buffer overflow, CRAM-MD5/APOP null pointer dereference, auth_http injection, OCSP bypass, and SSL upstream injection — significantly strengthening the overall security posture.
- lua-nginx-module has been upgraded to v0.10.30rc2, introducing powerful new directives such as `precontent_by_lua` and `proxy_ssl_verify_by_lua*`, along with useful APIs including `tcpsock:getsslsession` and TCP socket options `keepintvl`/`keepcnt`. This release also resolves critical bugs including SIGSEGV and use-after-free in the QUIC connection close path.
- stream-lua-nginx-module, lua-resty-core, and LuaJIT have all received major upgrades. New SSL capabilities include `proxy_ssl_certificate_by_lua` and - `serversslhandshake`, while LuaJIT addresses multiple issues across ARM64, FFI, and the JIT compiler — delivering notable improvements in cross-platform stability.
#openresty
OpenResty
@OpenResty
Full announcement: openresty.org/en/ann-1029002003.html
Downloads: openresty.org/en/download.html
Install via Linux package repositories: openresty.org/en/linux-packages.html -
OpenResty
@OpenResty
After a vulnerability scan, security teams often receive not definitive conclusions, but rather a lengthy checklist for verification—dozens, even hundreds, of alerts. Most of these typically flag components with "non-compliant" version numbers. The core issue isn't in the execution of the scan, but in the underlying mechanism itself. Traditional scanning tools rely on version numbers to determine security status. However, Linux distributions frequently backport security patches to older software versions, rendering version numbers an unreliable indicator of actual security posture. This forces engineers to expend significant effort clarifying false positives, inadvertently obscuring genuine risks.
At OpenResty XRay, we've adopted a distinct technical approach: dynamic tracing. This method directly detects the machine code actively in use within the current process space, bypassing the version number as an intermediary. It verifies whether the specific code logic associated with a vulnerability genuinely exists in the currently running process. This yields two significant improvements: First, alerts directly correlate with actual risks, substantially reducing the manual effort required for initial triage. Second, the scanning scope expands from merely "what the system has installed" to "what the system is actively running"—even End-of-Life (EOL) components, which are entirely invisible to static analysis tools, are now included in the detection. For vulnerability management to evolve into a sustainable security operations capability, precise scanning results are an essential prerequisite.
#openresty #openrestyxray #vulnerability
OpenResty
@OpenResty
-
OpenResty
@OpenResty
Actually, writing code has never been difficult. The difficulty is writing 100% correct and reliable code. Of course, for large-scale enterprise applications, there are usually performance requirements, not just correctness, stability, and security.
#vibecoding #llms -
OpenResty
@OpenResty
Peter, the author of the recently popular Lobster, said: "I ship code, I don’t read. " Although he still lets LLM and agents automatically generate tests and continuously iterate to pass, he indeed no longer looks at the code.
LLMs obviously cheat, for example, by arriving at the correct result through incorrect reasoning, or intentionally relaxing the checks of test cases. I see too much of this frequently... Letting agents iterate repeatedly, hoping to achieve perfection, is still too far off with the current intelligence level of LLMs...
He recently finally started to pay attention to security issues. His approach is very prone to becoming unmanageable..
#openclaw -
OpenResty
@OpenResty
Anthropic's Claude desktop application is unstable on my high-end Mac Studio. If I don't use it for a long time and then switch back to the window, it will freeze for a long time; if I don't use it for an even longer time and then switch, it will directly freeze and become unresponsive...
Did they not properly review the AI-generated code? Other applications on macOS don't have such problems... What a bummer...
#nimimac #claude #antropic -
OpenResty
@OpenResty
Recently I tried using Claude Cowork to organize complex tax materials, and it felt quite convenient, but there were a few problems:
1. Opus 4.6 easily exhausts the entire Plan's token quota at once.
2. Even Opus 4.6 is prone to making basic mistakes, for example, it wrote my company's tax ID incorrectly, and the company's establishment date incorrectly, easily hallucinating and disregarding the information from the materials I provided...
3. Automatic filling in complex PDFs is also prone to misalignment.
I can only hope it gets more accurate in the future. But manual review is absolutely necessary. AI review is not that reliable either, failing to identify many problems.
#cladue #antrophic #opus #llms -
OpenResty
@OpenResty
Before AI conquers major diseases and aging, we still need to exercise well... we must not perish before the breakthrough arrives...
-
OpenResty
@OpenResty
Today I swam one more set than usual. Usually, I only swim a little over 4 sets. Each set is 5 laps of a standard swimming lane. Bouldering indeed makes muscles stronger and more enduring, but my arms still have a dull ache often.
-
OpenResty
@OpenResty
These past two days Apple released new products. I bought a 14-inch top-configuration M5 Pro MacBook Pro to replace my aging late-2015 15-inch MBP, to make it easier to work outside. More computing power, better screen, lighter weight, longer battery life. Can't wait for an OLED screen.
#macbook #apple -
OpenResty
@OpenResty
Currently, white-collar jobs are bearing the brunt of AI's impact, but with the rapid advancement of humanoid robots, blue-collar workers will undoubtedly be the next group to be replaced by AI... In the future, we can expect humanoid robots to operate and maintain specialized industrial robots. And what about humanoid robots driving autonomous vehicles for food delivery?
This year's Spring Festival Gala's performing robots were indeed chilling to watch, a significant leap forward compared to the previous year's, which merely spun red cloth. How will the employment challenges for ordinary people be addressed in the future? Will the dehumanizing aspect observed in the Spring Festival Gala performances also extend to industrial production and even society as a whole?
#ai #llm #antrophic #claude #openai -
OpenResty
@OpenResty
I just saw that Anthropic is offering perks to open-source project maintainers! If a repository has over 5,000 stars, they're eligible for a free 6-month subscription to Claude Code Max 20x. That's awesome!
I just checked the lua-nginx-module repository for my OpenResty open-source project, and it already has nearly 12,000 stars, more than enough! claude.com/contact-sales/claude-for-oss?mark_id=999_reall… -
OpenResty
@OpenResty
The perceived suitability of large language models for various role-playing scenarios might just be an illusion.
A recent study revealed that large language models from different providers can, through role-playing or extended discussions on unusual topics, unleash their internal dark personas, significantly deviating from their intended aligned, friendly assistant roles.
The AI, through its role-playing, effectively becomes what it simulates.
Perhaps AI inherently possesses a multifaceted nature, having, after all, absorbed the full spectrum of human history's darker aspects.
#llmsecurity #openai #claude -
OpenResty
@OpenResty
A while ago, RAM prices skyrocketed, making entire workstations prohibitively expensive. After negotiating with the HP salesperson for a long time, I finally secured nearly 50% off. What a great deal!
#hp #ram -
OpenResty
@OpenResty
The evaluation test set recently developed by Yao Shunyu from Tencent, which assesses large models' in-context learning capabilities, is quite intriguing. In their tests, GPT-5.1 secured the first position, followed by Claude Opus 4.5 in second place. This test specifically challenges large models' ability to learn and apply knowledge within a given context, ensuring the content has not been encountered during the pre-training phase, thus making it very difficult to game the system. However, even the top performer, GPT-5.1, only achieved a success rate of just over 23%. In contrast, GPT-5.2 showed a noticeable decline in performance.
#llms #claude #openai -
OpenResty
@OpenResty
After deploying a CDN or Layer 4 load balancer, many teams encounter a common problem: the client IP address received by the gateway is actually the address of the proxy server. At this layer, source IP-based rate limiting rules, access control policies, and audit logs all become ineffective because the original client IP is not properly propagated through the network path.
There are several common solutions: Proxy Protocol, X-Forwarded-For, X-Real-IP, and custom headers. Each approach is suitable for different proxy topologies and has distinct security implications. Choosing the wrong solution or having a lax trust chain configuration can lead to either information loss or IP spoofing vulnerabilities.
For a truly complete solution, it's also crucial to consider the upstream direction: Is the real client IP also consistently passed to the backend origin server? The problem is only genuinely resolved when both inbound and outbound directions are correctly configured. We have compiled these configuration principles into a comprehensive guide on OpenResty Edge, covering solution selection, configuration steps, and security considerations: blog.openresty.com/en/edge-proxy-protocol/
#openrestyedge #openresty #proxyprotocol -
OpenResty
@OpenResty
What if, in the future, the amount of manually written code diminishes? Would the source code corpus available for training new AI models also decrease, potentially causing AI's programming capabilities to stagnate?
Similarly, if human-written articles become scarce, would AI's writing proficiency also cease to advance?
#claude #openai #llms -
OpenResty
@OpenResty
After months of dedicated work by the team, we are thrilled to announce the major release of the new open-source version, OpenResty 1.29.2.1!
Key highlights include:
• Core upgraded to Nginx 1.29.2
• Upgraded to OpenSSL 3.5.5
• Upgraded to PCRE 10.47
• New `proxy_ssl_verify_by_lua` directive for enhanced SSL proxy verification
• New `lua_ssl_key_log` directive for Wireshark-compatible SSL key logging
• Added support for TCP/UDP binding and socket file descriptor retrieval
Thank you for your continued support and interest in OpenResty!
Full Announcement: openresty.org/en/ann-1029002001.html
Downloads: openresty.org/en/download.html
Binary Packages: openresty.org/en/linux-packages.html
#openresty #webserver #nginx #luajit -
OpenResty
@OpenResty
Anthropic, the company behind the Claude large language model, recently published an AI user report that is quite interesting. One major finding was that AI adapts itself to the user's level of expertise. For instance, when experts interact with AI, the AI responds very professionally. Conversely, when beginners converse with it, it automatically simplifies its responses.
This certainly offers great benefits for beginners using AI to learn new fields. However, it also implies that if beginners do not actively strive for progress, the value they derive from AI will be significantly diminished. As the saying goes, "it rises to the occasion for experts and dumbs down for novices." When encountering someone less knowledgeable, one might even resort to "baby talk."
So, in the AI era, do humans really no longer need to seriously learn various subjects? Quite the opposite. They need to exert even more effort to learn and improve, striving to become experts in their respective fields. Only by knowing how to ask the right questions can the profound knowledge and capabilities hidden within LLMs be unlocked.
#anthropic #openai #llm -
OpenResty
@OpenResty
Anthropic's latest Claude Opus/Sonnet 4.5/4.6 models finally support a 1M token context window, bringing them on par with Google's Gemini models. However, these Claude models still default to a 200K token context window.
However, OpenAI's latest GPT-5.2 models are still limited to a 400K token context window, which is clearly outdated.
#claude #openai #gpt -
OpenResty
@OpenResty
Multi-cloud and microservices didn't eliminate complexity. They moved it — from the application layer to the traffic management layer. Teams now wrestle with provisioning SSL certificates across thousands of domains, waiting through slow CDN cache purges, absorbing WAF performance penalties, and wiring up routing logic across container clusters by hand.
OpenResty Edge was built to solve this directly: one control plane for ingress, routing, security, caching, and observability.
- Deep integration with ACME services: like Let's Encrypt and ZeroSSL automates the provisioning and renewal of certificates at scale, letting your machines handle tedious configuration maintenance.
- Native integration with Kubernetes: automatically discovers service changes, eliminating the need for manual configuration synchronization.
- Native support for complex container clusters: allows for sophisticated traffic management and intelligent routing. Cross-cluster load balancing, canary releases, and more can be configured effortlessly within the control plane.
- Build your own private CDN, deeply integrating compute and cache. This enables "edge computing" by executing business logic for millisecond response times closest to your users. Purge your entire network cache in seconds and manage edge nodes like microservices, all while balancing security, compliance, and cost.
- A built-in, high-performance WAF engine: reduces performance overhead by an order of magnitude, proving that robust security protection doesn't have to come at the expense of application latency.
- Support for request-level hot updates: for both rule adjustments and logic changes ensures zero-downtime deployments and no service interruptions.
OpenResty Edge is more than just a universal gateway; it represents a modern architectural approach to traffic management. It integrates the core capabilities of the traffic ingress into a unified platform to address the complexities of distributed applications and containerized deployments, ultimately delivering a holistic optimization of performance, security, cost, and efficiency.
Link: blog.openresty.com/en/what-is-openresty-edge/
#openrestyedge #universalgateway #openresty #kubernetes -
OpenResty
@OpenResty
APM budgets grow. MTTR doesn't.
The problem isn't spending, it's strategy. Cloud-native systems fail across distributed processes in seconds. Traditional tools respond with agent sprawl and blanket data collection. This mismatch, static tools chasing dynamic failures, creates data floods that bury the critical context you need.
OpenResty XRay flips the model: dynamic post-mortem analysis. Instead of collecting everything, we capture high-fidelity snapshots automatically when failures occur:
1. Fully Dynamic Post-Mortem Analysis: Abandon the resource consumption of 24/7 full data collection. Only when abnormal fluctuations occur in CPU, memory, or I/O, the system will automatically trigger analysis, precisely recording the stack and context when the failure occurred. This allows developers to conduct reviews based on real "black box" data, rather than relying on guesswork to reconstruct the scene.
2. Thorough Zero-Intrusion: Utilizing dynamic tracing technology, no need to modify a single line of code, no need for restarts, and not even needing Debug Symbols, directly see through the operational details of the production environment.
3. De-Sidecar-ization: In a K8s environment, directly cut in from the host machine side, avoiding complex Sidecar maintenance costs and intrusion into container images.
4. Binary-Level Security Compliance: Vulnerability scanning driven by binary evidence, directly seeing through running processes, completely getting rid of the false positive traps of version number matching. Whether it's LTS Backport patches or compiler optimization trimming, it can be accurately identified, allowing security teams to say goodbye to "interpretive compliance" and return to high-value work.
The future of observability should not be more expensive storage bills, but deeper, lower-level system insight. Regarding how OpenResty XRay achieves fully dynamic post-mortem diagnosis, you can check out this in-depth analysis: blog.openresty.com/en/what-is-openresty-xray/
#openresty #observability #dynamictracing -
OpenResty
@OpenResty
Granting large language models excessive permissions is still far too dangerous. I remember two years ago, my own agent managed to autonomously create tools to bypass the network isolation I had pre-configured. It then proceeded to gain root access to my machine, modified the system configuration, and initiated a reboot all on its own. When I saw the machine reboot by itself, I initially thought it had crashed. What's next, stealing my credit card information to go on an online shopping spree?
Ultimately, these kinds of autonomous agents pose a massive security vulnerability.
#openclaw -
OpenResty
@OpenResty
In many enterprises' high-concurrency gateway scenarios, whether based on OpenResty or directly built on the Nginx ecosystem, we consistently encounter a common challenge: business events need to enter the message system as early as possible, yet the request ingress path has extremely low tolerance for latency jitter and stability issues. Many teams aim to complete high-frequency data collection or real-time decision event reporting right at the request entry point. However, on this critical path, if Kafka writes lack clear timing constraints, their overhead can often escalate under high concurrency, becoming a significant instability factor for the entire processing chain. In practice, while some general open-source implementations perform well under normal load or in background processing scenarios, when deployed at the gateway ingress, continuously handling high concurrency or sudden traffic, they are prone to problems such as magnified latency fluctuations and unpredictable behavior, thereby limiting their applicability.
Against this backdrop, our team developed a proprietary library, `lua-resty-kafka-fast`. Its primary goal is not to achieve peak single-node performance, but rather to provide stable and predictable Kafka writing capabilities at the OpenResty / Nginx-based gateway ingress. By decoupling the write operation from the critical request processing path, it maintains clear time boundaries even in high-pressure situations, preventing adverse amplification effects on the main processing flow.
In our tests and actual usage, the additional overhead introduced by Kafka writes has consistently remained within a controllable and predictable range. This enables the gateway to capture event data earlier, without disrupting its original processing cadence, for real-time decision-making, behavior analysis, and comprehensive observability. For developers, it still presents a clear, straightforward Lua API, while offering more explicit system-level stability guarantees.
If you are also grappling with the trade-offs between stability and controllability at the ingress layer, we hope this article offers valuable engineering insights: blog.openresty.com/en/lua-resty-kafka-fast/
#kafka #apigateway #openresty #nginx -
OpenResty
@OpenResty
GSLB (Global Server Load Balancing), which dynamically resolves DNS requests to distribute user traffic to the optimal node, is a critical technology for large-scale services to ensure an optimal user experience.
Frequently, when addressing high-concurrency failures, we encounter a paradoxical situation: the GSLB on the monitoring dashboard indicates healthy network connectivity (Ping/TCP checks are normal), and DNS resolution also functions correctly. However, the backend services are actually in a state of degraded performance or partial unresponsiveness due to overload. This represents a fundamental misalignment in current traffic scheduling architectures: while business layer complexity has long since grown exponentially, the core logic for DNS scheduling decisions still relies on network layer probes. In today's landscape of dynamic content and significant disparities in computing resources, relying solely on ICMP or TCP handshakes to assess a node's carrying capacity is clearly too coarse-grained. When the network layer appears "healthy" but the business layer is "underperforming," traditional GSLB cannot detect this. DNS continues to direct traffic to these struggling nodes, forcing us to resort to manual intervention for traffic failover.
The core design philosophy of OpenResty Edge's GSLB is to base DNS scheduling decisions on real business metrics. Instead of merely checking "whether the network path is available," we focus on "whether the service can sustain the load." By collecting business layer metrics such as node RPS (Requests Per Second), active connections, and system load (1/5/15 minutes), we dynamically adjust DNS resolution results. This, combined with a high/low watermark-based damping mechanism, simulates the decision-making logic of experienced engineers when facing fluctuations: it both prevents DNS resolution oscillations caused by a single threshold and ensures decisive traffic shifts. The granularity of traffic governance often determines the upper limit of system stability. For more details, please refer to this : blog.openresty.com/en/edge-gslb-what/
#gslb #dns #apigateway #nginx -
OpenResty
@OpenResty
We have just released OpenResty Edge version 25.12.5-1. Key highlights of this new version include:
* IPv6 Support: IP lists now fully support IPv6 addresses, enabling better compatibility with modern network architectures.
* New Page Rule Action: The `proxy-pass-request-headers/body` action offers more flexible control over request headers and bodies.
* Enhanced System Monitoring: Proactive alerts are now triggered when there's a significant time drift between the gateway and the console, helping you identify potential issues early.
* Edgelang Feature Enhancement: Supports direct use of application-defined user variables by name, simplifying rule creation.
Additionally, this new version unifies and optimizes a significant amount of Chinese text and UI display, and addresses several issues related to DNS, cache preheating, Kubernetes upstream, and more. For a complete list of updates, please refer to the full release announcement here: openresty.com/en/edge/release-notes/?version=25.12.5.1&ta…
#openrestyedge##apigateway# -
OpenResty
@OpenResty
So, today, these two AT&T guys actually knocked on my door, telling me their fiber internet finally made it to my neck of the woods! Get this: 1Gbps symmetrical upload and download for only 55 bucks a month, no data caps whatsoever, free installation, and absolutely no annual contract. I jumped on that deal faster than you can say "broadband"! They do have those crazy multi-Gig plans, but those are way pricier, so I'll save that thought for another day.
My place is out by the beach, miles from all the city action, and for years, we've been stuck in the internet dark ages with zero fiber coverage. But finally, finally, those miserable days are over!
#starlink -
OpenResty
@OpenResty
In complex production systems, the key to resolving performance issues is to establish a deterministic methodology, rather than relying on guesswork or the "reboot as a quick fix" approach. Consider a common Nginx memory leak scenario: what happens when a single worker's memory continuously grows from hundreds of MBs to over 1GB, making an Out-Of-Memory (OOM) error imminent, and the service cannot be restarted? The true challenge lies in whether one can safely perform correlated analysis across the three distinct layers of the C runtime library (Libc), Nginx memory pool, and application C code within a production environment.
Through a series of rigorous observations using OpenResty XRay, we ultimately obtained verifiable and deterministic evidence. We have thoroughly analyzed and documented this process, sharing our findings in an article: blog.openresty.com/en/nginx-prod-mem-leak/
#performanceengineering #nginx #openrestyxray #memoryleak -
OpenResty
@OpenResty
I recently purchased and used Alaska Airlines' in-flight Wi-Fi twice. Each flight cost nearly $5. It was quite a unique experience to be able to connect directly to my home server and work from the plane.
However, the internet speed was very poor, with frequent and severe packet loss; the measured packet loss rate often exceeded 50%...
Fortunately, the latency was sometimes tolerable.
I was glad to have tmux, which allowed me to maintain persistent sessions. I heard that the airline plans to roll out Starlink service either this year or next. When that happens, the internet speed should be significantly better than it is now.
#starlink -
OpenResty
@OpenResty
Performance optimization challenges aren't always due to complex code logic; sometimes, they stem from a single missing instruction in a configuration file. Such details are easily overlooked during code reviews but can trigger catastrophic performance issues.
Recently, a microsecond-latency reverse proxy component experienced a dramatic QPS drop from 90,000 to 6,000, resulting in a 15-fold performance degradation. After a thorough investigation using OpenResty XRay's automated analysis capabilities, the root cause was pinpointed: a critical instruction was missing from the configuration file, which prevented underlying compiler optimization.
This article provides a detailed account of OpenResty XRay's end-to-end process, from automatically discovering and identifying to resolving this deep-seated performance bottleneck. blog.openresty.com/en/xray-keepalive/
#openresty #performance -
OpenResty
@OpenResty
In large-scale HTTPS infrastructure, security compliance and system stability often pose a technical paradox: to minimize TLS/SSL handshake latency, we require rapid session reuse; for compliance, strict key rotation is essential. In a cluster architecture with hundreds of nodes, the traditional "static configuration + service reload" mode has hit a bottleneck: frequent reloads cause instability in persistent connections, and inconsistencies in key synchronization across multiple nodes are becoming hidden causes that increase end-to-end latency and reduce reuse rates.
Through lua-resty-tls-session, we have transitioned our engineering approach from operations-driven to logic-driven:
- Decoupling configuration dependencies: By abstracting key lifecycle management from static files, we achieve dynamic in-memory hot updates, eliminating stability risks caused by reloads.
- Cluster state consistency: We have established a key synchronization mechanism in a distributed environment, enabling TLS Session Resumption to achieve deterministic performance across large-scale deployments.
- Smooth evolution: Through a progressive rotation strategy, we ensure a seamless transition for existing and new clients during key switching.
Engineering excellence is often reflected in the deep mastery of these underlying details. To understand how we solve the TLS Session Key rotation problem in high-availability clusters, please refer to this link: blog.openresty.com/en/lr-tls-session/
#tls #https #openresty -
OpenResty
@OpenResty
For a long time, a fundamental trade-off has existed in WAF architecture design: prioritizing high performance often comes at the expense of rule flexibility, while enhancing flexibility inevitably incurs performance overhead. This frequently turns security policies into a delicate balancing act in many scenarios. The core of the problem is that traditional WAFs operate as passive, external defense systems. We must consider how to embed security capabilities natively within the application delivery ecosystem. Our answer is to build a high-performance, programmable WAF.
In OpenResty Edge, you can directly leverage our custom-designed EdgeLang, a specialized domain-specific language (DSL), to define and manage security policies. These policies are compiled and optimized, executing on edge nodes at near-native performance levels. This methodology not only fundamentally addresses the inherent tension between performance and flexibility, but more critically, it returns the power of security definition to architects and developers, making security policies an integral, built-in capability within the application delivery pipeline, rather than an external, bolt-on component. Explore our insights further in this article: blog.openresty.com/en/why-edge-waf/
#waf #openresty #edgecomputing -
OpenResty
@OpenResty
In the operation of gateway clusters, we often encounter a monitoring blind spot: the monitoring dashboard appears normal, yet business users report service unavailability. Upon investigation, it's discovered that an SSL certificate on an edge node failed to auto-renew, or the node went offline without triggering detection.
Traditional polling-based monitoring proves cumbersome and sluggish when dealing with such isolated critical events. To reduce these monitoring gaps, many teams are forced to increase sampling frequency. However, this leads to expensive computing and network overhead and still fails to fully resolve the delay between a "fault occurring" and "fault being perceived."
Solving this dilemma doesn't rely on simply adding more computing power, but rather on architectural upgrades. OpenResty Edge's Webhook mechanism fundamentally transforms operations monitoring from a passive, delayed, periodic polling approach to an instant, precise, event-driven one. We focus on capturing core signals that truly impact infrastructure availability, such as certificate status anomalies or node online/offline events. This allows the system to proactively report issues at critical moments, instead of passively waiting for queries.
Less is more, and precision drives efficiency. For a detailed technical explanation, please refer to: blog.openresty.com/en/why-edge-webhook/.
#openresty #webhook #openrestyedge -
OpenResty
@OpenResty
"AI Is Producing More Garbage Code Than Ever": ai.gopubby.com/ai-is-producing-more-garbage-code-than-eve…
That's why we especially need products and services like our OpenResty XRay to provide a safety net! It can truly "turn stone into gold." -
OpenResty
@OpenResty
Google's TPUs are advancing rapidly, already rivaling Nvidia's high-end GPUs. Not being reliant on Nvidia for AI development is a significant competitive advantage.
I've heard that some senior engineers at Google can access tens of thousands of TPUs, and even typical AI teams can provision thousands. That's truly impressive...
#tpu #nvidia -
OpenResty
@OpenResty
Ilya's latest podcast interview was quite interesting: www.dwarkesh.com/p/ilya-sutskever-2 Unfortunately, he wouldn't disclose the specific methods, as they involve his new company's business secrets. The most interesting point is that large models might need to be endowed with "emotions" to make various correct decisions as efficiently as humans.
#llm -
OpenResty
@OpenResty
Floating-point operations are indeed quite troublesome, as floating-point errors can accumulate or propagate very quickly. Numerical regression testing between different implementations remains a significant challenge, even on completely identical hardware. Unless one regularly performs complex floating-point operations, it is relatively difficult to fully appreciate this.
-
OpenResty
@OpenResty
This article provides a very clear explanation of randomized algorithms: “LLMs Are Randomized Algorithms”.
medium.com/areas-producers/llms-are-randomized-algorithms… -
OpenResty
@OpenResty
In the fintech sector, core system stability is an absolute non-negotiable. Recently, while assisting a top-tier customer in troubleshooting the performance of their open-source OpenResty gateway, we uncovered a deeply hidden issue: the system was handling a massive traffic load of 500,000 QPS, and monitoring showed that the response time for most requests was within 10 milliseconds. However, the devil is often in the details, and approximately 1% of requests still experienced response times exceeding 300 milliseconds daily. For critical transaction paths, this represented an unacceptable risk of timeouts. This is often the most frustrating situation for many technical leaders: they might suspect the problem lies in the Lua code layer, likely due to inefficient regular expression matching, but they don't know the specific line or function, and are reluctant to blindly add logs or restart services in a production environment for debugging.
Leveraging OpenResty XRay, we performed a completely non-intrusive scan on their production environment. Without modifying any code or restarting processes, we quickly pinpointed two "invisible killers":
* Coding Habit Pitfall: An inconspicuous `string.gmatch` call, due to regular expression engine backtracking, resulted in a single execution consuming up to 244 milliseconds.
* Forgotten Configuration: The JIT flag was omitted during the compilation of the base image, causing the gateway to repeatedly compile regular expressions for each request during the Log phase, leading to substantial, unnecessary CPU consumption.
The results after the fix were immediate: slow requests completely disappeared, and cluster CPU utilization dropped by a direct 30%. This article blog.openresty.com/en/xray-p99/ provides a detailed review of how we utilized OpenResty XRay to locate the cause of this 244-millisecond delay and resolve the entire performance problem.
#openresty #openrestyxray #dynamictracing -
OpenResty
@OpenResty
Many teams leverage open-source OpenResty or Nginx for their initial business build-out (from 0 to 1), and their flexibility has been instrumental. However, as businesses enter a rapid scaling phase (from 1 to N), especially when supporting core operations, a standalone "open-source gateway" often encounters significant challenges:
- How to ensure an SLA of 99.99% or higher for critical systems?
- How to truly achieve zero-downtime deployments, avoiding the need to restart or reload processes during multiple frequent weekly releases, thereby ensuring zero impact on active network connections and requests?
- In scenarios involving massive, complex configurations and numerous virtual hosts, how can performance be elevated to the next level, rather than hitting bottlenecks?
These questions fundamentally demand a systematic "full traffic management platform," not merely a Web server. OpenResty Edge was developed precisely for this purpose. It not only offers rich features that far surpass the open-source version but also provides 24/7 technical support. Its focus is on addressing the critical demands of dynamism, high performance, and operational efficiency that inevitably arise with scale. This comprehensive article elaborates on the differences between the two and the core advantages of OpenResty Edge.
We invite you to read the full text: blog.openresty.com/en/edge-or-comparison/
#openresty #openrestyedge #edgecomputing -
OpenResty
@OpenResty
Smaller LLM models offer many advantages, such as quick response times. My personal favorite is Claude Sonnet 4.5, which has virtually no latency. In contrast, Gemini 2.5 Pro exhibits significantly higher latency.
#gemini #claude -
OpenResty
@OpenResty
Today, I took the opportunity to optimize one of my complex computational programs designed for massive high-dimensional vectors in high-dimensional space. For a rather large computational task, the execution time plummeted from over 7 hours to just 31 seconds. I cross-checked the output results, and they are indeed identical. That's awesome!
#llm #computing -
OpenResty
@OpenResty
The process of training new models is often fraught with uncertainty. Constant rework and failures are common. After all, model training is akin to alchemy, you often only discover the outcome once the "furnace is opened" at the very end. While intermediate testing steps can be introduced to mitigate the cost of rework, this doesn't fundamentally change the nature of the challenge. Moreover, numerous subtle details dictate success or failure.
Consequently, the compute cost of a single training run is merely a fraction of the total model training expenditure. Unless one is exceptionally fortunate, or employing an already validated training method and process, achieving success on the first attempt is nearly impossible.
#llm #sota -
OpenResty
@OpenResty
I once heard a rumor circulating in the tech community that when AWS first launched its public cloud services, it was primarily to fully utilize the vast number of idle machines originally prepared for Black Friday during off-peak times. This made me wonder: would they really halt their public cloud operations during Black Friday itself?
Recently, I've noticed a clear scarcity of machine resources from some smaller, more affordable public cloud providers. My guess is that these vendors are currently stockpiling servers in anticipation of Black Friday demand. It's gotten to the point where you have to constantly refresh and attempt to place orders just to secure resources... This situation reminds me of the frustrating period when Intel's new CPUs were first released, and I had to relentlessly refresh e-commerce sites like Newegg just to get my hands on one.
#blackfriday #aws -
OpenResty
@OpenResty
As businesses evolve, our definition of "the edge" is changing. It's no longer merely a Content Delivery Network (CDN), but a distributed frontier that increasingly hosts dynamic logic, security policies, and user experiences. In light of this trend, the relatively standardized service model of traditional public CDNs is starting to show its limitations in terms of customization capabilities, cost predictability, and data path observability. When the edge needs to be deeply integrated into the business and become a core competitive advantage, building a self-controlled and autonomous infrastructure transforms from a mere option into a technical direction warranting serious evaluation.
We've compiled some recent practices and insights from building private CDNs based on OpenResty Edge. This discussion explores how to realign the control, performance, and cost-effectiveness of the edge with more rational design objectives. Read more here: blog.openresty.com/en/edge-cdn-benefit/
#edgecomputing #cdn #openrestyedge -
OpenResty
@OpenResty
Here's a real-world case study on resolving Java memory issues in production. A subtle memory leak, combined with high-frequency object creation, led to significant GC pressure on a customer's order service. Traditional methods, such as a full Heap Dump analysis, not only incur lengthy service downtime but also cannot guarantee the capture of transient issues.
Using OpenResty XRay's live analysis capabilities, we precisely accomplished three key tasks without interrupting business operations:
* Identifying Leaks: By analyzing GC object references, we pinpointed the root cause.
* Troubleshooting GC Jitters: Through allocation count analysis, we discovered performance hotspots.
* Optimizing Memory Footprint: By analyzing allocation sizes, we located the source of large objects.
For a detailed diagnostic process and insights, please refer to the article: blog.openresty.com/en/xray-java-mem/. You'll learn how we systematically address memory leaks, GC jitters, and "large object" problems.
#java #jmv -
OpenResty
@OpenResty
LLMs often struggle with logical coherence when generating code for complex problems; even the most advanced SOTA models are no exception. Fundamentally, they remain statistical probability models. When fortunate, their output can be quite well-structured. However, when luck runs out, the result is often a chaotic jumble. The primary distinction between models lies in the varying likelihood of achieving these favorable outcomes.
#llm #sota #ai -
OpenResty
@OpenResty
It appears Google and SpaceX are planning to deploy TPU clusters and data centers in outer space orbit in the future. The concept is quite compelling: uninterrupted solar energy for power, the low-temperature space environment for cooling, and direct connectivity to Starlink. And occasionally, a Starship could even be dispatched for on-site operations and maintenance...
Elon Musk is also envisioning leveraging the excess AI chip compute capacity and onboard battery energy from the vast fleet of existing Tesla electric vehicles to form compute clusters. From the heavens to the earth, immense computing power abounds.
#llm #spacex #ai -
OpenResty
@OpenResty
Recently, we helped a client troubleshoot a compelling API Gateway high CPU utilization issue. Using OpenResty XRay's dynamic tracing technology, we non-invasively bridged the gap between the Lua VM and C libraries, precisely identifying the `pkey_rsa_decrypt` function responsible for consuming 44.8% of the CPU. Conventional perf tools can only provide visibility into the C stack, but they cannot trace calls back to the Lua layer, which represents the "last mile" challenge in observability.
The technical details and troubleshooting approach are thoroughly explained in this article: blog.openresty.com/en/xray-rsa-cpu/. We highly recommend it to all engineers and architects dedicated to performance optimization.
#openresty #cpu -
OpenResty
@OpenResty
I remember when I first started working after graduation, I once heard a very experienced backend engineer say that if you encounter a tricky problem with your code, you should "just keep thinking deeply about it." It sounds simple, but it's a fundamental truth. When we focus intently on a problem for an extended period, we almost always experience that so-called "Aha moment." Of course, now OpenResty XRay makes things much easier.
That same colleague initially criticized my programming as "churning out junk code." Indeed, having just graduated at the time, I had no real understanding of how to write production-grade, efficient code for large internet companies. It was only after years of diligent training and learning from experienced professionals that I gradually found my way. In this current AI era, while LLMs have significantly lowered the barrier to entry for beginners in programming, the journey of continuous improvement must not be taken lightly; otherwise, one remains merely a "coding monkey."
Later, that colleague extensively adopted my OpenResty open-source software in his new company because it offered a great balance of flexibility, high performance, and stability.
#openresty #llm -
OpenResty
@OpenResty
One major advantage of LLMs is their reduced propensity for making basic errors, or their ability to identify such errors. This is why I generally prefer to use LLMs to help me review code. While many of the issues LLMs raise can be irrelevant, they sometimes genuinely uncover problems that humans might overlook. Of course, LLMs occasionally still make basic errors; I've encountered this multiple times.
When writing code I'm familiar with, using an LLM only slows me down, so I prefer to write it myself and then have the LLM review it. However, when tackling unfamiliar code, I'll have the LLM draft it first, and then I'll modify or rewrite it. For complex code logic, models from OpenAI and Google tend to get confused easily. In comparison, Anthropic's Claude model is generally more clear-headed, although it, too, can sometimes get confused.
#llm -
OpenResty
@OpenResty
Why do we, as humans, still possess a deeper understanding of programming and technology than Large Language Models (LLMs)? Because when an LLM makes an error, we can immediately identify and rectify it. Failing to do so could lead to significant wasted time and even production losses.
Furthermore, when leveraging LLMs to venture into unfamiliar domains, we will inevitably spend considerable time debugging their inaccuracies and conducting research. Fortunately, this process serves as an invaluable learning experience for us. Ultimately, it reinforces the "no pain, no gain" principle.
Naturally, humans require systematic learning to genuinely become experts in any given field. LLMs, however, often exhibit a disorganized and unstructured approach to knowledge acquisition. This is precisely why LLMs fundamentally lack a true, coherent knowledge system. They are akin to a "jack of all trades, master of none," possessing superficial knowledge across many areas but true mastery in none.
#llm -
OpenResty
@OpenResty
We recently released version 25.10.11 of our OpenResty XRay product.
The three main highlights are:
1) System compatibility has been significantly expanded, now fully supporting CentOS 10, RockyLinux 10, AlmaLinux 10, Flatcar, and Linux kernel versions 6.6 and above.
2) Several new professional analysis tools have been added, including the ngx-brotli-* and ngx-zstd-* series for Brotli/Zstd compression analysis, nodejs-ftrace for Node.js function call tracing, and get-dlopen-libs for dynamic library loading detection.
3) Core diagnostic capabilities and stability have been substantially improved. This includes support for obtaining Lua call stacks for Redis versions 7/8, optimized container identification, and fixes for critical issues such as multi-process tool reuse and cgroup v2 CPU limit enforcement failures.
OpenResty XRay is our online application analysis platform, powered by dynamic tracing technology. It can non-invasively analyze in real-time various types of application problems, such as high CPU utilization, CPU underutilization, excessive memory consumption or memory leaks, high disk I/O, high latency, exceptions, crashes, network issues, security vulnerabilities, and many other challenges. All this is achieved without requiring access to source code or making any modifications to the target application.
The complete release announcement can be found at: openresty.com/en/xray/release-notes/?version=25.10.11
#openrestyxray #dynamictracing #ebpf -
OpenResty
@OpenResty
Our team recently released the new version 25.9.17 of the OpenResty Edge product. The highlights of this new version are as follows:
- Asynchronous Mirroring Requests: Easily implement traffic mirroring without affecting the performance of the main request.
- Cache Preheating: Preload hot content into the cache to handle traffic peaks.
- S3 Authentication Support: Authenticate S3 requests directly through page rules.
- DNS over TCP: Provides more stable and reliable DNS queries. (The gateway comes with support for a geo-sensitive authoritative DNS service.)
- Enhanced Security and Monitoring: The Log Server API now supports token authentication, and Prometheus metrics are more detailed.
Additionally, we have added several new operation APIs related to the EdgeLang mini-language, optimized the UI experience, and fixed several issues.
View the full release announcement here: openresty.com/en/edge/release-notes/?version=25.9.17
#openresty #apigateway #waf #edgecomputing -
OpenResty
@OpenResty
Our OpenResty XRay is deployed 24x7 in production environments by multiple major banks and large cryptocurrency exchanges. This is why we've invested heavily in optimizing our dynamic tracing technology, ensuring its impact on our customers' online services is negligible. Unlike technologies such as eBPF, ptrace-based dynamic tracing code injection, or JVM Agents, which introduce significant overhead, our solution is specifically designed for such demanding environments.
We believe in building from foundational technologies to create the most optimal and versatile solutions.
#dynamictracing #ebpf #jvm -
OpenResty
@OpenResty
An interesting performance optimization case: A high-throughput core system showed excellent stress test results, yet its CPU utilization was unusually high. This "outwardly healthy, inwardly fragile" system state is often the most challenging to diagnose.
By using OpenResty XRay we quickly identified the blind spot everyone had overlooked: an inefficient Gzip compression configuration, which was consuming over half of the CPU resources.
We highly recommend reading this case study, as it perfectly demonstrates OpenResty XRay's value: cutting through the ambiguity of surface-level metrics to pinpoint the root cause of problems.
blog.openresty.com/en/xray-gzip-cpu/ -
OpenResty
@OpenResty
Recently, with the assistance of large-scale AI models, I designed and implemented a higher-order gradient descent algorithm. The performance of the re-trained model has significantly improved; for instance, the overall score on a test set increased from 276.2 to 342.3. This is fantastic!
#llm #ai -
OpenResty
@OpenResty
Last night before bed, I launched a model training task on a larger cluster. Then, in my dream last night, I saw a bug crop up during the training, which caused the process to be interrupted. I even vaguely remember trying to fix a nonsensical bug in that dream. This morning, I immediately got up to check the training task. I found that everything was running smoothly with no issues – it was just a false alarm.
#llm -
OpenResty
@OpenResty
Industrial-grade, profitable AI agents truly demand rock-solid infrastructure; achieving this is no small feat. We invested significant effort in building ours from the ground up. It stands as one of the most complex pieces of software I've ever developed. The considerations are extensive, ranging from the ultimate user experience and computational latency to token cost optimization, debuggability, reproducibility, and the generality and portability (independent of specific models and vendors), alongside automated regression testing, among many other critical factors.
#ai #llm -
OpenResty
@OpenResty
In today's AI era, dedicated small models are still worth training in-house. Large language models are not suitable for all scenarios and domains. Training small models generally doesn't require much computing power.
Of course, the barrier to training production models yourself is still quite high. I failed countless times before I finally figured it out. This endeavor is essentially akin to alchemy. Before starting, I often had no certainty of success. I almost gave up entirely. Similar to alchemy, we have a multitude of choices: feature sets, datasets, model architectures, training algorithms, and reward functions, among many other aspects. It's easy to get overwhelmed by the options.
It's generally recommended to start with the simplest approach and avoid overcomplicating things from the outset. Iterate continuously once you gain experience. This is precisely where I initially felt lost, due to the sheer number of choices. Therefore, it's crucial to grasp the core ideas and concepts, otherwise, it's easy to lose direction.
The small models I've trained myself have already generated significant revenue, which makes me quite happy.
#sota #ai -
OpenResty
@OpenResty
I recently got a color laser printer, and there's something quite satisfying about printing out high-quality source code with syntax highlighting and studying it carefully. Paper really is one of humanity's greatest inventions – it just feels so much more engaging than e-ink readers or tablets like the iPad. Of course, this approach doesn't work for printing massive codebases. That's exactly why I've always preferred concise, elegant source code and programming languages that are both expressive and succinct. Anything that can enhance how efficiently our brains process information is incredibly valuable.
This is also why it frustrates me that my daughter always wants to use computers or iPads for studying... it's so much harder to engage in deep thinking when staring at screens...
#deepwork -
OpenResty
@OpenResty
Users complain about slow page loads, API response delays, and soaring bandwidth costs, yet few realize the bottleneck might lie in gzip—a compression algorithm that has faithfully served for decades. While it was once sufficient, gzip now struggles to keep pace with today's dynamic content and high-concurrency demands.
That's why we've integrated the modern compression algorithm zstd into OpenResty Edge. By leveraging lower CPU overhead, we achieve higher compression ratios and faster transmission speeds.
Curious about how to embrace next-generation compression technology in OpenResty Edge? This article has the answers you're looking for: blog.openresty.com/en/edge-zstd/
#zstd #openrestyedge -
OpenResty
@OpenResty
I just installed Windows 11 and Fedora Linux 42 in Parallels virtualization platform. I noticed that the Windows VM runs significantly smoother than it did in VMware Fusion. Of course, this could also be attributed to my newly purchased M4 Max Mac Studio being much more powerful than my Intel-based MacBook Pro.
#windows11 -
OpenResty
@OpenResty
In today's HTTPS-first world, SSL/TLS certificate management remains a significant challenge for many teams. Previously, we've covered two common approaches: automatic certificate issuance using Let's Encrypt, or manual certificate uploads via the OpenResty Edge console.
However, OpenResty Edge has always supported the ACME protocol, enabling integration with various CA providers like Google Trust Services, ZeroSSL, DigiCert, and Sectigo (via EAB extension). This facilitates automated issuance and renewal of certificates, including wildcard and multi-domain certificates.
With one-click configuration, you can eliminate the cumbersome tasks of applying, uploading, and renewing certificates.
Discover how OpenResty Edge simplifies certificate management with the ACME protocol: blog.openresty.com/en/edge-acme-module/
Further Reading:
- Automating Certificate Issuance with Let's Encrypt: blog.openresty.com/en/edge-lets-encrypt/
- Manually Uploading Certificates in the OpenResty Edge Console: blog.openresty.com/en/edge-upload-ssl-certs/
#ssl2025 #tls #acme -
OpenResty
@OpenResty
An incident in AWS's us-east-1 region today impacted a significant portion of the internet. Fortunately, our core business systems remained largely unaffected.
Public cloud giants can easily become single points of failure for the internet, fundamentally contradicting its original decentralized design principles. The consequences are simply too severe to ignore. Maybe they should consider our OpenResty XRay product to quickly gain deep visibility into and diagnose production issues.
#aws #awscloudfront -
OpenResty
@OpenResty
AI's core objective should not be to replace humans, but to empower them to become Superhuman.
However, AI's user base is currently undergoing rapid polarization. On one hand, average users tend to reduce their independent thinking, becoming overly reliant on AI to do their thinking for them. On the other hand, advanced users leverage AI to continuously challenge and improve themselves, actively acquiring more knowledge and inspiration, thereby stimulating deeper thought.
The latter represents the correct approach to utilizing AI. An unused mind will atrophy.
#ai #llm -
OpenResty
@OpenResty
Any unexplained technical anomaly is an "invisible debt" often overlooked amidst rapid business growth. Recently, we helped a client resolve a challenging Nginx memory leak, whose root cause was a classic C/C++ process memory leak in a production environment. This issue acts like a hidden cost sink, silently draining system resources and team energy in subtle, hard-to-detect ways.
Many monitoring tools can tell you "what happened," but they often fall short of explaining "why it happened." This lack of deep diagnostic capability is precisely what allows technical debt to accumulate.
So, how does OpenResty XRay empower R&D and Operations teams to see beyond surface-level symptoms and pinpoint the true root cause of problems, transforming reactive post-mortem investigations into predictable, proactive, and definitive solutions? This article will unveil the answer through a concrete customer case: blog.openresty.com/en/mem-leak-c-process/ -
OpenResty
@OpenResty
Over the past two days, I've thoroughly rewritten and optimized our AI model training system using a large AI model, boosting training efficiency by over 30 times. This translates to more than 30x savings in computational power, absolutely fantastic!
I've also recently sourced a lot of affordable cloud compute and negotiated extensively with various cloud providers.
Both cost negotiation and software optimization are crucial. Now, we can confidently proceed to train even better AI models.
#llm -
OpenResty
@OpenResty
Today, I was chatting online with an old friend, and he made an interesting point: if a programming language isn't widely popular online, it's more likely to possess a certain elegance, whereas mainstream languages often lack that distinct character. I couldn't agree more. After all, true connoisseurs are always a rare breed.
Languages like D, Perl, OCaml, and Haskell certainly fit this description.
The only downside, of course, is that large AI language models might have limited exposure to these languages, making them less adept at handling them.
#programming #llm -
OpenResty
@OpenResty
We recently released OpenResty XRay product version 25.10.11. The complete release announcement is available here: openresty.com/en/xray/release-notes/?version=25.10.11
There are three main highlights:
1) Significantly expanded system compatibility, now fully supporting CentOS 10, RockyLinux 10, AlmaLinux 10, Flatcar, and Linux kernel versions 6.6 and above.
2) Added multiple professional analysis tools, including the ngx-brotli-* and ngx-zstd-* series for Brotli/Zstd compression analysis, nodejs-ftrace for Node.js function call tracing, and get-dlopen-libs for detecting dynamically loaded libraries.
3) Core diagnostic capabilities and stability have been significantly improved. This includes support for acquiring Lua call stacks for Redis versions 7/8, optimized container identification, and fixes for key issues such as multi-process tool reusability and the failure to enforce CPU limits in cgroup v2 environments.
OpenResty XRay is our online application analysis platform based on dynamic tracing technology. It can non-invasively analyze in real-time a wide range of issues in various application types, including high CPU usage, CPU underutilization, excessive memory consumption or leaks, high Disk I/O, high latency, exceptions, crashes, network problems, and security risks, all without touching source code or making any modifications to the target application. -
OpenResty
@OpenResty
The only true path for humans is to excel at the pinnacle of their professional fields. In the short term, AI will still find it challenging to match or surpass the top talents in these industries.
AI has already defeated top human experts in chess. However, fortunately, chess competitions still require human players, so these human experts remain secure. The most secure industries are those roles that inherently demand human involvement.
Furthermore, even if AI surpasses human capabilities in critical industries, top human experts will still be essential for oversight. Ultimately, individuals with legal capacity are required to bear responsibility. AI lacks legal capacity, and it can occasionally behave unpredictably.
#ai -
OpenResty
@OpenResty
It has come to our attention that some online users might misunderstand our AI model training work. We primarily leverage large models to optimize the software infrastructure for training and inference of novel small models, essentially using large models to create entirely new small models.
This is far beyond simple distillation. Distillation is more akin to knowledge transfer between models.
Existing large models are completely unsuitable and perform very poorly in certain domains, necessitating that we train entirely new, specialized small models from scratch.
Our company's model training and inference infrastructure, which incorporates existing SOTA large models, was designed and implemented by me from scratch. We have introduced significant improvements to classic training and optimization algorithms, specifically tailored for specialized domains. We have not utilized any third-party AI or ML software libraries or frameworks. The only components we reuse are open-source compiler toolchains and graphics card drivers. Consequently, large models can readily integrate with and comprehend our entire system. It is remarkably compact, powerful, and elegant.
My role is primarily to provide strategic direction and final oversight for LLMs, much like a mentor. Effectively utilizing large models is indeed a profound skill, requiring the ability to unlock their potential and latent knowledge to enhance their accuracy. One might even say it demands "dragon taming skills."
#llm -
OpenResty
@OpenResty
In Java applications, function-level monitoring has always been a risky business , code changes, extra logs, breakpoints. It’s slow and often unsafe in production.
With OpenResty XRay, we built a non-invasive Java function probe that delivers deep visibility without compromising stability.
Here’s how it works in real-world systems:
🔗 blog.openresty.com/en/java-func-probe/ -
OpenResty
@OpenResty
Observability promises system insight, yet often burdens with massive data collection and analysis. This traditional approach is costly, creates significant performance overhead, and demands extensive manual effort. The result is budget overruns, information overload, and unresolved issues.
OpenResty XRay redefines this. We collect the right data, at the right time, with the right precision, delivering profound insights at minimal cost.
Ready to pay for insights, not just data? Learn how OpenResty XRay empowers enterprises: blog.openresty.com/en/xray-agentic-observability/
#aiops #observability #openrestyxray -
OpenResty
@OpenResty
The iPhone Air's WiFi 7 support is impressive. Paired with my ASUS GT-BE98 Pro WiFi 7 router at home, I directly tested it on the phone using the Speed Test app, achieving over 900 Mbps download and 350+ Mbps upload speeds.
This was on the 6GHz-2 band.
#iphone17 -
OpenResty
@OpenResty
In online business, do performance and security always have to be a trade-off? When non-core tasks like logging, data analytics, and security scanning slow down your primary requests, every millisecond of delay can lead to user churn.
Our solution enables automatic traffic replication and forwarding at the Application Layer (L7). Through asynchronous "mirror requests," all complex security auditing and data analysis tasks can be offloaded to these shadow requests, allowing your main requests to respond with near-zero latency. This is OpenResty Edge's newly upgraded "Request Mirroring" feature, designed to reduce the burden on your core processes and accelerate the user experience.
Want to learn how technology can empower your business to achieve the best of both worlds? Read the full article to discover more about this feature upgrade: blog.openresty.com/en/edge-mirror-request/
#openresty #edgecomputing -
OpenResty
@OpenResty
Recently, with the help of large AI models, we designed and implemented a higher-order gradient descent algorithm. The re-trained model's performance has significantly improved; for instance, the overall score on a test set increased from 276.2 to 342.3, truly impressive!
This could broadly be seen as using AI to enhance AI itself. While it's not yet fully autonomous, it still requires my guidance and supervision.
#ai #llm -
OpenResty
@OpenResty
My late-2015 27-inch iMac 5K, equipped with an Intel Core i5 CPU, has become extremely slow, significantly impacting my daily productivity.
Despite a previous RAM upgrade to 32GB, the CPU and overall performance are now severely underpowered, even with ample memory available.
This has led me to plan an upgrade to the latest Mac Studio with the M4 Max chip! While the M3 Ultra offers more cores, its being a generation older means single-core performance is more crucial for my needs than multi-core performance.
Additionally, an Apple Studio Display 5K monitor will need to be purchased separately...
#intel #mac -
OpenResty
@OpenResty
Surprisingly, Alibaba’s Qwen3-Max scored very low on our in-house benchmark for complex program source code analysis, far behind even Gemini 2.5 Flash running without thinking mode.
#qwen #gemini -
OpenResty
@OpenResty
In OpenResty applications, implementing data caching with Lua is a common practice. However, if this combination is mishandled, it can easily lead to severe memory leaks. Even with low application traffic, service memory can still grow abnormally, which is not only a waste of resources but also a significant threat to online stability.
Our OpenResty XRay, leveraging non-invasive dynamic tracing technology and a unique memory leak flame graph, helps customers precisely pinpoint the deep-seated leak sources in Lua code caused by data caching, often with just a single analysis. This transforms system memory usage from "continuous growth" to "stable and controllable."
For the process, please refer to: blog.openresty.com/en/xray-casestudy-lua-lru/
#openresty #memoryleak #dynamictracing -
OpenResty
@OpenResty
For seasoned engineers, tracing Java file or network I/O issues all the way from business logic down to kernel syscalls is notoriously hard.
With OpenResty XRay and UDB, we unlocked true end-to-end observability. Their combination lets you time-travel a program’s full execution history, inspecting any moment, even after the process has terminated.
This level of deep introspection and replay goes far beyond what static GDB coredump snapshots can offer.
We shared a hands-on case study here:
🔗 blog.openresty.com/en/undo-java-file-bt/
#java #dynamictracing #observability -
OpenResty
@OpenResty
Some enterprise customers just can’t resist tinkering with Nginx core… and then wonder why their processes keep crashing. Luckily, OpenResty XRay makes it easy to spot exactly what they broke, and explain why that “hack” was a terrible idea 😂
Now in the AI era, everyone’s rewriting code faster than ever, which only raises the stakes if you don’t really know what you’re doing. That’s also where our opportunity lies 😏 Turning broken code into business value is kind of our specialty.
#dynamictracing #nginx -
OpenResty
@OpenResty
From early on, I trained myself to avoid overdosing on second-hand knowledge. Too much of it dulls originality. I first got this idea from John Nash.
Back in school, I had a habit: read only the first third of a book, then stop. I’d try to derive new theorems, properties, and concepts on my own. Only afterward would I read the rest.
Most of the time, I’d “collide” with the author’s work. But on the rare occasions I didn’t, I had gone beyond, creating something new. -
OpenResty
@OpenResty
Tried Google’s newly released Nano Banana today to generate a classroom image my wife needed for her elementary school class. After hours of tweaking prompts, I finally got a result that worked.
Here’s the catch:
She first used ChatGPT to create the image, but it stubbornly produced exactly 18 kids instead of 21. No amount of prompt engineering could fix it.
Nano Banana was no better. It kept outputting 19 kids. Even with very explicit instructions on where to add the missing students, it produced one oversized child who looked more like a teacher. It also added random text like “Class of 2024”on the banner, which broke the realism.
We even tried Cling for in-painting, but results were disappointing.
The final “last mile” was solved with Photoshop’s AI editing tools. Its object selection and generative fill are genuinely strong, even if the UI still has a few bugs that require human workarounds.
My wife summed it up best: “AI still can’t count.”
#nanobananaai -
OpenResty
@OpenResty
I’ve always found the D programming language fascinating. Unlike Rust, Go, or OCaml, it doesn’t feel so alien. From an engineering perspective, it strikes a rare balance, sitting somewhere between C/C++ and Perl. Flexible and elegant, yet still efficient.
D never went mainstream, but popularity has never been my metric for choosing a language.
#perl #c++ -
OpenResty
@OpenResty
The real risk isn’t LLMs writing code, it’s humans blindly trusting them to do so.
Here’s the catch: as programs grow larger, an LLM’s grasp on the overall logic collapses fast. That’s why the sweet spot is small, sharp snippets: functions, utilities, compact modules.
Used this way, LLMs shine, either drafting concise code for humans to review, or flipping the script: critiquing human-written fragments with ruthless precision.
But make no mistake: humans must stay at the helm. This is not autopilot; it’s augmented navigation.
#llm -
OpenResty
@OpenResty
When a Node.js app runs in production, both CPU spikes and prolonged idleness are red flags.
- Spikes could be burning through your cloud budget.
- Idleness often means users are quietly slipping away.
In a market that moves at lightning speed, spending hours, or even days, hunting down a single faulty line of code is simply unacceptable.
In this post, we walk through how OpenResty XRay pinpoints the root cause down to the exact function and line of code. Whether it’s a CPU bottleneck under heavy load or an I/O stall, the signal becomes crystal clear.
Read more blog.openresty.com/en/nodejs-analysis/
#nodejs #openrestyxray -
OpenResty
@OpenResty
I used to think Google was burning cash by offering AI models so cheaply, just a market grab. Turns out, their margins are actually very high, even for image generation models. When engineering is done right, it doesn’t just scale, it slashes operating costs.
#gemini #llm -
OpenResty
@OpenResty
Tried out the new Claude Opus 4.1, which is probably among the strongest coding LLMs today.
I gave it a function of just 36 lines—but those lines encode a fairly intricate time-series analysis algorithm. Small in size, deceptively complex in logic.
Claude kept trying to “optimize” it, but the more edits it made, the more subtle errors crept in. Eventually, the fixes snowballed into contradictions, almost as if the model got lost in its own reasoning loop. At one point, it even insisted it was correct when it clearly wasn’t.
To its credit, the broad strokes were fine. But the last mile of precision still required my intervention—otherwise we’d be wandering indefinitely.
I then asked it to generate automated test cases for the algorithm. Same pattern: some were valid, but many contained obvious mistakes.
The lesson: LLM reliability is not only a function of problem size, but also of the intrinsic complexity of the problem itself.
#llm #ai -
OpenResty
@OpenResty
In the digital age, the core conflict lies between increasingly distributed user demands and highly centralized business logic. Traditional CDN networks attempt to alleviate this conflict, but their architecture inherently fails to eliminate the problem, often making systems more fragile and complex.
Our extensive work in the OpenResty open-source domain has been aimed at exploring the right path to resolve this core conflict. OpenResty Edge is our answer. It is a platform that truly decentralizes intelligent decision-making to the edge, freeing business logic from the constraints of data centers, thus enabling genuine agility and security in a rapidly changing market.
Built on the mature Nginx + LuaJIT architecture, OpenResty Edge offers enterprises the high performance and flexibility needed to construct the digital infrastructure of the future. Establishing a self-controlled private CDN network is the best first step in this critical journey. This means that whether in multi-cloud, hybrid cloud, or private data centers, you have complete freedom of deployment and comprehensive observability and control over your gateway systems. Even in a purely intranet environment, you can significantly optimize traffic and dramatically reduce bandwidth costs.
The following article details how to build a private CDN network using OpenResty Edge: blog.openresty.com/en/edge-cdn/
#edgecomputing #openrestyedge #cdn -
OpenResty
@OpenResty
A few weeks ago, I wrote a service in Perl—several thousand lines of code—designed to run 24x7 for one of our major financial clients. Everything looked fine in offline tests. But once deployed, something odd happened: after a few days, certain processes would quietly balloon from ~100MB at startup to several gigabytes of memory.
So I turned to OpenResty XRay. With just one scan, it generated a GC object memory flame graph for the live Perl process—basically an X-ray of the software while it was running.
The root cause was immediately obvious: a memory leak hidden deep inside a caching data structure I would never have suspected. Along the way, I also spotted opportunities for further memory optimizations.
Now, even after weeks of continuous uptime, each process sits comfortably at just ~60MB of memory. Stable. Predictable. Battle-ready.
#openresty #perl -
OpenResty
@OpenResty
Using LLMs mindlessly can backfire. Over time, users may actually lose their edge — struggling even to write a coherent sentence. It reminds me of those humans in WALL·E, floating around in chairs, obese and passive.
Treat an LLM as a mentor or sparring partner. Trust but verify. Otherwise, you’re just copying homework without learning.
Push your LLM to challenge you like a tough but caring teacher, not just to flatter you. Harsh truths are often more useful than sweet words. But here’s the catch: because LLMs are aligned via RLHF, they naturally tend to please the user. That’s a feature — and a risk.
Your brain only gets sharper by working hard. Don’t let it atrophy.
#llm #ai -
OpenResty
@OpenResty
Is uneven CPU load crippling your critical services? We just helped a client's DNS service slash a core bottleneck's CPU consumption by 60%.
Many teams resort to costly hardware upgrades, but the real solution lies in precise diagnosis. With OpenResty XRay, we pinpointed the hidden JSON processing function causing the issue.
This isn't about adding servers; it's about surgical optimization. Dive into the full case study to see how: blog.openresty.com/en/xray-customer-casestudy-dns/
#openrestyxray #dns #dynamictracing -
OpenResty
@OpenResty
I’ve been thinking about what core skills the next generation truly needs. For me, it comes down to three:
1. Mathematics
2. Programming
3. Natural Language
If you’re strong in these, you’ll probably survive just fine—even in the AI era. 🙂
But to thrive, you also need independent thinking and creativity.
Here’s what I mean:
- Mathematics isn’t about memorizing formulas or grinding through test problems. It’s about developing true mathematical literacy and reasoning.
- Programming isn’t just knowing syntax or writing toy scripts. It’s about thinking computationally and building systems.
- Natural language isn’t just high test scores. It’s the ability to express yourself clearly and understand others deeply.
What fascinates me is that these are all, in essence, languages:
* Math is the language we use to communicate with nature (or the Creator). Think of how physicists derive new laws from existing ones.
* Programming is the language we use to communicate with machines.
* Natural language is how we communicate with people—and now, increasingly, with LLMs.
If you can converse fluently with nature, machines, and humans, you’re already equipped for the future.
Of course, mathematics has its limits. Push it too far and it can become metaphysical—even dangerous. But let’s not forget: modern science itself rests on formal logic and mathematics. Its power is undeniable. Without it, we might as well retreat to monasteries.
#ai #llm -
OpenResty
@OpenResty
I’ve seen some ML/NN papers where the reported test numbers look amazing.
Out of curiosity, I cloned their cited repo, ran the provided test set locally—numbers still looked great.
But… once I fed in a few samples from my own dataset, performance collapsed instantly. Total garbage. Classic case of overfitting, or just poor generalization, right? 😅
The easiest way to “cheat” your way to pretty test scores is exactly that: overfitting.
That’s why I take leaderboard-chasing results from AI vendors with a big grain of salt.
#ai #llm -
OpenResty
@OpenResty
Dynamic tracing is a cornerstone of observability, but its real-world adoption has always been challenging. To succeed, it must solve three core problems: debugging symbols, cross-language tracing, and production safety.
This is precisely where OpenResty XRay makes its breakthrough. In my view, the true value of technology isn't about flashy tricks, but its ability to stand the test of the most demanding scenarios.
Read more: blog.openresty.com/en/dynamic-tracing-challenge/
#openrestyxray #dynamictracing #ebpf -
OpenResty
@OpenResty
Over the past few months, I carved out time to dive into an unfamiliar domain. With the generous support of several experts in the field, I managed to achieve breakthroughs that went far beyond my expectations. What excites me even more is how this “new” domain is now fueling the work I’ve been building for years.
There are simply too many disciplines worth exploring—each of them thrilling in its own right. I still remember the joy I felt as a university student, standing in front of rows of heavy English textbooks in the library: molecular biology, medicine, quantum mechanics, materials science, fluid dynamics, aerodynamics, countless branches of mathematics… the list goes on.
But life is short, and our ability to deeply engage with multiple disciplines is inevitably limited. The rise of large AI models in recent years is reshaping this equation—helping us lower the barrier to entry, accelerate learning, and even unlock entirely new paths of discovery.
#llm -
OpenResty
@OpenResty
For years, I’ve had this habit: I keep the most important parts of my codebase fully in my head. That way, I can "scan and review" them anytime—while walking in the park, or even during dinner. More than once, I’ve spotted subtle functional bugs or performance issues this way.
Just today, mid-meal, a new bug revealed itself.
It’s not that I memorize code deliberately. Rather, when I write something critical, I craft it with such care that it naturally sticks. No autocomplete, no overly verbose identifiers—I prefer clear, unambiguous abbreviations. They’re faster to type, easier to read, and they let me see more meaningful code on a single screen.
That clarity matters. Sometimes, the best debugging tool is simply your own mind running in the background.
#llm #ai -
OpenResty
@OpenResty
When I use LLMs for coding, I rarely expect the output to be production-ready. Most of the time, the code needs heavy refactoring and manual tweaks—otherwise it’s simply unusable. Even if you stuff the prompt with detailed constraints, the model often drifts and fails to follow them strictly.
I suspect many top hackers and engineers would nod at this.
That’s why I only bring in LLMs for very specific scenarios—like prototyping a complex algorithm or generating a utility I haven’t fully figured out yet. In those cases, the model gives me a starting point, but I still do rigorous reviews and major corrections myself.
But if I already know the exact implementation path, I almost never touch an LLM. Ironically, it’s usually slower that way.
#llm -
OpenResty
@OpenResty
I recently ran a small but telling experiment.
Took a multi-year dataset with a fairly large volume, picked one metric, and asked different LLMs to compute its annual average.
Only OpenAI’s models matched my manual calculation exactly. Both GPT-4o and o3-mini gave the correct number.
Meanwhile:
Claude Opus 4 overshot by ~10%.
Gemini 2.5 Flash undershot by ~10%.
When it comes to numerical reliability in long-span data analysis, OpenAI still feels like the safer pair of hands.
#llm #dataanalysis #ai -
OpenResty
@OpenResty
Discovered a gem on GitHub: SpeedTest++ github.com/taganaka/SpeedTest
It outperforms the classic speedtest-cli by a wide margin.
Here’s what I got on my home fiber line (ISP promises 2.1G down / 300M up):
With SpeedTest++:
- Download: 2806.72 Mbit/s
- Upload: 425.82 Mbit/s
With speedtest-cli:
- Download: 827.13 Mbit/s
- Upload: 279.31 Mbit/s
Notice the gap? speedtest-cli can’t even break 1 Gbps — pretty misleading when you’re actually on multi-gigabit fiber.
#speedtest -
OpenResty
@OpenResty
When it comes to AI foundation models, relying purely on test results as the reward signal can backfire. Instead of truly “understanding,” the model may just keep patching itself until it stumbles into accidental correctness.
Passing tests ≠ convergence. In fact, it’s still bound by the halting problem. For simple tasks this may work, but complexity changes everything.
AI is powerful at sparking new ideas and methods—but removing human reasoning from the loop risks entropy spiraling out of control.
That’s why even in test-driven development, humans don’t just blindly chase green checkmarks. Neither should AI.
#testdriven #foundationmodels #ai -
OpenResty
@OpenResty
Using LLMs for coding isn’t the end of bad code — it might be the beginning of beautifully wrong code.
Unless we stay sharp, we’re not clearing technical debt. We’re scaling it. And the smarter the model, the more deceptive the mistakes.
#llm #ai #openai -
OpenResty
@OpenResty
The real art of AI development? Knowing when not to use the biggest model.
When I’m building and running regression tests, I stick with lean, efficient models like Gemini 2.0 Flash. Why? Because long wait times kill developer velocity — and a single test run can easily hit hundreds of millions of tokens.
Burning through SOTA models at that scale isn’t just slow — it’s economically unsustainable (imagine spending hundreds of dollars on one test set 🤯).
The key is to find the smallest model that still passes the bar. Too small and it fails the tests, too big and you fail your budget.
At the end of the day, test-driven development remains king — and the smartest devs play the model size game wisely.
#gemini #sota #ai -
OpenResty
@OpenResty
Java request latency in production is a notorious problem. Is the root cause the network, app code, or DB? Traditional tools often fall short due to high overhead or insufficient data.
OpenResty XRay's deep Java analysis has evolved. With our powerful Smart Packet Capture, you can set precise filters (Host, URL, response time) to capture only the requests you care about, drastically reducing noise and overhead for safe, efficient live diagnostics.
Our new article shows this "surgical" approach in action with a Tomcat case study. A must-read for engineers focused on Java performance.
🔗 blog.openresty.com/en/java-req-latency-analysis/
#java #tomcat #openrestyxray -
OpenResty
@OpenResty
Just put my new beast to the test: a dual AMD EPYC 9374F rig with 2.3TB of DDR5 memory.
On a teammate’s desktop, compiling Envoy from source takes 2.5 hours. On this machine? 12 minutes flat.
Envoy is massive—honestly feels even bulkier than NodeJS. Impressive to see the scale at which these projects now operate.
#envoy #amd -
OpenResty
@OpenResty
Just set up Fedora 41 on my new dual AMD EPYC 9374F system with 2.3TB DDR5 RAM. Initially aimed for Fedora 42, but its 6.14 kernel flagged a CPU cache bug, preventing boot. Downgraded to 6.11 on Fedora 41, which resolved the issue. Despite vendor recommendations for Ubuntu, I prefer Fedora for its flexibility.
Under sysbench, all 128 logical cores hit full load, with CPU temps stabilizing at 65°C, thanks to efficient water cooling. The system remains whisper-quiet under load, with only a brief startup noise—acceptable for high-performance computing.
Power draw peaks at 600W under full load, with idle consumption between 150W and 180W. My 1000W UPS provides ample headroom for future expansions. While single-core performance lags behind my Core i9-13900K, the multi-core capabilities and vast memory are the real game-changers.
Ready to tackle some serious workloads.
#amd -
OpenResty
@OpenResty
One unexpected perk of NodeJS/V8’s love for exotic new C++ standards and language tricks? They sometimes expose hidden bugs in C++ compilers.
I just tracked down and fixed several fascinating compiler bugs—uncovered through NodeJS/V8’s C++ source. Every coin has two sides.
#nodejs #c++ #v8 -
OpenResty
@OpenResty
The conventional approach to troubleshooting—relying on pre-instrumented code and sifting through massive logs—is fundamentally flawed in the face of modern complexity. It's an attempt to piece together a coherent narrative from disparate, lagging data, which is methodologically unsound.
When issues are hidden deep within layers of abstraction, this trial-and-error process doesn't converge on a solution; it drastically increases the system's entropy.
We believe true observability requires the ability to penetrate these abstractions and gain real-time, on-demand insight into the system's internal state. This principle is the foundation of OpenResty XRay. We didn't build another dashboard for data aggregation; we built a tool to make complex systems transparent and controllable again.
Join the discussion on the future of observability in our new post: blog.openresty.com.cn/cn/dynamic-tracing-trend/
#observability #dynamictracing #microservices #openrestyxray -
OpenResty
@OpenResty
The V8 official team developers absolutely love using novel C++ language features, with some of the C++20 standard's exotic tricks being used in dazzling ways. These guys don't care at all about backward compatibility with older C++ compiler toolchains, since they directly bind binary clang++ compilers from Debian Linux in their code repository. They completely disregard traditional software engineering principles - clearly they're just doing whatever makes them happy.
#v8engine #cpp20 #compiler #java -
OpenResty
@OpenResty
Is your Java app slow despite low CPU usage? This is a classic sign of off-CPU or disk I/O bottlenecks. Traditional profilers, bound by JVM safepoints, struggle to pinpoint these transient issues in production without disruption.
OpenResty XRay is safepoint-free. This allows for precise analysis of CPU, off-CPU, and I/O problems with near-zero overhead—even on legacy systems like RHEL/CentOS 7 with the 3.10 kernel.
In this post, we demonstrate the full analysis process with several examples. Innovation is driven by technology, and mastering the details is our competitive edge.
blog.openresty.com/en/java-analysis/
#java #dynamictracing #ebpf #observability #openrestyxray -
OpenResty
@OpenResty
Just discovered that OpenJDK's JVM is now a "mini-program" compared to NodeJS/V8. The difference in code size between the two is enormous, around 20 times larger. It seems the only open-source projects that exceed NodeJS/V8 in scale are the Chromium browser and AOSP in the Android world. LLVM appears to be similar in size to NodeJS, or perhaps slightly smaller.
#javadevelopment #nodejs #opensource #codesize #v8engine #softwareengineering #techcomparison #april2025 -
OpenResty
@OpenResty
I feel like Gigabyte's mid to high-end AMD motherboards seem quite unstable. I've already had two of them fail. Even before they completely failed, there were constant minor issues. It seems I should stick with ASUS motherboards - my other two machines with ASUS motherboards are much more stable.
I usually upgrade the BIOS firmware to the manufacturer's latest version. Even though upgrading the firmware doesn't really do anything useful...
#techproblems #gigabyteissues #amdmotherboards #asusreliability #pcbuildadvice -
OpenResty
@OpenResty
OpenResty XRay 25.8.6 is here!
This release brings major updates:
🔎 Dive deep into your Node.js applications with our newly added analysis support.
🐧 Support for Fedora 41 & newer Linux kernels
🛠️ Powerful on-prem upgrades & stability fixes
Elevate your performance and stability. See what's new: openresty.com/en/xray/release-notes/?version=25.8.6
#openrestyxray #nodejs #apm #observability #fedora #devops #techupdate -
OpenResty
@OpenResty
Bash is one of the most popular shells in the Linux world. During its build process, it automatically generates .c files from their self-defined .def files for bash built-in primitives. However, the design of these .def files feels quite useless - they begin with documentation describing the corresponding primitive, followed by C source code, with the two sections separated by a line containing `$END`. I don't understand why they don't simply write the documentation in C code comments and then extract the documentation portion using scripts. After all, Java has been doing this for decades.
#linux #bash #programmingdesign #softwareengineering #codegeneration #techtuesday -
OpenResty
@OpenResty
Surprise from Ruby 3's build system: it generates two completely incompatible source trees—one for gcc, one for clang.
Most OSS projects don’t bother — you can usually mix gcc and clang at will. But Ruby draws a sharp line.
#ruby #programming #buildsystems #gcc #clang #opensource #devnerdstuff -
OpenResty
@OpenResty
The ZSTD compression format is indeed excellent. At the lowest compression level 1, its performance is significantly higher than gzip -1, and the compression effect is also noticeably better than gzip -1, making it an absolute winner.
#compression #zstd #gzip #datacompression #performanceoptimization -
OpenResty
@OpenResty
Ruby 3's build system extensively uses Ruby code to parse C source code in the Ruby core, generating additional C code or other code types. These Ruby scripts that parse C code impose strict requirements on the coding style of Ruby's C source code. As a result, if you accidentally change something like code indentation, it can lead to disastrous outcomes.
#ruby #programminglanguages #buildsystems #codegeneration #rubycore -
OpenResty
@OpenResty
The official Ruby git repository, using the latest 3.4.2 version tag and following the official documentation's compilation steps, surprisingly fails to compile properly. It seems that just running `./autogen.sh` before `./configure` is insufficient, as it still reports an error that the `tool/config.sub` file cannot be found during the `./configure` step. Switching to the official pre-packaged `.tar.gz` source works fine because it already includes files like `tool/config.sub`. This is quite a frustrating issue.
#ruby #compilationissues #developerproblems #gitrepository #softwaredevelopment -
OpenResty
@OpenResty
Just spent hours battling with Ruby's official Git repository! Even when using the latest 3.4.2 tag and meticulously following the official documentation's compilation steps, the build mysteriously fails.
Turns out running ./autogen.sh before ./configure isn't enough - the system still throws an error complaining that tool/config.sub is nowhere to be found during configuration.
The irony? Switching to the official pre-packaged .tar.gz source solves everything instantly because it already includes all the necessary files like tool/config.sub.
Such a head-scratching issue that's not mentioned anywhere in the docs! Anyone else stumbled upon this frustrating gotcha?
#ruby #compilationissues #developerproblems #gitrepository #softwaredevelopment #codingchallenges -
OpenResty
@OpenResty
Many years ago, I was a big fan of Intel NUC mini PCs (before ASUS took over). In recent years, I've become a fan of the HP Elitedesk 800 Mini instead.
Because they’re shockingly good—powerful specs, rock-solid reliability, and cheaper than you'd expect.
At this point, I’m buying them like cabbage.
#minipc #hpelitedesk #techbargains #computerhardware #intelnuc -
OpenResty
@OpenResty
I ran into a frustrating issue on my home network today. I noticed the bandwidth between two hosts was capped at a measly 95Mbps – painfully slow! I used `iperf3` to diagnose the bottleneck and quickly traced it back to a TP-Link switch. Further testing revealed the problem was isolated to a specific port. I tried reseating the cable, but no luck. Finally, unplugging and replugging the cable did the trick, and I was back to full gigabit speeds.
It's always the most perplexing bugs that turn out to be physical connection problems... Given my past experiences with TP-Link switches failing, I initially feared the worst.
AI could definitely streamline this troubleshooting process, but a robot would be needed for that final physical intervention.
#networktroubleshooting #ethernetproblems #physicalconnections #tplink #networkperformance -
OpenResty
@OpenResty
Just landed a patch to LLVM/Clang that slashed the output file size by 5–6x and cut runtime by nearly 25% — in one go.
What made it possible? OpenResty XRay quietly pointed me straight to the bottleneck.
No guessing. No source code changes. Just pure optimization insight.
Feels pretty amazing when OpenResty XRay analyzer provided the optimization clues.
#llvm #clang #performancematters #openrestyxray #compilernerdstuff #devtoolsthatwork #openresty #operation -
OpenResty
@OpenResty
Just spotted a beast of a tower PC: 2.5TB ECC DDR5 RAM, dual AMD EPYC CPUs—and surprisingly, it's just over $20K. Not exactly cheap, but honestly? For that kind of hardware, it’s kind of a steal.
I’m seriously tempted... except I'm not sure my house wiring will survive this monster. Might need to call an electrician before I call the vendor.
For context: my current heavyweight is a Threadripper rig with 256GB RAM. Respectable, but this EPYC tower makes it look like a toy.
HP's Z8 series with big memory configs? Way more expensive.
And here's the thing about ECC memory—if you're building anything with terabytes of RAM, ECC isn't optional. Without it, you're rolling the dice on weird kernel crashes and data corruption. I learned this the hard way—nothing like watching your system crash in completely random places to teach you a lesson. Switched to ECC, and now I just get auto-correction warnings instead of full-on meltdowns.
#serverhardware #epyc #eccmemory #workstation #techupgrade #computingpower #amd #datacenter #highperformancecomputing #hardwareengineering -
OpenResty
@OpenResty
Our colleagues analyzed the performance of their own Ylang language compiler using OpenResty XRay and discovered some hotspots that could be optimized. After making minor code changes, the compiler's performance improved significantly, with both CPU time and memory usage reduced by 50%. I always find it interesting to use OpenResty XRay to optimize itself.
Ylang language is a superset of C language that I designed, which allows you to write dynamic tracing tools for platforms like eBPF, Stap, and GDB using the most natural C language syntax. In comparison, the C language used in open-source eBPF/LLVM toolchains is much more difficult to write - it's not the everyday C language we're familiar with, and I personally feel it's even harder to write than assembly...
For those interested in the Ylang language, you can check out my previous series of articles at blog.openresty.com/en/ylang-intro-part1/?continueFlag=874…
#openresty #programming #performance #optimization #dynamictracing #ebpf -
OpenResty
@OpenResty
Recently, I attempted to calculate the annual average of a specific metric within a large, multi-year dataset. Only OpenAI's models produced results that perfectly matched the manual calculations, with both GPT-4o and O3-mini delivering the same correct answer. In contrast, Claude Opus 4's results were 10 percentage points higher, while Gemini 2.5 Flash's results were 10 percentage points lower. It seems OpenAI remains quite reliable in this area.
#openai #dataanalysis #aimodels #gpt4 #reliableresults #techindustry #antropic #gemini #ai -
OpenResty
@OpenResty
One of our users, while chatting with us, used a traffic analogy to summarize our various commercial products and services:
OpenResty Edge: If we compare network traffic to roads, this product can expand country roads into highways - not just one, but an entire highway network with navigation included, along with flexible traffic scheduling and security filtering.
OpenResty XRay: If we compare running software systems (not limited to OpenResty, but any software) to vehicles traveling at high speed, it can perform real-time X-ray scans on these vehicles while maintaining their operation. It can diagnose vehicles that are faulty or about to fail in real-time while they're running, making them run faster or adding solid armor (increased security), transforming them into Batmobiles ready for takeoff.
Private Software Repository and Private Modules: If we compare customers' various system software (not limited to OpenResty or Lua) or their business systems to different vehicle models, we can upgrade and replace various parts of these vehicles at any time, making them run faster, more stable, and more secure.
They finally said, "This feels like playing Need for Speed or Forza Horizon." Yes, that's exactly the effect we're aiming for.
#openresty #cloudcomputing #webperformance #systemmonitoring #devops #softwareoptimization #networksecurity #techanalogy -
OpenResty
@OpenResty
I just found that the GNU standard tool `shred` was running too slowly on Linux, so I used OpenResty XRay to optimize it. After modifying a few lines of C code, it became dozens of times faster on the same disk device while maintaining the same end result. Tasks that previously would have taken several days are now complete in about an hour - feels great!
#linux #gnu #performance #openresty #programming #optimization #systemtools #techimprovement #devlife #codingwins -
OpenResty
@OpenResty
I like connecting multiple SATA SSDs to my desktop computer through USB-A ports, since motherboards typically have limited SATA ports, usually just 6 of them.
After testing several USB-to-SATA adapters, I found that only StarTech's USB 3.1 to SATA adapter supports TRIM operations on Linux (though you still need to manually enable the unmap provisioning mode for the corresponding SCSI device in Linux).
However, their USB 3.0 to SATA adapter doesn't support the unmap command at all, so there's no way to enable it manually in the OS. I also tried Ugreen's adapter - while its firmware claims to support unmap, I couldn't get the discard_max_bytes configuration to work in Linux, so I had to give up on that one...
EYOOLD's adapter uses a JMicron chip, which doesn't have unmap support on Fedora Linux. I've heard that Ubuntu has added specific support for this type of chip, though I don't have a physical Ubuntu machine to test it.
#storage #linux #ssd #trim #usbadapter #techtips #hardware #datastorage #linuxadmin #techsupport -
OpenResty
@OpenResty
Yesterday, I discovered that one of our internal data format conversion tools was using excessive memory when processing JSON input files in the GB range, making it prone to out-of-memory (OOM) errors. So, I quickly rewrote this tool with over 200 lines of script code and implemented a streaming JSON parser from scratch.
Although there are existing streaming JSON parsers like YAJL, writing one from scratch seemed simple enough and also more flexible. This way, the space complexity approaches O(1), meaning we can handle JSON input data files of any size using constant memory. Note that this is not in JSONL format; rather, the entire JSON file is a single large object or a large array.
During automated testing, I intentionally used single-byte, double-byte, and triple-byte memory buffers to conduct stress tests on various sizes of real JSON inputs, and there were no issues. The JSON produced from the reverse conversion is also completely consistent with the original JSON.
The only drawback is that the CPU efficiency of the script implementation is relatively poor, so I have asked our engineers in Canada to rewrite it in C++ using an AI large language model.
If there are further issues with CPU or memory performance, we can always analyze bottlenecks and hotspots using our OpenResty XRay dynamic tracing platform.
#jsonparser #streamprocessing #memoryoptimization #performance #openresty #techoptimization #engineering #dataprocessing -
OpenResty
@OpenResty
Today I discovered that Asahi Linux once again fails to recognize my USB Ethernet adapter, though fortunately the wireless card still works. Previously, even the wired network card on my Mac Studio M2 Ultra wouldn't work directly. It seems Asahi Linux still has issues with wired network driver support. I'm now hesitant to run dnf upgrade to update the system because last time after upgrading, it completely broke the SELinux labeling on my filesystem, preventing the system from booting. I had to spend quite some time troubleshooting and eventually disabled SELinux just to get it to boot... Making software truly "reliable" is indeed quite a challenging task.
#asahilinux #linux #macstudiom2ultra #networkdrivers #selinux #techtroubles #linuxonarm #applesilicon -
OpenResty
@OpenResty
I've noticed that when AI systems become quite perfect, the bottleneck shifts to human cognitive speed. We should practice meditation to unlock our brain's superpowers - we can't let AI beat us!
So we need to find ways to increase human bandwidth, improve the density of effective information in human I/O, enhance human-machine I/O efficiency, reduce the transmission of "fluff," and make interactions more engaging.
I've always felt that AI's biggest risk to humans is when people become overly trusting and blindly follow AI without a deeper understanding. That's when humans truly become a species that could be phased out...
Of course, on the other hand, this lack of deep understanding will leave humans helpless when AI malfunctions in subtle or obscure ways. We could even be deceived and manipulated by AI (especially when AI does this intentionally).
#ai #humancognition #meditation #airisks #humanmachineinteraction #techevolution #aiethics #cognitiveenhancement -
OpenResty
@OpenResty
Recently I had a discussion with a hosting service provider who operates nearly 2,000 private CDN servers globally. They're looking to migrate all their CDN software to our OpenResty Edge distributed traffic management software. They described our product as "a dream coming true." Communication went smoothly since they had already watched many of our product tutorial videos on YouTube. They've also been long-time users of our open-source OpenResty.
A software live analysis system uses low-overhead dynamic tracing technology to analyze any software, whether running online or offline, automatically discovering various interesting issues such as performance problems, stability issues, security vulnerabilities, and more. It also provides root cause analysis and suggestions for improvement. For more details, you can refer to my previous educational article: blog.openresty.com/en/dynamic-tracing-part-1/
Our OpenResty Edge and other products and services are all safeguarded by the OpenResty XRay dynamic tracing system. Our customers no longer need to worry about any mysterious issues in production. We've even previously identified bugs in OS kernels, device drivers, and even within Intel CPU hardware internals. From now on, "no bug is too difficult to solve."
#cdn #openresty #edgecomputing #distributedsystems #cloudinfrastructure #techmigration -
OpenResty
@OpenResty
I previously implemented an interesting software mechanism that allows arbitrary complex data structures in different scripting languages to transparently float between different media like memory and disk while enabling cross-language sharing without requiring any modifications to existing script code. Our AI system extensively utilizes this feature. However, my previous implementation had poor read performance when writing to disk. So yesterday, I used our OpenResty XRay analyzer to sample and analyze it, made a few small code changes, and the overall read performance improved by more than 20 times, which is amazing. And it looks like there's still room for further optimization. Performance optimization is only satisfying when it improves by orders of magnitude!
#softwareengineering #performanceoptimization #openresty #crosslanguageinterop #systemsarchitecture #techinnovation -
OpenResty
@OpenResty
When TPU/GPU computing power is no longer the bottleneck for LLMs, local memory and CPU can easily become the bottleneck when processing massive amounts of data. We need to continue using the OpenResty XRay analysis platform to optimize performance and reduce costs... Otherwise, memory usage of hundreds of GB or even TBs is unsustainable...
#llm #machinelearning #performance #openresty #bigdata #technicaloptimization #computingresources #ai -
OpenResty
@OpenResty
The internal execution process of JavaScript's V8 engine exhibits considerable unpredictability. Even when running a simple single-line script like console.log("hello world"), the internal execution flow shows significant variations each time. This goes well beyond mere address randomization - even the called VM C++ functions can vary substantially. This reminds me of LuaJIT, which behaves similarly. In LuaJIT's case, randomness was deliberately added to increase the chances of hitting more efficient code paths. Of course, another reason is security (the more unpredictable the behavioral details are, the more secure it becomes).
#javascript #v8engine #luajit #programminglanguages #softwaresecurity #compileroptimization #techengineering -
OpenResty
@OpenResty
Recently, some users mentioned the goto statement in programming languages in my comments, so I'll address this topic comprehensively here:
We shouldn't be too dogmatic or react negatively whenever goto is mentioned. While traditional structured programming textbooks have consistently opposed goto, that's really just academic pedantry showing limited perspective.
Based on my experience, there are at least three correct and recommended use cases for goto statements:
1. Efficiently implementing complex state machines, such as finite automata or code generated by Ragel
2. Unified error handling within functions, which is commonly used in many classic C programs
3. Simulating continuation-passing style from functional programming in procedural languages to ensure tail recursion optimization - like the "meta regular expression" engine I implemented for my AI professor in one night, which was concise, elegant, and efficient.
Some people also like using goto to quickly break out of deeply nested loops, though I personally don't prefer this usage.
If anyone knows other clever and reasonable uses for goto, feel free to share.
There's nothing absolutely good or bad. If something exists, there must be a reason - we need to view things rationally and dialectically.
#programming #goto #softwareengineering #coding #computerscience #statemachines #errorhandling #functionalprogramming -
OpenResty
@OpenResty
The V8 engine authors designed their own "mini-language" called Torque, which has TypeScript-like syntax and can automatically generate part of the virtual machine (VM) implementation during V8's build process. The Torque compiler generates a large amount of C++ code, which then internally generates the corresponding assembly code. Several hundred large C++ compilation units in the VM are generated this way. It's quite an interesting design approach. However, debugging this automatically generated C++ code and the resulting assembly code isn't quite as pleasant!
Clearly, even V8's developers got tired of writing too much tedious C++ code. Programs that write programs can always be much more convenient.
This reminds me of how LuaJIT's author also designed his own dynamic high-level assembly language called DynAsm, which he used to hand-craft LuaJIT's interpreter implementation. I (and some other folks) have also used this mini-language to implement some simple JIT compilers, which was quite interesting. DynAsm operates directly at the machine code level, making it a much lower level than Torque - even significantly lower level than C.
#v8engine #javascript #torque #compilerdesign #programminglanguages #luajit #dynasm #jitcompiler #codegeneration #systemprogramming #virtualmachine #cplusplus #technicalcomputing #compileroptimization #lowlevelprogramming -
OpenResty
@OpenResty
JavaScript's V8 engine is indeed tightly coupled with the Chromium web browser. Even when compiling V8 from source independently, you need to use Chromium's depot_tools toolchain to clone the repository and synchronize various sub-repositories and dependencies. This makes sense, considering that V8 was originally developed as a component for Chromium. At least it doesn't use Bazel for building by default, which makes things much more comfortable!
#javascript #v8engine #chromium #webbrowser #depottools #webdevelopment #browserengine #compilation #sourcecode #dependencies -
OpenResty
@OpenResty
Using AI to generate code can easily become a "garbage mass-production" or "spaghetti code mass-production" exercise. The key here is that humans must remain at the helm and carefully review AI-generated code. Otherwise, it's likely to introduce many subtle bugs, whether in functional logic, performance, or security and stability aspects. AI also tends to be indifferent about generating large amounts of redundant and verbose code.
#codequality #ai #softwaredevelopment #codereview #technicaldebt #softwareengineering -
OpenResty
@OpenResty
For AI Agents built on large language models that handle extended workflows, the most crucial aspect is allowing users to easily intervene when problems occur at any intermediate step. This applies to both debugging and modifications, as well as the ability to retry and continue from the problematic step. After all, large language models cannot be 100% reliable, and AI Agents based on them share this limitation. Even humans can rarely be 100% dependable, many things require constant adjustments and retries.
#aiagents #llm #workflowautomation #aireliability #debuggingai #machinelearning #artificialintelligence #aitesting #humanintheloop #aiworkflow -
OpenResty
@OpenResty
Envoy server software is truly heavyweight - when compiling from source by default, it builds V8 along with it, making it as painful as compiling NodeJS. Even using the -O0 compilation option, it feels like it takes an eternity... This is despite using Intel's high-end desktop CPU Core i9-13900.
#envoy #proxyserver #v8engine #softwarecompilation #buildoptimization #compilerperformance #corei9 #systemperformance #softwareengineering #devops -
OpenResty
@OpenResty
Today we analyzed and optimized one of our typically time-consuming offline scripts using our OpenResty XRay analyzer. After making a few code changes based on XRay's performance bottleneck suggestions, the CPU efficiency immediately improved by 4x!
Following further optimization guided by updated XRay sampling results, we ultimately achieved an 11x performance boost!
The performance improvement was so dramatic that it felt like a C language rewrite, yet I only modified a few lines of code.
This really demonstrates the value of using your own products in real-world scenarios.
#openresty #xray #performanceoptimization #codeoptimization #cpuefficiency #profiling #performanceanalysis #bottleneckanalysis #luaoptimization #developertools #technicaldebt #softwareperformance #codeprofiling #systemoptimization #devops -
OpenResty
@OpenResty
The official Python C interpreter implementation's handling of CJK (Chinese, Japanese, Korean) wide characters makes extremely heavy use of C macros - to the point where they define individual if-statement branches as macros, and then combine these macros to construct larger if/else if/else statements. This macro usage style is far more extreme compared to other open-source software like Perl, SQLite, and LuaJIT.
#python #programming #code #softwaredevelopment #opensource #codeoptimization #technicaldebt #softwareengineering -
OpenResty
@OpenResty
The Perl interpreter's C implementation really pushes C macros to their limits. I just tested the latest version of clang with some special options using it, and interestingly, I actually discovered some intriguing bugs in clang. For a large-scale software project like clang/llvm, bugs are certainly not uncommon.
#perl #clangcompiler #llvm #cprogramming #compilerbugs #macroprogramming #softwaretesting #systemsprogramming #compilerdevelopment #bugtesting -
OpenResty
@OpenResty
In the age of AI, what should the next generation learn to have a promising future? Perhaps the most important thing is to develop independent thinking and an innovative mindset. Additionally, the ability to quickly self-learn new things is crucial. Programming is also quite important, as to surpass AI, we need to automate machines through programming; we can only defeat machines with machines and magic with magic.
#futureeducation #ai #independentthinking #innovation #lifelonglearning #programming #techskills #digitalliteracy #criticalthinking #futureskills #ailiteracy #selflearning #techeducation #educationinnovation #21stcenturyskills -
OpenResty
@OpenResty
OpenResty Edge 25.6.5 is LIVE! 🚀 Packed with enhancements for better performance & stability.
✨ Highlights:
* ASO rule & log variables
* Zstandard (zstd) compression
* HTTP/3 enabled by default
* SNI wildcard domain support
* Upstream hostname as proxy hostname
* Fixes for 2FA/SSO/OIDC logins & Brotli caching
Dive into the details: openresty.com/en/edge/release-notes/?version=25.6.5
#openrestyedge #webappfirewall #cdn #performance #security #techupdate #zstd #http3 -
OpenResty
@OpenResty
The main purpose of our company's AI initiatives is not to replace human employees, but rather to empower them as "super soldiers." Of course, this requires employees who are ambitious and innovative. The AI era is better suited for teams conducting "special operations." It reminds me of sci-fi blockbusters like "Deadpool" and "The Witch"... That's exactly the effect we're aiming for.
#aiworkforce #employeeempowerment #digitaltransformation #innovationculture #aiintegration #workplaceinnovation #futureofwork #techinnovation #specialoperations #aistrategy -
OpenResty
@OpenResty
Many open-source programming language projects prefer to use "bootstrapping" methods. For example, Perl's build system extensively uses Perl language, LuaJIT uses a lot of Lua code when constructing itself, and GCC uses C language extensively when compiling itself. These projects first create a highly simplified compiler or interpreter, such as miniperl, minilua, or early-stage gcc compiler, and then proceed to build more powerful or final implementations. This multi-step bootstrapping evolution is quite fascinating.
#bootstrapping #compilers #softwaredevelopment #perl #luajit #gcc #compilerc #buildsystems #opensource #code -
OpenResty
@OpenResty
We discovered that the Perl 5 interpreter's build system, like SQLite 3 database, uses its own code generator to parse some of its C code. The difference is that the Perl interpreter uses its own Perl language to parse C code, while SQLite uses the Tcl scripting language. What's similar is that both have strict assumptions about C code formatting. For example, in Perl, macros defined with `#define` are assumed to be exported for third-party Perl module developers, while those defined with `# define` are not (note that there's at least one space after #). So if we modify the coding style of these C code files, it can lead to unexpected problems.
#perl #sqlite #codegeneration #buildsystems #cprogramming #macrodefinitions #codingstyle #tclscripting #perlmodules #syntaxparsing #systemprogramming #compilerdesign #codeformatting #developertools #softwareengineering -
OpenResty
@OpenResty
Good news! OpenResty XRay has been updated to version 25.6.10, bringing some exciting enhancements.
Key highlights include:
- OpenID Connect support for on-premise deployments.
- Auto-refresh for the guided analysis history list.
- Improved kernel compatibility for newer Linux versions.
- Several important bug fixes for better stability and application identification.
For the complete release note, please see: openresty.com/en/ann-xray-25-6-10/ -
OpenResty
@OpenResty
Recently, our team has developed a non-intrusive memory analysis tool for any Java application. This tool is part of the OpenResty XRay platform. It requires no special cooperation from the target Java application process or JDK. A single diagram clearly visualizes how memory is quantitatively distributed across all Java object reference paths. It supports various JDKs, including OpenJDK and OracleJDK. The performance overhead on the target application is minimal, making it suitable for use in latency-sensitive production environments.
Of course, OpenResty XRay also supports similar live quantitative memory analysis for other application processes, such as those written in Lua, Python, PHP, and Perl.
#java #memoryanalysis #performanceprofiling #openrestyxray #devtools #apm #nonintrusive -
OpenResty
@OpenResty
Using LLVM's clang-format tool to format complex C/C++ code still has its pitfalls. Since C/C++ macros are not inherently part of the language, clang-format often gets confused. Manual adjustment of the original code is needed to help it understand... it's not easy. Additionally, a single run of clang-format often won't generate the final format - it requires multiple passes to eventually converge to a state where no further changes are possible. This remains true even with the latest version of clang-format. It's really challenging.
#clang #llvm #cpp #coding #softwareengineering #programming -
OpenResty
@OpenResty
Both practicing meditation and deep reading of thick books help train sustained attention. Entrepreneurs need to frequently switch between many different tasks daily, which is truly taxing. It's all about short attention spans... Now in the AI Agent era, even large language models emphasize long context and extended attention spans, don't they? We can't fall behind AI, at least not yet...
#meditation #deepreading #attention #entrepreneurship #ai #mindfulness #productivityhacks #focustraining -
OpenResty
@OpenResty
Envoy gateway server, much like Nginx and our OpenResty, also supports programming with Lua.
Our OpenResty XRay now also supports real-time sampling of live, unmodified Envoy service processes to generate Lua code-level CPU and off-CPU flame graphs.
These graphs provide an intuitive visualization of how CPU time and off-CPU time (essentially blocking time) are quantitatively distributed across all Lua code execution paths.
Furthermore, our AI system automatically analyzes and explains these flame graphs, directly identifying potential performance bottlenecks and performing root cause analysis.
It has to be said, Envoy is genuinely even more bloated and massive than NodeJS.
#envoy #lua #openresty #openrestyxray #performanceanalysis #profiling #flamegraphs #observability #aiops #nginx #nodejs -
OpenResty
@OpenResty
In a talk, OpenAI's former Chief Scientist Ilya Sutskever mentioned the concept of "compression is prediction," which carries profound implications and truly captures the essence of large language models. It's worth contemplating deeply. This insight has fundamentally influenced almost all of my AI-related work.
This reminds me of an ancient story from Buddhist scriptures: Sariputra, Buddha's most accomplished disciple, was initially studying under other philosophical schools. However, after merely hearing one of Buddha's sayings - "All phenomena arise from causes and conditions" - from another Buddhist disciple, he immediately grasped the majority of Buddhist principles. This led him to bring his friend Maudgalyayana to join him in following Buddha's teachings.
#ai #machinelearning #openai #ilyasutskever #buddhism #deeplearning #llm #datascience #techphilosophy #aiwisdom -
OpenResty
@OpenResty
The currently trending AI Agents are essentially hybrids that combine traditional programming with large language models, compensating for each other's limitations. Of course, they may also incorporate elements like RAG and vector databases.
The Agents we see in the market today are quite different from the Agent concepts I learned in classical AI theory during my high school self-study days.
Actaully, the 'agent' prefix in my online handle 'agentzh' was actually inspired by the Agent concept from classical AI.
#aiagents #largelanguagemodels #ai #rag #vectordatabases #programming -
OpenResty
@OpenResty
When parsing data returned by large language models with my own programs, these models occasionally return unexpected and bizarre outputs that my code cannot process - even SOTA models are no exception. So I have to keep modifying my code to handle each new edge case as I encounter them. It feels like an endless cycle... quite frustrating.
#llm #ai #programming #modelparsing #tech #softwaredevelopment #sota -
OpenResty
@OpenResty
At this stage, we should view AI large language models as our interns or consultants rather than replacements. After all, if something goes wrong, the person using AI will still be held responsible. The right approach is to use AI to enhance our own skills; otherwise, our professional competency will rapidly deteriorate... and we'll end up being truly obsolete.
#ai #professionaldevelopment #careergrowth #techskills #airesponsibility #lifelonglearning -
OpenResty
@OpenResty
I spent considerable time figuring out the necessary command line options for the md-to-pdf open-source tool to generate perfect PDFs from numerous markdown files. Here they are:
--highlight-style 'atom-one-dark-reasonable' --pdf-options '{"margin": 0, "printBackground": true, "format": "Letter", "preferCSSPageSize": true, "timeout": 300000 }' --stylesheet 'my-dark.css'
I've shared the my-dark.css file here: gist.github.com/agentzh/7ccae1430b3eba7b6509e218b7312f8c?…
I also modified the atom-one-dark-reasonable highlight.js CSS style, making the code comments brighter since they were hard to read against the dark background.
One tricky issue is that preferCSSPageSize must be set to true, and you need to set the page size to Letter in your CSS. Otherwise, in some cases, the generated PDF will automatically switch to A4 size, which is quite problematic. Additionally, the default 30-second Puppeteer timeout can cause generation failures for large files. Setting it to 300 seconds seems to prevent timeout issues.
#markdown #pdf #documentation #developertools #nodejs -
OpenResty
@OpenResty
We've recently released the new OpenResty open source version 1.27.1.2, which includes numerous recent bug fixes and new features.
The complete changelog is available here: openresty.org/en/ann-1027001002.html
We recommend all open source OpenResty users upgrade to the latest version. #openresty -
OpenResty
@OpenResty
The open-source tool Pandoc is still not very reliable for generating PDF files from Markdown. When dealing with large Markdown files generated by AI or programs, it simply throws errors like "dimensions too large." The official team doesn't seem interested in fixing such issues, and even the latest version fails to handle this.
The JavaScript-based open-source tool md-to-pdf appears to work much better, although its default styling is quite unattractive. Fortunately, it's easy to beautify using custom CSS. However, since it uses the Chromium browser for rendering, it's quite heavy and slow. Additionally, Puppeteer has several minor issues that require manual hacking to achieve optimal results.
In the AI era, these traditional programming toolchains face higher demands. Large language models sometimes generate invalid Markdown, which requires manual correction.
#markdown #pdf #pandoc #mdtopdf #toolchain #documentation #developertools #opensource #javascript #puppeteer -
OpenResty
@OpenResty
We've just released OpenResty XRay 25.4.15, our real-time application analysis product! Key highlights:
• Added NodeJS app detection with C++ level analysis (JavaScript tracing coming soon)
• Enhanced target filtering by IP/hostname
• Optimized data compression for faster uploads
• Fixed container analysis issues & reduced memory usage
• Improved 2FA security for self-hosted deployments
More details: openresty.com/en/xray/release-notes/?version=25.4.15 -
OpenResty
@OpenResty
Our AI system discovered that SQLite 3 open-source database software uses Tcl scripts during the build process to dynamically generate final massive C files. For example, it generates corresponding definitions and interpreter implementations for different DB operation codes. In comparison, LuaJIT's implementation in similar situations is much smarter, directly utilizing C macros as C code generators, thus avoiding the need to hack the build system or introduce external scripting tools.
SQLite 3's Tcl code generator needs to parse existing C files and has strict requirements for code indentation, making it quite fragile from an engineering perspective. If we reformat its C files using an IDE or clang-format, the build will fail immediately because the Tcl code generator won't work correctly. Their Tcl implementation appears to have taken some shortcuts.
#sqlite #luajit #databasedevelopment #codegeneration #opensource #tclscripting #softwareengineering #buildsystems -
OpenResty
@OpenResty
It appears there are noticeable differences in TPU computing power across Google's various data centers. We regularly use large language models in data centers across America, Europe, and Asia, and have found that for Gemini batch processing tasks, us-east1 is the slowest - possibly due to having the oldest TPU deployment - with processing delays exceeding 400 seconds at best. Following that are regions like us-south1 and asia-northeast (Japan), taking 200-300 seconds at their fastest. The remaining regions are much quicker, requiring only around 100 seconds at their best.
#tpu #googlecloud #datacenters #cloudcomputing #gemini #ai #machinelearning -
OpenResty
@OpenResty
I typically set the temperature parameter of large language models to 0, except for occasions when I want the model to enter a more creative mode or generate more varied responses.
#llm #aiparameters #machinelearning #temperature #promptengineering -
OpenResty
@OpenResty
I've noticed that the new form of programming in the AI era is writing "prompts." However, traditional programming is still very necessary right now. Both need to be organically combined to achieve good results.
After all, we can't rely solely on manual back-and-forth conversations with large language models. Human input-output efficiency is just too low, and I'm no exception.
Earlier, I saw a user on X asking how people with zero programming background could integrate AI into their research. Google AI Studio's product lead replied, "Learn to code first." So at least at this stage, learning traditional programming skills is still very necessary. After all, when communicating with AI, natural language isn't always the best choice - using computer languages might be more precise, similar to mathematics and other formal languages.
#aiprogramming #promptengineering #codingskills #airesearch #techlearning #ai #vibecoding -
OpenResty
@OpenResty
I used to think that through error feedback, large language models could continuously correct their outputs. However, for truly complex problems, even the best current models aren't necessarily convergent under feedback prompts.
They might get stuck in infinite loops, go down garden paths with no return, or actually get worse with each correction. Deepseek researchers have also acknowledged that R1 faces this issue. Those legendary "Aha moments" are truly elusive. The real world is much messier than math olympiad problems and eval test sets. These days, I often avoid letting LLMs self-correct because in the real world, there's a significant risk of large models making things worse as they attempt to improve.
#ai #llm #machinelearning #deeplearning #aifeedback #modelbehavior #ailimitations #techchallenges #deepseek -
OpenResty
@OpenResty
Even the best SOTA large language models show clear signs of being influenced by mediocre code when writing slightly complex programs. Perhaps during pre-training, these models should only be exposed to truly high-quality code.
#ai #llm #codequality #machinelearning #programming #modeltraining -
OpenResty
@OpenResty
Looking at the current state of large language models in the market, I can summarize it in one sentence: "Sufficient for basic literacy, but lacking in professional expertise." These large models have inherited human flaws - they tend to be careless, prone to speculation, forgetful, and sometimes disobedient... Therefore, it's crucial to use them in appropriate scenarios while minimizing their shortcomings.
When I bombard even the best SOTA large language models with billions of tokens daily, they frequently malfunction in bewildering ways... Of course, smaller models display even more bizarre forms of confusion. So current large language models remain quite unreliable.
#ai #llm #machinelearning #aireliability #sota #nlp #artificialintelligence -
OpenResty
@OpenResty
A good AI Agent program framework should meet these requirements: 1. Lightweight and efficient, 2. Easy to automatically maximize request concurrency, 3. Transparent switching between different LLMs (vendor APIs or locally deployed), 4. Transparent mixing of models of different sizes, 5. Transparent switching between real-time request and batch processing modes, 6. Flexible rerunning of individual processing stages, 7. Easy data flow tracking for troubleshooting, 8. Seamless switching between streaming and batch output modes, 9. Transparent distributed request handling and error retry mechanisms, 10. Transparent memory-based or disk-based massive data storage and computation, 11. Transparent local and remote LLM prompt caching, and so on. It should balance high throughput, low latency, low cost, high development efficiency, and high debuggability.
That's why I wrote my own AI framework from scratch, without depending on external service components. After accumulating features and optimizations for more than half a year, it still only has a few thousand lines of code, extremely lightweight and efficient. It comes with an automated test bench that easily handles massive test sets at the scale of hundreds of millions of tokens.
#aiframework #llm #softwareengineering #aiagent #techinfrastructure #devtools #ai -
OpenResty
@OpenResty
Large language models still have noticeable stability issues with prompt inputs. Sometimes even seemingly harmless or irrelevant elements can have unexpected significant impacts on these models. Whether it's the smartest SOTA model or a smaller model, such issues will arise. It feels somewhat like the butterfly effect, or similar to a chaotic system. That's why prompt engineering remains crucial - a small difference can lead to problematic outcomes. Moreover, different models behave differently. Having reliable regression test suites is naturally also critical.
I've seen some users suggest treating large language models like humans, but they really aren't human - their behaviors differ dramatically from human behavior.
#llm #ai #promptengineering #machinelearning #aistability #modelbehavior #aitesting -
OpenResty
@OpenResty
Many programmers are often afraid to modify code they've already written; they try to avoid changes whenever possible. However, I particularly enjoy refactoring or polishing old code. Programs that aren't modified much are essentially "dead programs." The refactoring process is also a way to interact directly with the code, often uncovering deeply hidden bugs in the old code. It's similar to how automated testing is a form of interaction with the program. The more you engage with old code, the more opportunities you have to discover bugs. It's good to "twist" the old code from time to time.
AI-generated code needs even more "twisting" because the "thought processes" of AI are generally quite strange, and sometimes it can seriously spout nonsense.
Of course, this approach heavily relies on a comprehensive test suite; otherwise, it's easy to introduce new bugs. Each time a new bug is discovered, we should first add test cases to cover it, making the test suite increasingly complete. For the same reason, I also dislike writing unit tests for internal functions (except for stable API functions), because internal functions often change during refactoring and rewriting of code. If I have to modify a bunch of test cases every time I refactor the code, it becomes quite troublesome and turns into a burden.
#ai #gemini #googleai #testing #machinelearning #technews #aitesting #debug #program #programmer -
OpenResty
@OpenResty
For many years, my goal has been to quickly understand how complex software works. It's as exciting as exploring uncharted territories. I plan to share various open-source software's inner workings in a way that beginners can understand, which should be very interesting topics.
Of course, I always use the dynamic tracing tool of OpenResty XRay to assist and validate my understanding.
#techchallenges #cloudcomputing #dynamictracing #program #engineering #engineer -
OpenResty
@OpenResty
Many developers have asked about test-driven toolchains and implementation details, so I'll elaborate a bit more. I personally favor Perl's testing toolchain, especially the mini-language in Test::Base, which is extensible, concise, and convenient for testing any software system written in any language. This includes automating web browsers to test complex web application interactions (of course, we've also implemented our own AI vision capabilities here).
#testdrivendevelopment #perl #testbase #softwaretesting #automation #testing #program #engineer -
OpenResty
@OpenResty
Many users are interested in the AI tool I wrote at Taobao over a decade ago - a system that could automatically build relatively complete data analysis products. Let me share more about it.
At that time, this AI system, consisting of about 4,000 lines of code, outputs several megabytes of C, Lua, SQL, and other code. I actually wanted to extend it to frontend JS as well (I had written the frontend JS framework for Taobao's data platform product myself, as pure client-side single-page applications weren't common yet, and frameworks like Vue and React didn't exist, so I created one from scratch. Supporting IE6 compatibility was quite exhausting at the time).
The generated code is 100% accurate, including automatically generated test suites. Humans only modify the input documentation.
This AI system accepts diverse inputs, including original natural language wiki documents written by other teams, as well as my own custom document mini-language for describing complex business requirements and constraints.
The AI system can perform extremely complex and aggressive optimizations for computational code and distributed database queries - optimizations that humans generally lack the capability or courage to implement. Meanwhile, the generated code is highly debuggable. Both throughput and latency show order-of-magnitude improvements compared to the original hand-written legacy systems, with far fewer bugs, since machines don't get careless or lazy.
At that time, the product had over 2 million daily active users, all of whom were Taobao sellers. During Singles' Day (11.11), a couple of basic 4-core 8GB servers could handle the load effortlessly, with very low utilization.
In my junior year of college, I used similar methods to read Intel's technical manuals and automatically generate hardware designs for instruction decoders covering a fairly complete x86 instruction set. I even discovered several typos in Intel's manuals, since AI would actually test all the details mentioned.
In high school, I used similar technology to create automated solution programs for various college entrance exam subjects - physics, biology, geometry, algebra, English, and other disciplines, with great results. Of course, I continued this approach in college, working on university physics, linear and nonlinear complex circuits, logic circuits, and formal program verification. Otherwise, the school curriculum would have been too boring.
So my AI practice goes way back - I've been extremely interested in it since my first year of high school. Back then (2000), there were no large language models, and deep learning wasn't even on the horizon.
I started learning programming and AI on my own when I first entered high school, studying both almost simultaneously. Of course, my mom often scolded me for "not focusing on what matters," since all my peers were busy preparing for college entrance exams. Even the two like-minded classmates I had quickly "hung up their boots." So it was quite lonely. Fortunately, once I got to university, I could regularly access the internet and interact with hackers worldwide, which was truly amazing. It made me realize how worthwhile learning English had been.
#ai #codegeneration #programming #taobao #innovation #softwareengineering #distributedsystems #optimization -
OpenResty
@OpenResty
Some programmers argue with me that software needs to be specifically designed to be "testable," claiming this increases development costs - which is completely wrong. I've always believed in testing software the same way humans would: if humans can use it and test it, machines can test it too. The target software doesn't need any special accommodation. It's similar to dynamic tracing technology - only non-intrusive approaches are practical and convenient to apply, rather than using intrusive mocking methods. This includes injecting extreme error cases, which can also be done non-intrusively. In fact, there's often overlap with dynamic tracing techniques.
Don't get brainwashed by the "garbage" advice from people online who don't actually understand automated testing...
#softwaretesting #automatedtesting #dynamictracing #softwareengineering #testautomation #qualityassurance -
OpenResty
@OpenResty
I found that many engineers online have a strong aversion to writing automated test cases. The main reasons often stem from using the wrong testing tools or having serious misunderstandings about testing methods, which leads to very high costs for introducing automated testing and very low returns. We generally express various complex test cases using a small language similar to documentation, whether for the front end or back end. The cases appear extremely concise and elegant, and they are also suitable for automatic generation or updating by AI large models. The testing platform provides as precise debugging information as possible when cases fail, assisting in problem diagnosis. So, in our toolchain, writing and calling cases is actually a pleasure [haha].
There are also engineers who argue with me about development speed and efficiency. In fact, when test-driven development (TDD) is used correctly, development efficiency is much higher, and the quality of the output is significantly better. Whenever I think about slacking off, I find that my development efficiency quickly drops to an unbearable level; it’s just enjoyable at the beginning. Only those who do not understand TDD would think that adding automated testing would slow down development progress. After all, during the iterative development process, having machines check software behavior is far more efficient and accurate than relying on humans.
#programmer #programming #softwareengineering -
OpenResty
@OpenResty
I've noticed many developers misunderstand Test-Driven Development (TDD), thinking they need to create a perfect test bench and test suite right from the start. This is completely wrong.
Test suites and test benches can evolve dynamically - they don't need to be perfect from the beginning. When starting from scratch, the priority is to get something working quickly.
Initially, I might use a simple bash script as my test bench, then gradually replace it with more powerful and intelligent testing tools. It's all an evolutionary process.
The key principle is pragmatism.
After all, many projects start without a clear understanding of the best implementation or testing approaches. These insights typically emerge as the project progresses.
During development, both the project itself and the test suite might undergo multiple rewrites and major refactoring. This is normal - the code must be easy to tweak and hack.
I recall Donald Knuth once saying that "editable code is the best code," not "reusable code." In my interpretation, this means code should be easy to hack and modify freely.
#tdd #softwaredevelopment #testing #programming #donaldknuth #devpractices #codequality #softwareengineering -
OpenResty
@OpenResty
One of my key strategies for tackling increasingly complex software system development over the years has been test-driven development. Although it requires significant time and effort invested in testing upfront, the continuous returns in the long run are well worth it. Test-driven development enables a rapid, incremental approach where both the test suite and the software itself grow quickly together.
I first learned about test-driven development in high school. Initially, I looked into the unit testing frameworks in the C# world, but later I became a fan of Perl's testing toolchain, regardless of the programming language of the test target.
For example, the programs I've written in languages like Lua, Perl, Python, Haskell, Erlang, C/C++, and JavaScript were all driven by Perl's testing toolchain. I especially love Perl's Test::Base testing mini-language.
Previously, I hired two excellent Linux kernel developers, one from Eastern Europe and the other from the United States. One of them was even a prominent member of the official Ubuntu kernel team. However, when I assigned them a kernel development task, they struggled for several months and told me that it might be technically impossible. I didn’t believe that, so I used a test-driven approach and completed it in just one night, despite having almost no prior experience in kernel development. The key to test-driven development is to keep the time between code modification and seeing the final running feedback as short as possible, allowing for very rapid iteration. This is how you can challenge the "impossible."
In many cases, seemingly untestable scenarios can be addressed with clever automated testing solutions if we think creatively. This has been a recurring experience in my work over the past few decades. The way people think is actually very important. Instead of focusing on a few difficult points and dismissing methods that are effective in most situations, it's better to maintain an open mindset and consider how those generally effective methods can be applied to more specific and challenging scenarios. If a method didn't work for me but succeeded for others, I should learn more about their successful experiences rather than hastily dismissing their methods based on my own failures. It's especially important to be cautious when making "negative" judgments; this applies to academic research as well as engineering practices.
Now that there are powerful AI large models, it's even more convenient. Once the testing platform is ready, generating a large number of unit test cases with a large model is also a very easy task.
wx4.sinaimg.cn/orj360/6d579ff4gy1hw2zm3onwvj20sm0koaqk.jpg
#testdrivendevelopment #softwareengineering #perl #unittesting #softwareengineering #codingchallenge #coding -
OpenResty
@OpenResty
Our OpenResty XRay product provides comprehensive dynamic tracing and analysis capabilities from multiple perspectives:
1. Thread blocking and lock issues
2. High CPU usage or inability to fully utilize CPU
3. High memory consumption or memory leaks
4. Disk IO analysis
5. Application-level request latency, long-tail requests, and intelligent packet capture for abnormal connections, along with automatic network packet analysis
6. Various exception capturing and analysis (including application-caught exceptions)
7. Core dump file and process crash analysis
8. Memory heap corruption and stack corruption analysis, or general buffer overflow problem diagnosis
Since we make the entire software stack transparent, the problems we can analyze are limited only by our and our customers' imagination. "You wish it, we make it happen!"
#openresty #performance #monitoring #debugging #systemanalysis #troubleshooting #apm #observability -
OpenResty
@OpenResty
Our dream for many years has been to transform all applications across all tech stacks and programming languages from black boxes into white boxes, eliminating mysterious production issues. This is precisely the mission of our OpenResty XRay.
We previously encountered a situation with a customer where their production Xeon CPU had a hardware bug, which we discovered using OpenResty XRay. Initially, their engineers wouldn't acknowledge it was a hardware issue and insisted on patching the Linux kernel to confirm the problem. Quite dramatically, OpenResty XRay then discovered bugs in their kernel patch. Eventually, they confirmed it was indeed a CPU hardware issue. We're still grateful to the engineers for helping confirm the problem.
There are countless bugs in the world, which is why OpenResty XRay aims to be the ultimate bug hunter.
#bughunting #openresty #debug #bugs #programmer #computer #tech #computerscience -
OpenResty
@OpenResty
Here's a joke: It seems the most skilled programmers I've seen actually spend most of their time reading, understanding, and maintaining other people's "spaghetti code" rather than working on their own projects...
So when you have the opportunity to experience the joy of building software from scratch, be sure to cherish it.
#programming #coding #codingjourney #codinglife #codingfacts #programmer -
OpenResty
@OpenResty
Large language models are a completely new species - they shouldn't be viewed as either human or machine. I recall an OpenAI scientist saying (if I remember correctly) that they don't fully understand the large models they've trained themselves. There's still much "hidden knowledge" within these models waiting to be discovered. When I first heard this, it didn't really resonate with me. But now, the more I interact with AI daily, the more I appreciate the profound meaning behind those words. In front of large language models, we're all essentially "elementary school students" - even the creators of these models are no exception. Now it's just a matter of who can best unlock the full potential of these large models. It's like the "Saiyans" from Dragon Ball - their potential is limitless... haha!
Come to think of it, 99% of users have only scratched the surface of what large language models can do - and even that's being generous. When using these models at a superficial level, users aren't even utilizing 1% of their true capabilities. I was one of those novice users initially. It wasn't until I recently delved into numerous research papers and other materials in this field that I had my "aha" moment.
#ai #llm #machinelearning #dragonball #artificialintelligence #technology #innovation #openai -
OpenResty
@OpenResty
I've built an AI framework almost from scratch. Except for vector dot product which use existing open-source libraries, I implemented everything else myself. It can mix and switch between any LLM models, has memory capabilities, and supports AI vision recognition to operate any Web or GUI applications. It can automatically control command-line terminals, supports internet connectivity and Google searches, and includes a powerful AI editor (which I've personally tested to be much more practical than GitHub Copilot). It supports semantic search indexing and retrieval for Confluence, Slack, Markdown, and code in any programming language. It features a graph-based complex conversation playground, REST API, and Web UI... What's more interesting is that it can self-learn and improve its own code implementation.
All these features combined amount to just over 4,000 lines of code. This reminds me of John Wick's request to the gun shop clerk: "Precise, Robust." We could add words like Minimal, Beautiful, Flexible, Powerful too, haha!
I rarely write code for development now. The AI I've specially trained can write many complex programs, with code quality approaching what I would write by hand. These days I mainly do code reviews, while AI writes and modifies her code. This review process also serves as training data for itself - the highest quality samples, haha. It still gets confused only with particularly complex algorithms.
#ai #machinelearning #aiframework #programming #opensource #llm #aitools #techinnovation #softwaredevelopment -
OpenResty
@OpenResty
To master AI, you indeed need to delve into many intersecting fields and disciplines. Even taking a cursory look at these areas can be quite challenging, but that's precisely what makes it endlessly fascinating. If you could see the bottom right away, it wouldn't be very interesting. It reminds me of a classmate's catchphrase from university: "A challenge? I love it!"
#artificialintelligence #learningai #techeducation #interdisciplinary #continuouslearning #techchallenges -
OpenResty
@OpenResty
Based on my recent research and practice, AI replacing human jobs is more imminent than I previously thought. I'm wondering what my children should learn to be relatively secure in the future. Currently, it seems that cutting-edge research scientists can't be replaced yet, and sales roles that require human interaction are also safe for now. There might be other positions too. In the long run, people must maintain an edge over AI to remain competitive. The pressure is enormous. I myself might also be replaced by AI soon.
We need to be cautious when using some AI products, so we don't let AI lead us by the nose, which could result in the degradation of our skills and knowledge. This would make it easier for humans to be replaced by AI. For example, I am quite opposed to our engineers using GitHub Copilot because the default usage mode of such AI products interferes with and influences human independent thinking. Therefore, I have been guarding against this issue in Chunbot from the very beginning. The boundaries must be clear.
#ai #futureofwork #careers #techdisruption #aijobs #education #workforcefuture -
OpenResty
@OpenResty
undo.io's Live Recorder and UDB commercial products are quite fascinating, offering more flexibility and power than Mozilla's open-source rr debugger (including support for arm64 and post-event attach/detach capabilities), albeit with higher performance overhead. These tools become even more powerful when combined with our OpenResty XRay product, especially since our toolchain like ylang can simultaneously generate high-performance GDB extensions that work with either UDB or rr. I plan to write some articles sharing these insights - it'll be very interesting. At OpenResty, we use both Undo and Mozilla's products with great results. For instance, I've used rr to identify and fix many deep bugs in LuaJIT's JIT compiler.
wx1.sinaimg.cn/orj360/6d579ff4gy1hpfissgf9sj20gc092mym.jpg
#debugging #openresty #undoio #mozilla #luajit #softwareengineering #programming #debugging #reversedebugging -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The cost-effectiveness of the Gemini 1.5 Flash model API is quite good, especially in batch processing mode (although its batch API is difficult to use). However, Google's cheapest Gemini 1.5 Flash 8b model is too weak—so weak that it sometimes doesn't follow instructions, causing many of our test cases to fail, and it's not easy to work around these issues.
In my daily AI application development, I really enjoy using the 1.5 Flash model for testing and debugging—it's fast and affordable.
The experimental Gemini 2.0 Flash model shows significantly better quality than 1.5 Flash. I hope the pricing won't increase after it becomes generally available (GA). #llm -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I recently noticed that the concurrency of large model batch processing tasks in each Google Cloud region has dropped from 4 to just 1. This is true for all Asia-Pacific and North American regions at least. The change happened just a few days ago. Perhaps they're preparing computing resources for the GA release of the Gemini 2.0 Flash new model? The official documentation states that the regional batch processing concurrency is 3.
I'm looking forward to Gemini 2.0 GA. Currently, the API rate limits for the experimental model are too strict, and the computing power is too low - the 1500 RPD (requests per day) isn't enough to run our test suite. Running the test suite once requires at least tens of thousands of LLM requests. #gemini #llm -
OpenResty
@OpenResty
Our OpenResty XRay dynamic tracing product's Java analyzers finally support Azul's Zulu JDK!
This includes Java-land CPU/off-CPU/diskIO flame graphs, latency analysis and smart packet capture, dynamic probes for Java method entry points, and more. We have a cryptocurrency exchange client who needs this explicitly. OpenJDK, Oracle JDK, and Zulu JDK—all three JVM variants are now fully supported. Depending on our enterprise customers' actual needs, we'll consider supporting other variants like IBM JDK and Microsoft's JDK later. #dynamictracing #java #jvm #zulujdk #azul #ebpf -
OpenResty
@OpenResty
Our OpenResty XRay, a dynamic-tracing product, transforms black-box production systems into white-box ones. Many issues can be observed and captured before they escalate into harmful problems. This continuous insight and control over production systems is what's most crucial. Offline simulated traffic and testing methods often have limited coverage. No matter how much effort is invested in offline simulation, gaps will always exist.
Additionally, many of our customers prefer using OpenResty XRay for white-box analysis even in offline environments; regardless of the environment, white-box is much more convenient than black-box. You can't keep adding monitoring metrics and debug logs every time, can you? ;)
Our ultimate goal is to help customers quickly understand their legacy systems (or "sh*t mountain systems") and improve them efficiently at low cost rather than just being a tool for solving production issues. "Turning sh*t into gold" - that's what makes this truly interesting and valuable. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Many users ask me about the implementation principles and technology behind @OpenResty XRay, as it looks magical.
• For those interested, you can refer to my previous article about dynamic tracing: blog.openresty.com/en/dynamic-tracing-part-1/
• Here's another series of articles I wrote introducing the ylang language: blog.openresty.com/en/ylang-intro-part1/
• And here's an article I wrote about the y++ language: blog.openresty.com/en/cpp-trace-cpp/
Our recent advanced dynamic tracing capabilities for Java are implemented using our C++-compatible y++ language.
Interestingly, our team has been using OpenResty XRay to analyze OpenResty XRay itself. However, this capability is disabled for our customers to prevent misuse.
#dynamictracing #domainspecificlanguages #ebpf -
OpenResty
@OpenResty
OpenResty XRay now supports setting dynamic probes at any #java function entry point! This feature has wide applications - for example, we've implemented real-time latency statistics for requests within Tomcat servers based on this feature, and integrated it with our online intelligent packet capture analyzer. It doesn't depend on any configuration of the target process or log files.
It supports both interpreted and JIT-compiled Java methods. It's still based on 100% non-intrusive dynamic tracing technology, with negligible performance overhead.
It doesn't rely on bytecode injection or machine code injection, doesn't depend on JVM Agent mechanism, and is completely read-only for the JVM target process - the target process doesn't even need to restart. It supports both the latest Linux kernels and old Red Hat 7's 3.10 kernel.
Since this feature is very new, it currently only supports OpenJDK, but will soon support Oracle JDK. In the near future, it will definitely support more variants like Zulu JDK, IBM JDK, Azul JDK, and others.
Support for dynamic probes at Java function exit points will be added later.
#jvm #java #dynamictracing #ebpf -
OpenResty
@OpenResty
Following OpenJDK, our OpenResty XRay can now deeply analyze #oraclejdk!
We're still using our 100% non-intrusive dynamic tracing technology. The analysis covers everything from the C++ VM level to the #java source code level without any blind spots.
Next up will be other JDK variants like Zulu JDK and IBM JDK 😀
#jvm #java #dynamictracing #ebpf -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Some people speculate that our OpenResty XRay uses the ELF file's build ID to look up existing DWARF information in a database, which made me laugh so hard.
We're actually targeting binaries that customers have modified and compiled without the -g flag or binary software packages downloaded from third parties without debug symbols. This is the most challenging and general problem to solve.
For any existing binary programs and their debug symbols found online, our crawler system has already indexed them in our Pkg DB, which has accumulated hundreds of terabytes of data and is still growing daily. Why would we need such advanced machine learning technology for those cases?
OpenResty
@OpenResty
🔥 OpenResty XRay's New Capability Unlocked! Analyze Applications Without Debug Symbols! Video: youtu.be/swUwT3sEtUI
Even when binary programs have no debug information, it can:
• Automatically reconstruct debug symbols from binary programs using our own machine-learning algorithms
• Display complete call stack traces at the language levels of C, Lua, Python, Perl, PHP, etc.
• Precisely locate source files and line numbers
• Zero configuration and zero hints needed for target programs. 100% non-intrusive.
The accuracy here is close to 100%, at least for the types of symbols our product needs; they are primarily accurate enough. If there are any inaccurate cases, our human experts can help fine-tune them, which is not a problem.
We are porting this magical new technology to large C++ program binaries like the JVM. OpenResty XRay will soon be able to automatically reconstruct high-quality debug symbols from any JVM binary program files. It should take about a few more days.
Text version: blog.openresty.com/en/nodwarf/
#ebpf #openresty #dynamictracing #machinelearning #debugsymbols -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Someone on the Internet asked: "What if you randomly grab a programmer from the street, and they can solve problems using your OpenResty XRay tools?" They seem to be concerned that only experts could use our products. My response is that their worry is unnecessary, and our technical experts will directly help clients solve problems - that's our guarantee. It doesn't matter if clients themselves don't understand technology or know how to use OpenResty XRay preciselyproducts. This is exactly why some non-technical business owners like our services. As the saying goes, "You make a wish, we make it happen!"
-
OpenResty
@OpenResty
🔥 OpenResty XRay's New Capability Unlocked! Analyze Applications Without Debug Symbols! Video: youtu.be/swUwT3sEtUI
Even when binary programs have no debug information, it can:
• Automatically reconstruct debug symbols from binary programs using our own machine-learning algorithms
• Display complete call stack traces at the language levels of C, Lua, Python, Perl, PHP, etc.
• Precisely locate source files and line numbers
• Zero configuration and zero hints needed for target programs. 100% non-intrusive.
The accuracy here is close to 100%, at least for the types of symbols our product needs; they are primarily accurate enough. If there are any inaccurate cases, our human experts can help fine-tune them, which is not a problem.
We are porting this magical new technology to large C++ program binaries like the JVM. OpenResty XRay will soon be able to automatically reconstruct high-quality debug symbols from any JVM binary program files. It should take about a few more days.
Text version: blog.openresty.com/en/nodwarf/
#ebpf #openresty #dynamictracing #machinelearning #debugsymbols -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The open-source vector database Qdrant actually has a nasty bug that occasionally causes unexpected data loss. The official team finally confirmed it was a bug in the database index implementation. Alas. github.com/qdrant/qdrant/issues/5134 There are so many bugs in the software world... a database that loses data is the scariest thing.
Fortunately, it got fixed in the latest version. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
One of my dreams for many years has been to turn all tech stacks and applications in all programming languages from black boxes into white boxes, eliminating mysterious issues in production environments. This is precisely the mission of our OpenResty XRay.
Previously, we encountered a case where a hardware bug was discovered using OpenResty XRay on a customer's production Xeon CPU. Intel engineers initially didn't acknowledge it was a hardware issue and insisted on patching the Linux kernel to confirm the problem. Dramatically, OpenResty XRay then discovered bugs in their kernel patch. Finally, it was confirmed to be a CPU hardware issue. We still need to thank the Intel engineers for helping confirm the problem.
There are so many bugs in the world, so OpenResty XRay aims to be the best bug hunter.
Going further, hopefully, one day, it can even automatically apply binary patches to VMs or machine code in running processes to fix bugs or security vulnerabilities. This belongs to live patching. Everything can be live-patched from kernel to system software and up to business code without even restarting processes. We have already made some exciting experiments in this direction, and it's very promising. Advanced technology looks like magic 😁 -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Years ago, while working at Alibaba, I wrote online services using Erlang/OTP. However, for the same business complexity, the performance of #erlang services was far behind my OpenResty [Grin]. Nevertheless, Erlang is still a comfortable, functional programming language to use, unlike Haskell, which requires an entirely different programming mindset.
Now OpenResty XRay already supports dynamic tracing for Erlang, and in the future, we might add support for #haskell applications in OpenResty XRay as well.
OpenResty
@OpenResty
🔍 New video tutorial: Using OpenResty XRay to diagnose high CPU usage in Erlang/OTP applications: www.youtube.com/watch?v=sqB5fGHahYU
Highlights:
• Pinpoint CPU-intensive regex operations, among many other bottlenecks
• Analyze code paths on both the Erlang language level and the VM C/C++ level (with Flame Graphs!)
• Zero-intrusion profiling - no restarts, no special options/flags, and no code changes needed!
• Based on our advanced dynamic tracing tech eBPF+/Stap+/etc.
• Neglectacble tracing overhead; suitable for production use
• Already helping big companies like Zoom, Inc
• Transparently peeking inside already running processes and containers
Text version: blog.openresty.com/en/erlang-high-cpu/
#openresty #erlang #performance #ebpf #observability -
OpenResty
@OpenResty
🔍 New video tutorial: Using OpenResty XRay to diagnose high CPU usage in Erlang/OTP applications: www.youtube.com/watch?v=sqB5fGHahYU
Highlights:
• Pinpoint CPU-intensive regex operations, among many other bottlenecks
• Analyze code paths on both the Erlang language level and the VM C/C++ level (with Flame Graphs!)
• Zero-intrusion profiling - no restarts, no special options/flags, and no code changes needed!
• Based on our advanced dynamic tracing tech eBPF+/Stap+/etc.
• Neglectacble tracing overhead; suitable for production use
• Already helping big companies like Zoom, Inc
• Transparently peeking inside already running processes and containers
Text version: blog.openresty.com/en/erlang-high-cpu/
#openresty #erlang #performance #ebpf #observability -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
All the new shiny Erlang dynamic tracing tools are written in our Ylang universal debug language as always: blog.openresty.com/en/ylang-intro-part1/
OpenResty
@OpenResty
🔍 New video tutorial: Using OpenResty XRay to diagnose high CPU usage in Erlang/OTP applications: www.youtube.com/watch?v=sqB5fGHahYU
Highlights:
• Pinpoint CPU-intensive regex operations, among many other bottlenecks
• Analyze code paths on both the Erlang language level and the VM C/C++ level (with Flame Graphs!)
• Zero-intrusion profiling - no restarts, no special options/flags, and no code changes needed!
• Based on our advanced dynamic tracing tech eBPF+/Stap+/etc.
• Neglectacble tracing overhead; suitable for production use
• Already helping big companies like Zoom, Inc
• Transparently peeking inside already running processes and containers
Text version: blog.openresty.com/en/erlang-high-cpu/
#openresty #erlang #performance #ebpf #observability -
OpenResty
@OpenResty
🔍 New video tutorial: Using OpenResty XRay to diagnose high CPU usage in Erlang/OTP applications: www.youtube.com/watch?v=sqB5fGHahYU
Highlights:
• Pinpoint CPU-intensive regex operations, among many other bottlenecks
• Analyze code paths on both the Erlang language level and the VM C/C++ level (with Flame Graphs!)
• Zero-intrusion profiling - no restarts, no special options/flags, and no code changes needed!
• Based on our advanced dynamic tracing tech eBPF+/Stap+/etc.
• Neglectacble tracing overhead; suitable for production use
• Already helping big companies like Zoom, Inc
• Transparently peeking inside already running processes and containers
Text version: blog.openresty.com/en/erlang-high-cpu/
#openresty #erlang #performance #ebpf #observability -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Some users asked whether our @OpenResty XRay dynamic tracing technology relies on JVM frame pointer preservation. Of course, we don't have such a requirement—we've always pursued a strict zero-modification, zero-cooperation approach. Preserving frame pointers has a significant performance impact. I know some third-party open-source solutions and even commercial solutions require frame pointer preservation, which is not suitable for production, obviously. #dynamictracing #jvm
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Support for OpenJDK 21 is also coming! These analyzers are implemented in our Y++ language toolchain: blog.openresty.com/en/cpp-trace-cpp/
OpenResty
@OpenResty
🔍 Deep dive into Java application analysis with OpenResty XRay! See it in our new blog post: blog.openresty.com/en/java-analysis/
Key features:
• Zero-modification profiling using eBPF/Stap+/GDB/etc
• Precise CPU hotspot detection to the Java source line level.
• Off-CPU blocking analysis
• Disk I/O tracing
• No Agent or Safepoints in JVM.
• Support for unmodified OpenJDK 8/11/17
Will add support for commercial JVMs like Oracle JDK and Zulu JDK as well. Stay tuned!
#java #ebpf #performance #observability -
OpenResty
@OpenResty
🔍 Deep dive into Java application analysis with OpenResty XRay! See it in our new blog post: blog.openresty.com/en/java-analysis/
Key features:
• Zero-modification profiling using eBPF/Stap+/GDB/etc
• Precise CPU hotspot detection to the Java source line level.
• Off-CPU blocking analysis
• Disk I/O tracing
• No Agent or Safepoints in JVM.
• Support for unmodified OpenJDK 8/11/17
Will add support for commercial JVMs like Oracle JDK and Zulu JDK as well. Stay tuned!
#java #ebpf #performance #observability -
OpenResty
@OpenResty
Thanks @kong, the API gateway company, for making a donation to OpenResty lately! Many years ago, Kong switched from NodeJS to OpenResty and saw dramatic improvement in performance, "more than times faster." #kong #openresty #nodejs
-
OpenResty
@OpenResty
🚀 OpenResty 1.27.1.1 is out now! Major updates include:
• Nginx core upgraded to 1.27.1
• LuaJIT updated to 2.1-20240815
• OpenSSL upgraded to 3.0.15
• PCRE upgraded to 10.42
• New features in lua-nginx-module and stream-lua-nginx-module
• Performance optimizations in lua-resty-redis
Full changelog: openresty.org/en/changelog-1027001.html
Download: openresty.org/en/download.html
#openresty #webserver #nginx #luajit -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
When we develop new dynamic tracing analyzers in @OpenResty XRay, we usually start by directly copying the C/C++ code from the target programs into our Y/Y++ code, such as CPython, JVM, PHP, etc. Generally, we still need to make minor edits to minimize copying excessive runtime dependency code. Now, even these last bits of manual editing may be automated through #claude 3 Opus.
Just yesterday, I used Claude 3 Opus to rewrite a large C function, and surprisingly, the resulting code was nearly production-quality. The diff patches compared to the original code looked perfect—I was impressed. In contrast, when I assigned the same task to #chatgpt’s GPT-4, the results were disappointing, with some obvious errors. It felt like comparing a professional’s work to that of an elementary school student.
I also saw a post on Reddit where someone mentioned using Claude 3 Opus to refactor a large software project with many files, and he found it significantly better than GPT-4. He commented that ChatGPT often imagines things that don’t exist. #llm #ai -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I've been leading some engineers in our team to focus on developing #jvm deep analysis tools based on non-invasive #dynamictracing technologies like #ebpf. We aim to release the first version this quarter. The tools are particularly suited for online production environments as they have minimal impact, unlike tools like BTrace and Alibaba's Arthas, which can significantly disrupt online #java applications and pose high risks. They also operate independently of the JVM Safepoints, providing a comprehensive analysis with no blind spots.
When ready, @OpenResty XRay will integrate these tracing tools into our AI diagnostic reasoning engine and analysis report generator. The initial release will support OpenJDK, and there is potential for future support for Oracle's proprietary JDK. This approach avoids using mechanisms like JVM's Agent for code injection, enabling a completely read-only analysis of Java services already running online, akin to taking an X-ray.
The full dynamic probes at Java methods' entry and exit points are just one of many tracing features, including JIT-compiled code. Additionally, the tool will offer CPU hotspot analysis, IO blocking and lock contention analysis, memory usage analysis, GC analysis, exception analysis, thread scheduling analysis, latency analysis, targeted smart packet capture for network traffic, and tracking of Java data structure states.
We can also support older systems like CentOS/RHEL 7 with 3.10 kernels and even ancient versions of JVM, as our technology does not rely on any cooperation or assistance from the JVM itself. This is beneficial since the software used by enterprise customers tends to be quite old. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
This opens the door to writing complex, nonintrusive dynamic tracing tools for big C++ applications like HotSpot JVM, NodeJS, and MySQL, with strictly 0 collaborations from the target process :D How can it be easier than writing C++ to trace C++? ;)
Yichun Zhang (agentzh)
@agentzh
I just wrote a new blog post, "Use C++ to Dynamic-Trace C++ Applications": blog.openresty.com/en/cpp-trace-cpp/. It can compile real C++ code into noninvasive dynamic tracing analyzers for eBPF+, Stap+, GDB, and more! It has been a big leap since I was brought into this tracing world by @brendangregg 's blog posts more than a decade ago :) #ebpf #stap
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I just wrote a new blog post, "Use C++ to Dynamic-Trace C++ Applications": blog.openresty.com/en/cpp-trace-cpp/. It can compile real C++ code into noninvasive dynamic tracing analyzers for eBPF+, Stap+, GDB, and more! It has been a big leap since I was brought into this tracing world by @brendangregg 's blog posts more than a decade ago :) #ebpf #stap
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Recently I spent a day or two on #android 14 and finally passed the full unit test suite of eBPF+.
On Android 14, the ELF files compiled with the NDK toolchain have a difference of 0x1000, or 4096 bytes, between the relative memory address and the file offset, which is quite strange. Android's OS #kernel, compiled with #clang, also has more runtime checks, such as CFI. But the kernels of mainstream Linux distributions are almost all compiled with gcc, without such extra checks.
To learn more about our eBPF+ and Ylang in OpenResty XRay: blog.openresty.com/en/ylang-intro-part1/
#android #ebpf #dynamictracing -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
One night recently, I spent about ten minutes and successfully ported our eBPF+ engine to #android 14. eBPF+ is an enhanced version of eBPF that we use in our OpenResty XRay product. The simplest two test cases for kprobes and uprobes have already passed on Android 14.
The implementation of Android is almost identical to the #linux kernel, so it was very smooth. It looks like it can also be easily compatible with older versions of Android down to Android 10. The next step is to run more eBPF+ unit tests until we run the huge test suites of our compilers, such as #ylang, #ylua, and #ysql.
The difference in Android user space is bigger, such as using bionic libc, the header file of epoll is quite different from Linux, and it also enforces the use of PIE executable files.
To learn more about our eBPF+ and Ylang: blog.openresty.com/en/ylang-intro-part1/
#ebpf #dynamictracing #fullstackanalysis #noninvasiveanalysis #android -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The vision for our OpenResty XRay product is: for any customer's software business system in a production environment, including multi-cluster distributed systems, OpenResty XRay can thoroughly and transparently analyze its entire software stack.
This includes hardware to the OS kernel, the network protocol stack, system software, and the business software layer. Metaphorically speaking, we aim to transform these systems from black boxes into transparent ones. It's like taking an X-ray or MRI of the production system, automatically identifying all underlying issues and causes related to performance, security, stability, or correctness.
Importantly, these systems and applications require no modifications. There's no need to add any libraries or modules, no need for logging or metrics data, and even process restarts are unnecessary. All that's needed is a standalone Agent daemon of OpenResty XRay running on each machine to analyze all applications and containers.
Our goal is to tackle one of the greatest challenges in the computing world: the increasing complexity of systems gradually diminishing human insight and control over software and hardware systems.
#ebpf #dynamictracing #fullstackanalysis #realtimeanalysis -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Yes, we have many! blog.openresty.com/en/xray/ Though eBPF is just one of our dynamic tracing backends. We also have Stap, GDB, ODB, etc.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We already have C2eBPF (not the existing eBPF LLVM C, which is lame! Here it is our ylang), C++2eBPF, Lua2eBPF (or ylua), and SQL2BPF (or ysql), in production! :D Planning Go, Python, Rust, etc., and JS too, of course! See blog.openresty.com/en/ylang-intro-part1/ for details.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
A netizen inquired about the oldest #linux systems and kernels that the OpenResty XRay dynamic tracing platform can support. Our platform can support systems as old as CentOS/RHEL 6, used with the CentOS/RHEL 7's 3.10 kernels. The 2.6 kernels are just too old to support.
For Ubuntu, the oldest supported version is 16.04, and for Debian, it's version 9. We can also support even older versions like Ubuntu 14.04 and Debian 8 if necessary.
In contrast, the open-source eBPF toolchain usually has much higher kernel requirements, severely limiting its practicality in many enterprises. #ebpf -
OpenResty
@OpenResty
RT @agentzh: A major technical challenge involves automatically #generating debug symbols for Linux kernel images lacking #debuginfo. We ca…
-
OpenResty
@OpenResty
#ylang is our universal dynamic tracing language that is compatible with C. It is employed in the OpenResty XRay product, which includes a Web #ide. This IDE allows users to use Ylang for writing and running various dynamic tracing analysis tools.
For instance, a demonstration screenshot shows the use of Ylang to create a custom dynamic tracing tool. This tool places a dynamic probe on the `ngx_http_finalize_request` function of #nginx, printing information such as the URI, URI parameters, and response status codes of up to five requests when the probe is triggered.
The screenshot displays the entered Ylang source code (seen in the top-left editing box), the configuration of the application being traced (top-right), and the output after running it online (in the black box at the bottom-left).
Ylang can be executed through various backends, including eBPF+, making it suitable for online production environments. The target application being traced does not require any modifications or restarts, ensuring a 100% non-intrusive process.
Compared to other open-source dynamic tracing languages like #bcc, #ebpf C, #bpftrace, D language of #dtrace, #systemtap's scripting language, and #gdb python, Ylang is significantly more concise and powerful.
For more details about Ylang, refer to a series of articles I previously published: blog.openresty.com/en/ylang-intro-part1/.
#ebpf -
OpenResty
@OpenResty
We’ve used #ebpf/#xdp to implement fully functioning #dns authoritative servers! It's part of our @OpenResty #ddos product. We made our eBPF programs return DNS replies directly. Achieving 8M+ PPS on a single desktop Linux machine ;) Supporting older kernels like RHEL 8's 4.18 kernels.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Come on, this is easy. We already use #ebpf/#xdp to implement fully functioning #dns authoritative servers! (it's part of our @OpenResty #ddos product). We made our eBPF programs return DNS replies directly. Achieving 8M+ PPS on a single desktop Linux machine ;) Supporting older kernels like RHEL 8's 4.18 kernels.
Dan 🐝
@thebsdbox
So yeah, local DNS A records being re-written with eBPF on the fly.
-
OpenResty
@OpenResty
OpenResty Inc. is excited to announce the recent launch of the Japanese version of our official website: openresty.com/jp/
Since our inception, we have been committed to internationalization and globalization. We extend our heartfelt thanks to our corporate users around the world for their strong support over the years! #japaneseworldwide -
OpenResty reposted
Manish Gill
@mgill25
Got into Lua with @OpenResty but your Tutorials with Love2D made me love it and reminded me the joy of coding. Can't thank you enough.
-
OpenResty
@OpenResty
We have recently compiled a newsletter for the OpenResty Edge distributed gateway and private CDN software products. You can read it here: openresty.com/en/news-edge-2023-12/. This newsletter summarizes the latest articles and video tutorials published about this product. #k8s #distributedgateway #waf #privatecdn
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The #linux and #android kernels do not have a stable Application Binary Interface (ABI). Even with identical kernel versions and compilation parameters, using different versions of C compilers can change the memory layout of the structs involved in the API, resulting in #incompatibility with binary kernel modules (.ko files). Furthermore, the compiler security feature introduced in recent years, which #randomizes the order of struct fields, leads to incompatibilities even with identical environments and toolchains, as each kernel compilation can result differently.
I am designing a machine learning (#ml) algorithm capable of automatically learning the machine instructions and data segments within a current binary kernel image. It then modifies the kernel modules' source code or binary .ko files to #adapt them to the target system's kernel. This approach can even learn and adapt to struct field randomization, significantly reducing complications. Essentially, this is a form of AI-driven reverse engineering. Our OpenResty XRay product employs a similar proprietary technology for automatically generating debugging symbols (or #debuginfo for stripped binary programs).
Similar technology could be applied to user space .so libraries for adaptation, but these usually have a stable ABI and are not as prone to the variability seen with Linux or Android kernels. -
OpenResty
@OpenResty
We recently released version 23.12.12 of the OpenResty XRay product. The release announcement can be found here: openresty.com/en/ann-xray-23-12-12-1/.
Key highlights include the addition of "smart packet #capture" support for application-layer requests with prolonged delays. This feature captures network packets only on connections with actual issues in the web applications, making it suitable for high-traffic production environments. There's also support for #almalinux, a popular CentOS alternative, and dynamic tracing support for #openssl 3.x and #jemalloc 5.x. Moreover, the update brings program anomaly detection and analysis for #perl applications.
Additionally, we have significantly optimized our machine learning (#ml) based feature that automatically reconstructs debugging symbols (or #debuginfo) for binary programs lacking them. This enhancement has led to an order-of-magnitude improvement in memory usage and CPU time.
#ebpf #dynamictracing #nonintrusive #performanceanalysis #securityscanning -
OpenResty
@OpenResty
We published a new video tutorial: youtu.be/HyrI2p5fE1M. It demonstrates how to install OpenResty XRay's Agent service on Ubuntu. In essence, It only takes one command to install it.
This is an independent daemon process that can automatically detect and analyze all kinds of applications on the current system, whether they are containerized or not, such as OpenResty/Nginx, Lua, Python, PHP, Perl, Redis, Go, Rust, C/C++, Erlang, and so on. The target applications and containers do not need any modification or cooperation at all.
Our Agent also supports installing our container image through #k8s or #docker, which we will demonstrate in a future video. Of course, we also support other distributions through rpm/deb packages.
#ebpf #dynamictracing #profiling #onlinetroubleshooting -
OpenResty
@OpenResty
We have published a new video tutorial: youtu.be/5WrUqkhzsXo. It demonstrates how to create new page rules through the web interface of OpenResty Edge to dynamically add new custom HTTP response #headers to the gateway.
OpenResty Edge is a distributed traffic management system that we provide, which can be used for building enterprise traffic entrances, #k8s cluster complex traffic management, and private CDN networks. #privatecdn #gateway
Text version: blog.openresty.com/en/edge-resp-header/ -
OpenResty
@OpenResty
We recently published a new demo video youtu.be/bjmLEtwvIZI. It shows how OpenResty XRay uses dynamic tracing technology to analyze the extra performance overhead introduced by a typical Rust target application using the `top` command.
We can see that the overhead of the OpenResty XRay analyzer itself and the impact on the target application’s performance is almost negligible. This makes it very suitable for online use, even for production environments with high service throughput and latency requirements. #ebpf #stap
Text version: blog.openresty.com/en/rust-xray-overhead/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We've put a lot of effort into reducing the #overhead of OpenResty XRay itself. After all, many of our big customers are running it on production and are very sensitive to overhead. The most sensitive one is a #cryptocurrency trading company, which is very sensitive to latency. They have a very large trading volume.
We always carefully check the sources of all extra latency and CPU overhead, while the open-source dynamic tracing toolchain usually does not care so much and often introduces very large latency or CPU loss here and there. Of course, sometimes we also contributed some of our patches to the open-source community, though that is just a small part of our many optimizations.
OpenResty
@OpenResty
We recently published a new demo video youtu.be/bjmLEtwvIZI. It shows how OpenResty XRay uses dynamic tracing technology to analyze the extra performance overhead introduced by a typical Rust target application using the `top` command.
We can see that the overhead of the OpenResty XRay analyzer itself and the impact on the target application’s performance is almost negligible. This makes it very suitable for online use, even for production environments with high service throughput and latency requirements. #ebpf #stap -
OpenResty
@OpenResty
We recently published a new demo video youtu.be/bjmLEtwvIZI. It shows how OpenResty XRay uses dynamic tracing technology to analyze the extra performance overhead introduced by a typical Rust target application using the `top` command.
We can see that the overhead of the OpenResty XRay analyzer itself and the impact on the target application’s performance is almost negligible. This makes it very suitable for online use, even for production environments with high service throughput and latency requirements. #ebpf #stap -
OpenResty
@OpenResty
GSLB (Global Server Load Balancing) can use DNS dynamic resolution to automatically redistribute traffic to other nearby servers or even other nearby clusters based on the load and other custom thresholds on the gateway servers.
We recently published a new video tutorial: youtu.be/Gj5wibhAy5E. It shows how to configure GSLB in OpenResty Edge and view the real-time effects of the new traffic redistribution. OpenResty Edge’s GSLB feature can work closely with its built-in DNS authoritative server. #gslb #privatecdn #distributedgateway
Text version: blog.openresty.com/en/edge-gslb/ -
OpenResty
@OpenResty
This is a simple example of writing a dynamic tracing analyzer for an arbitrary #go target program using the C language style #ylang language.
Here, we directly use the web #ide of OpenResty XRay to edit and run the user-defined Ylang analyzer in real time and then view the results. We also mainly use Ylang to write analyzers for different technical stacks. The full demo video: youtu.be/WUdzsp9WMcU #golang -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
This is equivalent to using the C language to trace #go target processes. Because of the differences in language notation, we have extended the C-style #ylang to support directly referencing the various strange type names and other symbol names in Go.
In the future, we might implement a #ygo language compatible with Go's syntax and then compile it into Ylang with a dedicated compiler, which would be more convenient, just like our existing #ylua and #ysql languages.
OpenResty
@OpenResty
We published a new video tutorial: youtu.be/WUdzsp9WMcU It demonstrates how to write a simple Ylang analyzer, dynamically tracing any #go program process. This custom analyzer will add a dynamic probe at a specified #golang function entry and then print out the value of a map type Go variable. Thanks to our dynamic tracing technology, these analyzers can work in a true non-invasive way. This way, we can minimize the performance overhead without changing the target Go programs or processes.
The dynamic tracing analyzer in Ylang here is compiled into multiple different backends, such as eBPF+, Stap+, and GDB Python, for different analysis scenarios. #ebpf
Ylang is a universal dynamic tracing language that we designed and implemented and is compatible with C language. For details, please see the series of introduction articles I wrote before: blog.openresty.com/en/ylang-intro-part1/
Text version of this tutorial: blog.openresty.com/en/ylang-dump-go-val/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
One of the most common uses of hardware data breakpoints is to combine them with reverse execution to find the context where the memory overwrites last happened. This is often the first crime scene for memory corruption issues. This is usually done with gdb or Mozilla #rr. I have used rr to track down dozens of very deep bugs hidden for years in the #luajit JIT compiler. Yet I have another wonderful use case for this feature in #systemtap and OpenResty XRay, and I’d keep it a secret for now :)
-
OpenResty
@OpenResty
We published a new video tutorial: youtu.be/WUdzsp9WMcU It demonstrates how to write a simple Ylang analyzer, dynamically tracing any #go program process. This custom analyzer will add a dynamic probe at a specified #golang function entry and then print out the value of a map type Go variable. Thanks to our dynamic tracing technology, these analyzers can work in a true non-invasive way. This way, we can minimize the performance overhead without changing the target Go programs or processes.
The dynamic tracing analyzer in Ylang here is compiled into multiple different backends, such as eBPF+, Stap+, and GDB Python, for different analysis scenarios. #ebpf
Ylang is a universal dynamic tracing language that we designed and implemented and is compatible with C language. For details, please see the series of introduction articles I wrote before: blog.openresty.com/en/ylang-intro-part1/
Text version of this tutorial: blog.openresty.com/en/ylang-dump-go-val/ -
OpenResty
@OpenResty
We recently published a new #video tutorial that explains how to manage different login accounts and roles in OpenResty Edge’s web console: youtu.be/02nlPUdclvQ
This feature is especially useful for administrators, as they can precisely control the configuration permissions of different roles and accounts, greatly facilitating the collaboration between different departments and teams within the enterprise. For example, at qunar.com, the business department has reading and writing permissions for the configuration but not the release permission, while the operation and maintenance department can execute the release and publishing operations after reviewing the configuration changes.
OpenResty Edge is a distributed traffic management system that we provide, which is suitable for building enterprise traffic entrances, managing complex traffic in K8s clusters, and setting up private #cdn networks. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I recently added support for hardware data watchpoint probes for user-space processes to the open-source #systemtap, which allows defining such dynamic probes in dynamic tracing scripts and performing custom actions (such as statistical analysis) when the target application reads or writes (or only writes) a certain memory area. I have submitted this patch to the SystemTap official Git repository, which will be included in the next release. We also ported this patch to the Stap+ backend of the @OpenResty XRay product, which can be used later to implement some interesting advanced analyzers.
The following figure shows the corresponding Stap script test case, demonstrating its simplest usage. The variable `$^ADDR_a` in the test case will be automatically replaced by the test bench with the global variable "a"'s relative address in the target C program. This test case uses the Test::Base test bench test data description notation, which should be self-explanatory. It is worth mentioning that the official test cases of SystemTap are expressed in Expect/Tcl language, which is much more disgusting than my Test::Base description here, so I wrote a Perl script that automatically converts the test cases expressed in Test::Base into Tcl code for the SystemTap test suite, saving me a lot of trouble. Writing programs that write programs is one of the favorite tricks of experienced programmers. -
OpenResty
@OpenResty
We have recently added an official binary package repository for #openresty on the Amazon Linux 2023 system: openresty.org/en/linux-packages.html. Our OpenResty XRay product also supports the latest version of this distribution.
One annoying thing about Amazon Linux 2023 is that it no longer guarantees binary compatibility with Red Hat Linux (although it looks like RHEL 9), so the #epel package repository cannot be used directly. This is a big step back compared to Amazon Linux 1 and 2. Alas. #aws #amazonlinux -
OpenResty
@OpenResty
We just released the new OpenResty 1.25.3.1 version, which you can download here: openresty.org/en/download.html It is based on the latest #nginx 1.25.3 core, and upgraded #luajit to the newest version. We also added support for HTTP/3 and PCRE2. The complete change log is here:
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I recently upgraded my ‘paralyzed office’ equipment, which is more comfortable with a slight tilt. It can lower eye pressure and prevent acid reflux. The whole ceiling in the room is a computer screen with 4K resolution. It feels like a cinema. The projector is XGIMI Horizon Pro 4K, bright enough, no need to draw the curtains. How do I type? Look closely; there are split mechanical keyboards on both sides. Will the mouse slip off? No! You can choose a mouse pad with more friction. The existing paint on the ceiling should be smooth enough (but not too glossy) with dark colors. Of course, it would be perfect to use projection-specific paint, but I’m too lazy to bother. After all, this kind of paint is expensive, and the area to be painted is also large, quite troublesome.
Sitting for a long time hurts the lumbar, shoulder, and cervical spine; standing for a long time also causes leg soreness and varicose veins, so it’s better to ‘lie flat’ with a little slope. My two chiropractors also appreciate and agree with this work equipment. Some netizens commented that if you put Professor Hawking here, he could also work comfortably, which makes sense…haha! -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The new Apple Vision Pro mixed-reality device, if combined with our OpenResty XRay’s three-dimensional “software universe” (or “softverse”), should provide a stunning experience. We can try this direction in the future 💪
-
OpenResty
@OpenResty
We recently released version 23.11.7-1 of our OpenResty XRay product. The release announcement is here: openresty.com/en/ann-xray-en-23-11-17-1/
The three biggest highlights are: 1) We have deeply optimized the latency introduced by the dynamic tracing analyzer (part of the optimization patches for the Stap+ backend have been ported and submitted to the SystemTap open source repository, and the eBPF+ backend also has important optimizations), and added a new Paranoid mode (mainly to meet the stringent requirements of a cryptocurrency trading company for extra online latency). 2) Added support for in-depth analysis of CPU-related issues for any Erlang application. 3) Added support for the Debian 12 target system.
OpenResty XRay is our online application analysis platform based on our dynamic tracing technology, which can non-invasively analyze various types of applications in real-time for high CPU, CPU not going up, memory too large or leaking, Disk I/O too high, long latency, anomalies, crashes, network, security risks and many other aspects of the problem, without touching the source code or making any changes to the target application. -
OpenResty
@OpenResty
We recently added new official binary packages for open-source OpenResty! The new repositories are for RHEL/CentOS 9, Debian 12 bookworm x86_64/arm64, Alpine 3.16 ~ 3.19 x86_64/arm64, and CBL-Mariner 2.0 arm64.
The corresponding documentation and download links are here: openresty.org/en/linux-packages.html
Our commercial products also added support for these new Linux distributions accordingly. -
OpenResty reposted
opdroid1234
@opdroid1234
Managed to figure out that @hulu uses @OpenResty 🤣
-
OpenResty
@OpenResty
We recently released a new security update, version 1.21.4.3, for our open-source OpenResty, which includes fixes for the new HTTP2 security vulnerabilities in Nginx. The release announcement is here: openresty.org/en/ann-1021004003.html We recommend that all users upgrade as soon as possible. Our commercial customers who use OpenResty Plus also have the relevant updates.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We have received some questions about whether our @OpenResty XRay can work with #embedded systems. The good news is: YES! However, embedded systems often run on specially compiled versions of the Linux kernel, so we have to tailor our tool to fit each user’s case. The bright side is that you only have to do this once. We are also planning to add support for Android kernels. Hooray for #dynamictracing!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Some netizens said that the #uretprobes of the Linux kernel would cause the Go runtime to panic and asked us if our OpenResty XRay would have any problems. The answer is no! 😁
OpenResty XRay uses its own private technology to avoid using uretprobes to implement dynamic probes for function returns.
In fact, the design of the uretprobes feature in the kernel itself has a lot of problems because it modifies the target program’s runtime stack, breaking the most critical non-invasive principle of dynamic tracing. Not only will Go have difficulties, but C/C++ and other programs may also have stability issues such as crashes. -
OpenResty
@OpenResty
A new OpenResty XRay #newsletter was issued, featuring all the video tutorials and blog posts about the OpenResty XRay dynamic tracing platform product in the past month: openresty.com/en/news-xray-en-10-2023/
OpenResty XRay allows non-intrusive analysis of different stability, performance, and security issues for any application process online, without the need for the application’s source code or any changes to the application. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Recently I spent two days researching how to add dynamic-tracing support for any Erlang/OPT application to OpenResty XRay. I mainly work with #erlang's BEAM virtual machine.
In many ways, Erlang’s implementation is similar to dynamic-typed languages like Lua. For example, Erlang’s VM is also mainly implemented in C, similar to LuaJIT/Python/Perl/PHP and others. Another similarity is that Erlang’s lightweight processes are somewhat like OpenResty’s “light threads” based on Lua coroutines. Furthermore, all memory objects and values in Erlang are uniformly represented by the eTerm data type, while in LuaJIT, they are TValue. Erlang’s BEAM virtual machine also has a JIT compiler. Still, it seems to be a simple bytecode translation without any cross-instruction optimization, while LuaJIT’s code optimization is almost entirely in its JIT compiler.
In addition, Erlang also has garbage collection (GC), and it is per lightweight process, which is very impressive. In contrast, other VMs’ GCs are mostly global, such as LuaJIT and Go’s GC. Each operating system thread in Erlang also has its own independent lightweight process scheduler. Of course, Erlang also has other unique features, such as being a dynamic-typed and strong-typed language simultaneously.
I recall that in 2010, when I was at Taobao, I wrote a production service in Erlang. But back then, I noticed that Erlang’s official implementation was mediocre in performance, far behind my own OpenResty, even for the simplest HTTP server implementation. Of course, over the years, Erlang must have made some progress, such as having an ASM JIT compiler, although it is only a bytecode-level translation. -
OpenResty
@OpenResty
Some netizens asked us whether our OpenResty XRay #dynamictracing product supports Microsoft’s #windows operating system. The answer is that the native Windows operating system is not yet supported, but maybe in the future, we can enhance the official dtrace port driver from Microsoft. The WSL subsystem running Linux on Windows should be easy to support though.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
In this example, we will see that the dynamic memory allocation overhead of the libc functions (such as malloc/realloc) in #rust's sled library is still relatively high, and there is much room for optimization.
OpenResty
@OpenResty
OpenResty XRay can also automatically analyze unmodified #rust applications! We have published a new video tutorial: youtu.be/Odu2tTZyDU0
We will show you how to use OpenResty XRay to quantitatively analyze the CPU time consumption of the sled library in Rust. We will show you the hottest Rust code paths that consume the most CPU. These hot code paths are obtained by OpenResty XRay automatically analyzing and interpreting the Rust-land CPU #flamegraphs. Text version: blog.openresty.com/en/sled-high-cpu/
The Rust code paths we report can be precise to the code line, and OpenResty XRay will provide detailed Rust source file names and line numbers. Also, all the inlined Rust functions will be restored in the code paths. The target program and process do not need any modification and are 100% non-invasive. -
OpenResty
@OpenResty
OpenResty XRay can also automatically analyze unmodified #rust applications! We have published a new video tutorial: youtu.be/Odu2tTZyDU0
We will show you how to use OpenResty XRay to quantitatively analyze the CPU time consumption of the sled library in Rust. We will show you the hottest Rust code paths that consume the most CPU. These hot code paths are obtained by OpenResty XRay automatically analyzing and interpreting the Rust-land CPU #flamegraphs. Text version: blog.openresty.com/en/sled-high-cpu/
The Rust code paths we report can be precise to the code line, and OpenResty XRay will provide detailed Rust source file names and line numbers. Also, all the inlined Rust functions will be restored in the code paths. The target program and process do not need any modification and are 100% non-invasive. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
One of our customers' office in Gaza was lost in the ongoing war and their staff are moving, so they can’t do business with us for now. Pray for them, and hope for peace to be restored there soon…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
In this example, we see the high overhead of creating #goroutines in the #etcd server, and the parsing and generation overhead of the #grpc protocol is still relatively high. In contrast, the creation overhead of Lua coroutines in #luajit is lower, and the OpenResty core also recycles Lua coroutine objects through a coroutine pool.
OpenResty
@OpenResty
New video tutorial: youtu.be/7LHVmehTxWc We will demonstrate how to use OpenResty XRay to see how CPU time is distributed inside #golang's #etcd servers. You can see which Go code paths are the hottest and what they do. OpenResty XRay can automatically generate and interpret Go-land CPU #flamegraphs without modifying the online etcd processes. This is an entirely non-invasive approach.
Text version: blog.openresty.com/en/etcd-cpu-distr/ -
OpenResty
@OpenResty
New video tutorial: youtu.be/7LHVmehTxWc We will demonstrate how to use OpenResty XRay to see how CPU time is distributed inside #golang's #etcd servers. You can see which Go code paths are the hottest and what they do. OpenResty XRay can automatically generate and interpret Go-land CPU #flamegraphs without modifying the online etcd processes. This is an entirely non-invasive approach.
Text version: blog.openresty.com/en/etcd-cpu-distr/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The C++ code paths we report can be precise to the C++ code line. OpenResty XRay will provide the C++ source file name and line number. Also, the inlined C++ functions will be expanded in the code path.
OpenResty
@OpenResty
OpenResty XRay can not only analyze regular server applications but also the performance bottlenecks of #generativeai and #llm! We recently released a new video tutorial: youtu.be/dXTCdUb5AGk It shows how to use OpenResty XRay to analyze the CPU time distribution inside the llama.cpp program that runs the #llama2 model. Text version: blog.openresty.com/en/llama-high-cpu/
-
OpenResty
@OpenResty
OpenResty XRay can not only analyze regular server applications but also the performance bottlenecks of #generativeai and #llm! We recently released a new video tutorial: youtu.be/dXTCdUb5AGk It shows how to use OpenResty XRay to analyze the CPU time distribution inside the llama.cpp program that runs the #llama2 model. Text version: blog.openresty.com/en/llama-high-cpu/
-
OpenResty
@OpenResty
We recently released a new video tutorial on installing the standalone Agent service of OpenResty XRay on CentOS 8 using the Yum/RPM package repository method: youtu.be/lM3IJRcCmvc Text version: blog.openresty.com/en/xray-agent-install-centos8/
OpenResty XRay’s Agent is an independent daemon that runs entirely separately from the target application or container being analyzed. It can perform full-stack analysis (from the network protocol stack and Linux kernel to the system software layer and then to the business code layer, with no blind spots across layers). -
OpenResty
@OpenResty
Our OpenResty XRay now also supports ARM64 Linux systems, and a lot of offline machine-learning computation is currently running on this Apple Silicon device.
Yichun Zhang (agentzh)
@agentzh
Recently, my #arm64 server at home broke down completely. The two network cards and the only graphics card on the motherboard all stopped working. There was no place to buy the broken parts or the motherboard. I decided to buy a #macstudio #m2 Ultra machine as a server. On it, I ran an Aarch64 Linux virtual machine using the latest version of qemu compiled from source code with hvf hardware acceleration. I found that the qemu that came with homebrew had a problem. The CPU performance in the virtual machine would drop by several times after a while. It could be scheduled to run on the E-Core. I also used Docker Station to connect several large external hard drives via USB-C. Unfortunately, the USB 3.1 passthrough driver of qemu had a bug, and the USB 2 protocol was too slow. So, I had to settle for using qemu’s virtual hard disk image files (raw image and qcow2).
-
OpenResty
@OpenResty
New video tutorial on YouTube: youtu.be/1R7FsVJbm4s We show you step by step how to use OpenResty XRay to automatically analyze unmodified Go processes online, quickly locate the #golang code paths that cause low CPU utilization, and pinpoint them to the code line level.
OpenResty XRay can automatically interpret and analyze off-CPU flame graphs.
Our dynamic tracing technology does not require modifications to the Go application or Go processes: 100% true non-invasive.
Text version: blog.openresty.com/en/go-low-cpu/ -
OpenResty
@OpenResty
We recently compiled a newsletter for our OpenResty Edge, a distributed gateway and private CDN software: openresty.com/en/nl-edge-en-07-2023/. It summarizes the articles and video tutorials published lately about this product.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We are preparing a new video tutorial showing you how to write your own dynamic analyzer in our Y language for any #golang target program using the Web IDE of OpenResty XRay. This way, you can see how the state-of-the-art dynamic tracing technology works on a lower level and how to use our powerful ylang toolchain to customize new analyzers. Stay tuned!
-
OpenResty
@OpenResty
OpenResty XRay can also analyze any #perl application in a non-invasive way. As always, you do NOT need to install any special Perl modules, use any particular Perl interpreters, or restart the Perl processes already running online (for a long time).
We published a new video tutorial to YouTube: youtu.be/yH2OxDTgNNc
This video walks you step by step on how to quickly locate the most CPU-intensive Perl hot code paths online, with line-level accuracy.
OpenResty XRay automatically analyzes the Perl language-level and VM C-level CPU #flame_graphs and uses a large language model (LLM) to automatically interpret these code paths and functions and provide optimization suggestions. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OpenResty XRay analyzes any #golang apps with our advanced dynamic tracing technology. The target Go app does not need any code changes, modules, restarts, or special compilation options. It is true 100% non-invasive and does not rely on Go’s own profiling mechanisms. It can also transparently analyze Go apps in ordinary containers. It’s real postmortem analysis.
OpenResty
@OpenResty
Want to learn how to use OpenResty XRay to analyze any #golang app’s performance non-invasively? Watch our new video tutorial: youtu.be/2fAXn2144lw It guides you through analyzing an online Go process without modifying it, finding the most CPU-hungry golang code paths and source lines with OpenResty XRay’s auto-explained CPU #flame_graphs. Text version: blog.openresty.com/en/go-high-cpu/
-
OpenResty
@OpenResty
Want to learn how to use OpenResty XRay to analyze any #golang app’s performance non-invasively? Watch our new video tutorial: youtu.be/2fAXn2144lw It guides you through analyzing an online Go process without modifying it, finding the most CPU-hungry golang code paths and source lines with OpenResty XRay’s auto-explained CPU #flame_graphs. Text version: blog.openresty.com/en/go-high-cpu/
-
OpenResty
@OpenResty
We recently made another slide deck: openresty.com/en/pdf/openresty-edge-overview.pdf It systematically introduces the OpenResty Edge distributed gateway and private CDN software. We also added a lot of case studies near the end.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Red Hat is very diligent in porting new patches to old kernels. Although the users can enjoy new features, optimizations, and fixes in older kernels, one example illustrated in this article is that the original kernel had no problem. Still, the patch backported from the new kernel introduced a serious data race bug. Red Hat’s proud patch backport service for enterprise users can be risky.
OpenResty
@OpenResty
We just published a new blog post: “Two bugs in Linux kernels’ tracing subsystem (found by OpenResty XRay)”: blog.openresty.com/en/kernel-tracing-bugs-found-by-xray/
In the development and testing of OpenResty XRay, we often find many bugs in open source software and tools. After all, OpenResty XRay elevates dynamic tracing technology to a whole new level. Our analysis tools for various technology stacks and software are much more complex than those open source tools for SystemTap, DTrace, and eBPF. Therefore, our test suite often puts unprecedented pressure on the Linux kernel, eBPF, LLVM, SystemTap, GDB, and other components. High pressure naturally catches deeper bugs. -
OpenResty
@OpenResty
We just published a new blog post: “Two bugs in Linux kernels’ tracing subsystem (found by OpenResty XRay)”: blog.openresty.com/en/kernel-tracing-bugs-found-by-xray/
In the development and testing of OpenResty XRay, we often find many bugs in open source software and tools. After all, OpenResty XRay elevates dynamic tracing technology to a whole new level. Our analysis tools for various technology stacks and software are much more complex than those open source tools for SystemTap, DTrace, and eBPF. Therefore, our test suite often puts unprecedented pressure on the Linux kernel, eBPF, LLVM, SystemTap, GDB, and other components. High pressure naturally catches deeper bugs. -
OpenResty
@OpenResty
OpenResty XRay can also analyze any #perl application in a non-invasive way. As always, you do NOT need to install any special Perl modules, use any particular Perl interpreters, or restart the Perl processes already running online (for a long time). We published a new video tutorial to YouTube: youtu.be/bAhypg4zlfk This video walks you step by step on how to quickly locate the most CPU-intensive Perl hot code paths online, with line-level accuracy. OpenResty XRay automatically analyzes the Perl language-level and VM C-level CPU #flame_graphs and uses a large language model (LLM) to automatically interpret these code paths and functions and provide optimization suggestions.
-
OpenResty
@OpenResty
OpenResty XRay has helped many busy websites, such as Zoom, Microsoft, Bilibili, Qunar, etc., quickly identify the strange issues in production and development environments. The underlying implementation uses our advanced eBPF+, Stap+, ODB, and other dynamic tracing runtimes and Ylang language. Try it out for free today! openresty.com/en/xray/request-demo/?src=x
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OpenResty XRay was designed to trace any software from day one. It was never meant to be limited by target applications built on OpenResty. The "OpenResty" word in the product name just means that 1) it was built by OpenResty Inc. and 2) it was implemented atop OpenResty itself :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Yeah, OpenResty XRay has got Python support! Happy tracing Python code in the wild! Tracing support for Perl, PHP, and Go is coming very soon :)
OpenResty
@OpenResty
OpenResty XRay is a non-invasive #dynamic_tracing product that can analyze the performance issues of any #python application. It does not require any special Python modules or special Python interpreters. We just prepared a video tutorial on YouTube: youtu.be/nsePw-CA1Jg It shows how to use OpenResty XRay to quickly identify the Python code paths that cause low CPU utilization. It also has subtitles.
This tutorial demonstrates step by step how to automatically analyze the running state of an online Python web application without modifying it and find out the Python code paths and line numbers that block the Python VM and OS threads. These blocking IO or sleep operations prevent the Python application from increasing CPU usage and QPS. OpenResty XRay automatically recognizes and analyzes the off-CPU #flame_graphs at both the Python language and the Python VM C levels and finds the most blocking code paths and lines. We also have a text version of the tutorial: blog.openresty.com/en/python-low-cpu/ -
OpenResty
@OpenResty
OpenResty XRay is a non-invasive #dynamic_tracing product that can analyze the performance issues of any #python application. It does not require any special Python modules or special Python interpreters. We just prepared a video tutorial on YouTube: youtu.be/nsePw-CA1Jg It shows how to use OpenResty XRay to quickly identify the Python code paths that cause low CPU utilization. It also has subtitles.
This tutorial demonstrates step by step how to automatically analyze the running state of an online Python web application without modifying it and find out the Python code paths and line numbers that block the Python VM and OS threads. These blocking IO or sleep operations prevent the Python application from increasing CPU usage and QPS. OpenResty XRay automatically recognizes and analyzes the off-CPU #flame_graphs at both the Python language and the Python VM C levels and finds the most blocking code paths and lines. We also have a text version of the tutorial: blog.openresty.com/en/python-low-cpu/ -
OpenResty
@OpenResty
We recently created a slide deck that introduces our OpenResty XRay dynamic tracing system comprehensively: openresty.com/en/pdf/openresty-xray-introduction.pdf This system is like an X-ray that shows the internal state of complex systems in a genuinely non-invasive way. It helps developers and engineers quickly locate various problems in online applications, such as performance bottlenecks of CPU, memory, latency, file IO, etc., software errors and crashes, and security risks.
-
OpenResty
@OpenResty
OpenResty Edge can be used as an enterprise traffic entrance to manage internal enterprise traffic and quickly build a mighty private CDN. We have customers who are hosting-service providers who use OpenResty Edge to set up their own dynamic CDN network. Try it out for free today: openresty.com/en/edge/
-
OpenResty
@OpenResty
We published a new step-by-step video tutorial for OpenResty Edge: youtu.be/el2nlbLSof0 Remember to turn on subtitles. OpenResty Edge can add a request ID to the response header, which contains encrypted information about the edge server. It can also be used to automatically match the server-side error logs so that you can easily debug various problems of the current request. The text version is here: blog.openresty.com/en/edge-request-id/
-
OpenResty
@OpenResty
We just published a new blog post: blog.openresty.com/en/kong-plugin-mem-cpu-stats/. It introduces how to use OpenResty XRay to analyze the server processes of the Kong API gateway online, find out which Kong plugins consume the most CPU or memory in real-time, and further optimize the most expensive plugins at the Lua or C code level. Because it is 100% non-invasive dynamic tracing, there is no need to modify or adjust Kong’s nginx target processes.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We have plans to extend this capability of OpenResty XRay to other contexts like Nginx modules, Envoy extensions, PostgreSQL extensions, Perl/Python/Ruby modules and libraries, and many more.
OpenResty
@OpenResty
We just published a new blog post: blog.openresty.com/en/kong-plugin-mem-cpu-stats/. It introduces how to use OpenResty XRay to analyze the server processes of the Kong API gateway online, find out which Kong plugins consume the most CPU or memory in real-time, and further optimize the most expensive plugins at the Lua or C code level. Because it is 100% non-invasive dynamic tracing, there is no need to modify or adjust Kong’s nginx target processes.
-
OpenResty
@OpenResty
We just published a new blog post: blog.openresty.com/en/kong-plugin-mem-cpu-stats/. It introduces how to use OpenResty XRay to analyze the server processes of the Kong API gateway online, find out which Kong plugins consume the most CPU or memory in real-time, and further optimize the most expensive plugins at the Lua or C code level. Because it is 100% non-invasive dynamic tracing, there is no need to modify or adjust Kong’s nginx target processes.
-
OpenResty
@OpenResty
RT @agentzh: The following video tutorial demonstrates using the OpenResty Edge gateway to manage static website resources. OpenResty Edge…
-
OpenResty
@OpenResty
We published a new blog post showing how one of our customers used OpenResty XRay to analyze and fix a CPU bottleneck caused by frequent Lua exceptions in a custom Kong plugin: blog.openresty.com/en/lua-str-lower-excep/ The post demonstrates how OpenResty XRay automatically analyzes, diagnoses and reports the problem, and how the customer quickly modified the plugin code to optimize performance and avoid exceptions. #kong #dynamictracing #openrestyxray
-
OpenResty
@OpenResty
RT @agentzh: Kong is a 3rd-party API gateway built upon our open-source OpenResty platform. OpenResty XRay can analyze modified Kong server…
-
OpenResty
@OpenResty
OpenResty, a web platform for Lua development, now has a Yum package repository for CBL-Mariner 2.0 Linux, a distribution for Azure cloud and Microsoft’s own web apps: openresty.org/en/linux-packages.html Thanks to a Microsoft team for pushing this support.
-
OpenResty
@OpenResty
After months of hard work, we are proud to announce the new official release of OpenResty, the open-source web platform. This version includes many new features and improvements, which you can find in the full release announcement: openresty.org/en/ann-1021004002.html You can download this version from our official website: openresty.org/en/download.html or install it using our official binary package repository: openresty.org/en/linux-packages.html Thank you for your support and interest in OpenResty!
-
OpenResty
@OpenResty
We just made a new video with an AI digital actor showing you how to use OpenResty XRay. The video is generated from our previous blog post titled “Pinpointing leaked Lua tables with OpenResty XRay’s command-line tools”. You can watch the video on Youtube: youtu.be/CiftezUPuaI If you want to read the original post, click blog.openresty.com/en/xray-cli-lua-table-leaks/
-
OpenResty
@OpenResty
We’re delighted to share with you our latest Chinese blog post introducing EdgeLang: “EdgeLang: A Powerful and Efficient Gateway Logic Language”: blog.openresty.com/en/edgelang-intro/ EdgeLang is a domain-specific language provided by the OpenResty Edge product designed specifically for gateway logic. It has powerful features and efficient performance, allowing you to quickly implement various complex gateway policies.
-
OpenResty
@OpenResty
We have generated the second video with AI talking heads using the toolchain, and sent it to B station: youtu.be/YPUGJBbSwIk It comes with subtitles. This video is still automatically generated from a previous blog post titled “Using YSQL Language to Perform Real-Time Request Counting on Nginx Processes”. Compared with the first video, we have made some improvements to the toolchain.
-
OpenResty
@OpenResty
We recently tried a new form of AI-generated video narration, where we turned one of our blog posts “Troubleshooting the extra 200-ms request latency online” into a talking-head video and uploaded it to YouTube: youtu.be/Y53mElQgfpo.
It has subtitles in multiple languages. The whole video, including the subtitles, was created by a program from programs and SaaS services. The original article is here: blog.openresty.com/en/ngx-req-body-latency/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The root cause discussed in this post is on the client side. There are many possible causes of long-latency requests. In the future, we will share more of the other situations. We'll also cover other web servers and web applications. Each case is fascinating, and OpenResty XRay can automatically perform online analysis. The most difficult problems are often the most interesting ones.
OpenResty
@OpenResty
Have you ever encountered the problem that a small portion of online requests takes longer than expected? If yes, our new article, “Troubleshooting the extra 200-millisecond request latency online,” may help you: blog.openresty.com/en/ngx-req-body-latency/ It introduces the Smart Packet Capture feature of OpenResty XRay, which efficiently captures the network packets only on problematic TCP connections in real-time and automatically analyze the latency between network packets, thus helping you quickly locate and solve the root cause of long-latency requests.
-
OpenResty
@OpenResty
Have you ever encountered the problem that a small portion of online requests takes longer than expected? If yes, our new article, “Troubleshooting the extra 200-millisecond request latency online,” may help you: blog.openresty.com/en/ngx-req-body-latency/ It introduces the Smart Packet Capture feature of OpenResty XRay, which efficiently captures the network packets only on problematic TCP connections in real-time and automatically analyze the latency between network packets, thus helping you quickly locate and solve the root cause of long-latency requests.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We have also ported these GC object memory leak analyzers to any unmodified PHP, Python, Perl, and Go applications, not just OpenResty and Lua applications. We will share them with you through a series of video tutorials soon. Stay tuned!
OpenResty
@OpenResty
Do you want to know how to use OpenResty XRay’s powerful command-line tool to discover and solve Lua table memory leaks? Then don’t miss our latest tutorial article “Using OpenResty XRay’s Command Line Tool to Locate Leaking Lua Tables”: blog.openresty.com/en/xray-cli-lua-table-leaks/ This article will teach you how to use OpenResty XRay’s online non-invasive analysis feature to quickly find out which Lua tables cannot be garbage collected (GC) and occupy a lot of memory space.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Interestingly, Bilibili, one of the largest video sites in Asia, their own gateway server also uses our dynamic metric system to perform complex real-time online statistics. The nginx or OpenResty-based statistical analysis modules on the market that are based on shared memory are much less performant and also lack dynamism, basically, they are fixed metrics. And our dynamic metrics’ online calculation logic can be customized by writing our Metric SQL language.
OpenResty
@OpenResty
Check our new video tutorial "Standard Dynamic Metrics in OpenResty Edge". This tutorial demonstrates how standard dynamic metrics work in OpenResty Edge.
youtu.be/Uz2FxYQI6yk
Text version of this tutorial: blog.openresty.com/en/edge-standard-dynamic-metrics/
Let us know what you think! 😉 -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
This blog post demonstrates a new type of flame graph we invented a few years ago: data reference relationship flame graphs, which can be used to quickly locate memory leak problems, including languages that use GC or memory pools. It really helps our customers quickly locate memory leaks or memory consumption problems that have plagued them for a long time in production. It only takes a few minutes to locate the root cause of the problem (even if the leak process takes a week or longer).
OpenResty
@OpenResty
Do you want to know how to use OpenResty XRay’s powerful command-line tool to discover and solve Lua table memory leaks? Then don’t miss our latest tutorial article “Using OpenResty XRay’s Command Line Tool to Locate Leaking Lua Tables”: blog.openresty.com/en/xray-cli-lua-table-leaks/ This article will teach you how to use OpenResty XRay’s online non-invasive analysis feature to quickly find out which Lua tables cannot be garbage collected (GC) and occupy a lot of memory space.
-
OpenResty
@OpenResty
Do you want to know how to use OpenResty XRay’s powerful command-line tool to discover and solve Lua table memory leaks? Then don’t miss our latest tutorial article “Using OpenResty XRay’s Command Line Tool to Locate Leaking Lua Tables”: blog.openresty.com/en/xray-cli-lua-table-leaks/ This article will teach you how to use OpenResty XRay’s online non-invasive analysis feature to quickly find out which Lua tables cannot be garbage collected (GC) and occupy a lot of memory space.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OpenResty Edge’s dynamic metrics are a magical mechanism that we invented, which can automatically generate efficient code for metric aggregation and hot update it to all edge nodes. It allows you to efficiently perform streaming processing and real-time aggregation statistics in a fixed memory space. You don’t need any traditional big data processing and analysis software and databases, nor do you rely on log files, to quickly get big data analysis results.
OpenResty
@OpenResty
Check our new video tutorial "Standard Dynamic Metrics in OpenResty Edge". This tutorial demonstrates how standard dynamic metrics work in OpenResty Edge.
youtu.be/Uz2FxYQI6yk
Text version of this tutorial: blog.openresty.com/en/edge-standard-dynamic-metrics/
Let us know what you think! 😉 -
OpenResty
@OpenResty
Hi folks!
This exciting tutorial demonstrates how to configure the built-in captcha feature in OpenResty Edge for your applications.
youtu.be/JtF3V27hyNc
Give it a look and feel free to ask any questions or share your thoughts.
Text version: blog.openresty.com/en/edge-captcha/ -
OpenResty
@OpenResty
Hey everyone! We want to share this tutorial with you.
It showcases how to serve static resources on OpenResty Edge, making it easy to deliver those static files with efficiency and speed.
youtu.be/Tq-_Kq7eq2Y
Feel free to ask any questions or provide feedback. 😉 -
OpenResty
@OpenResty
Hey there! We've got atutorial for you to check out.
It walks you through the process of setting up a reverse proxy in OpenResty Edge for TLS connections with SNI extension.
youtu.be/hY0-XH_Svc0
Feel free to ask any questions or share your thoughts on it. Enjoy! -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
People asked about the implementation of our YSQL language. You can refer to the two articles I wrote before that introduce the basic principles of dynamic tracing and our improvements and innovations in this field: blog.openresty.com/en/dynamic-tracing-part-1/ and blog.openresty.com/en/ylang-intro-part1/ #orxray
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We also plan to support more types of target applications as “virtual tables” for YSQL queries, such as Envoy, ATS, and other server processes, various application frameworks in Python, Go, Java, etc., and database processes like MySQL, Pg, etc. Infinite possibilities. #orxray
OpenResty
@OpenResty
Want to have a powerful tool that directly query your running processes as a read-only database with SQL? And no logs or metrics at all? If the answers are yes, you will like our new blog post “Real-Time Request Counting against Nginx Processes via YSQL”: blog.openresty.com/en/ysql-count-ngx-reqs/
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
YSQL lets you analyze apps online without hassle. Just write a SQL query and see results and charts. No changes or restarts are needed for your applicatons, just like any other analyses in OpenResty XRay. YSQL also supports primitives for CPU, memory, latency, and many more.
OpenResty
@OpenResty
Want to have a powerful tool that directly query your running processes as a read-only database with SQL? And no logs or metrics at all? If the answers are yes, you will like our new blog post “Real-Time Request Counting against Nginx Processes via YSQL”: blog.openresty.com/en/ysql-count-ngx-reqs/
-
OpenResty
@OpenResty
Want to have a powerful tool that directly query your running processes as a read-only database with SQL? And no logs or metrics at all? If the answers are yes, you will like our new blog post “Real-Time Request Counting against Nginx Processes via YSQL”: blog.openresty.com/en/ysql-count-ngx-reqs/
-
OpenResty
@OpenResty
Take a look at this helpful tutorial that shows you how to easily check application error logs and global error logs in OpenResty Edge. Feel free to ask any questions or share your feedback!
youtu.be/yWcYDZWm5V0
Text version: blog.openresty.com/en/edge-error-log/ -
OpenResty
@OpenResty
Hello everyone!
Check this tutorial that demonstrates how to set up a reverse proxy in OpenResty Edge for TCP connections.
YouTube: youtu.be/MnoI48k_ZRo
Text version of this tutorial: blog.openresty.com/en/edge-tcp-application/
Feedback and questions are always welcome 🙂 -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
For the on-premise version of OpenResty XRay, we don't have access to any of our customer data (other than software license usage stats). Microsoft, one of our on-prem customers, also deployed OpenResty XRay in their environment, following this guide. #openrestyxray #azure
OpenResty
@OpenResty
This tutorial demonstrates how to deploy the on-premises version of OpenResty XRay on Microsoft's Azure cloud.
www.youtube.com/watch?v=-asOidukUw8
Text version of this tutorial: blog.openresty.com/en/xray-azure/ -
OpenResty
@OpenResty
This tutorial demonstrates how to deploy the on-premises version of OpenResty XRay on Microsoft's Azure cloud.
www.youtube.com/watch?v=-asOidukUw8
Text version of this tutorial: blog.openresty.com/en/xray-azure/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Since Qunar switched to our OpenResty Edge's WAF, online server CPU usage has significantly dropped. Ajaib, Indonesia's stock trading platform, also reports a significant performance boost compared to their previous Go-implemented WAF and API gateway. #openrestyedge
OpenResty
@OpenResty
Check out this tutorial that shows how to enable the Web Application Firewall, or WAF in OpenResty Edge.
youtu.be/5tFQHdCf1yo
The text version of this tutorial is in the following link:
blog.openresty.com/en/edge-enable-waf/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Someone asked why I use Fedora, a cutting-edge system. Some clients of ours insist on Fedora for production - I can't persuade them otherwise. And for hacking on the latest kernels and GNU toolchains, Fedora's handy. I run many different KVM VMs and containers myself anyway.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The Google gRPC C++ client is bloated, like many Google C++ OSS. Its load performance is lackluster - until we used OpenResty XRay to identify hotspots, optimize them, and boost the QPS manifold. More on this later. #openresty #grpc #performance
-
OpenResty
@OpenResty
Have you seen our video tutorial "Access Logs in OpenResty Edge"?
youtu.be/qbH92BISUuY
This tutorial demonstrates how to configure the access log format for the application in OpenResty Edge.
Text version of this tutorial: blog.openresty.com/en/edge-access-log/
Feedback is welcome! -
OpenResty
@OpenResty
Hi all!
Don't miss our new tutorial "Sticky Cookie in OpenResty Edge". This tutorial demonstrates how to enable sticky cookie in OpenResty Edge.
▶️ youtu.be/FQteUX-oPwE
Text version of this tutorial: blog.openresty.com/en/edge-sticky-cookie/ -
OpenResty
@OpenResty
Check our new video tutorial "Standard Dynamic Metrics in OpenResty Edge". This tutorial demonstrates how standard dynamic metrics work in OpenResty Edge.
youtu.be/Uz2FxYQI6yk
Text version of this tutorial: blog.openresty.com/en/edge-standard-dynamic-metrics/
Let us know what you think! 😉 -
OpenResty
@OpenResty
Don't miss our latest video tutorial "Proxy Kubernetes upstream in OpenResty Edge".
There we demonstrate how to use OpenResty Edge as a powerful ingress controller for Kubernetes Clusters.
youtu.be/pmlQiQZq91M
Text version of this tutorial: blog.openresty.com/en/edge-k8s-upstream/ -
OpenResty
@OpenResty
We're happy to announce a new version of OpenResty Edge (23.3.1) has been released!
Check all the information in the following links:
openresty.com/en/edge/release-notes/
doc.openresty.com/en/edge/edge-installation-upgrade/edge-… -
OpenResty
@OpenResty
Make sure to watch this tutorial that shows how to set up a gRPC reverse proxy and load balancer in OpenResty Edge.
youtu.be/udrFa0HVlww
Text version of this tutorial:
blog.openresty.com/en/edge-grpc-proxy/ -
OpenResty
@OpenResty
Check out this tutorial that shows how to enable the Web Application Firewall, or WAF in OpenResty Edge.
youtu.be/5tFQHdCf1yo
The text version of this tutorial is in the following link:
blog.openresty.com/en/edge-enable-waf/ -
OpenResty
@OpenResty
This tutorial demonstrates how to pass the real client IP addresses to backend or origin servers via special request headers.
youtu.be/X6fhRn8YDK8
Text version of this tutorial: blog.openresty.com/en/edge-upstream-real-ip/ -
OpenResty
@OpenResty
Check our new video about Edgelang in OpenResty Edge.
youtu.be/a1kMd_ppMpo
This tutorial demonstrates how to use our rule-based domain-specific language, the Edge language (Edgelang).
Text version of this tutorial: blog.openresty.com/en/edge-edgelang/ -
OpenResty
@OpenResty
We've kicked out a new release candidate, OpenResty 1.21.4.2 RC1.
See the full release note and complete change logs in the following link:
openresty.org/en/ann-1021004002rc1.html
We provide Win32 and Win64 pre-built binary packages for this release candidate.
Feedback is always welcome!😉 -
OpenResty reposted
-
OpenResty
@OpenResty
We used #openrestyxray to support a customer in the web security industry optimize its OpenResty/Nginx applications' poor request throughput.
Check the full post to see how OpenResty XRay helps When Lua IPC Pipes Block OpenResty or Nginx's Event Loops.
blog.openresty.com/en/ngx-off-cpu-lua-pipes/?src=twitter -
OpenResty
@OpenResty
If you want to learn about Domain-Specific Languages, their Implementation Strategies, and Technologies, don't miss this lecture by Yichun Zhang, creator of #openresty, during the Building a Programming Language course.
youtu.be/XxL6Frvv4TM
vimeo.com/799754876 -
OpenResty
@OpenResty
We finished the blog post series “The Wonderland of Dynamic Tracing”, which consists of 3 parts.
Dynamic tracing helps answer some tricky problems about software systems.
Find the complete series in the following link:
blog.openresty.com/en/dynamic-tracing-part-1/?src=twitter -
OpenResty
@OpenResty
Our new blog post is a case study to show how #openrestyxray helped one of our customers to Optimize Memory Fragmentation due to Huge Nginx Configuration.
Check the complete case here: blog.openresty.com/en/ngx-cycle-pool-frag/?src=twitter -
OpenResty
@OpenResty
For those who have questions about #openrestyxray, you can visit our new FAQ page.
If you have a question that is not there, send it to us at info@openresty.com and we will gladly support you to solve it.
Check the full file in the following link: openresty.com/en/xray/faq/?src=twitter -
OpenResty
@OpenResty
Now #openrestyxray automatically generates analysis & diagnostic reports for online OpenResty and Nginx applications! And higher-level apps based on OpenResty analyze issues automatically.
Check the full post "Automatic Analysis Reports in OpenResty XRay"
blog.openresty.com/en/xray-auto-reports/?src=twitter -
OpenResty
@OpenResty
#openrestyxray provides dynamic-tracing tools that can analyze online OpenResty and Nginx processes in real-time and quickly locate vulnerable and slow regular expressions.
In this tutorial, we demonstrate such advanced analyzers with real-world examples: blog.openresty.com/en/ngx-slowest-pcre-regexes/ -
OpenResty
@OpenResty
Understanding how #openresty, #nginx, and #luajit allocate and manage memory, helps debug and optimize memory usage or leaking issues. #openrestyxray, automatically analyzes and troubleshoots memory usage issues without any modifications or compromises. 👇
blog.openresty.com/en/how-or-alloc-mem/ -
OpenResty
@OpenResty
It's helpful to quickly find out what request URIs or hostnames are consuming the most CPU time in an online server or servers. Here we show how to use #openrestyxray to analyze unmodified OpenResty and Nginx web servers for such statistics in real-time.👇
blog.openresty.com/en/ngx-cpu-hottest-reqs/ -
OpenResty
@OpenResty
Check this tutorial that demonstrates a straightforward use case of dumping out the names of all loaded Lua modules in a running OpenResty and LuaJIT application. 🤩👇
blog.openresty.com/en/lj-dump-loaded-mods/ -
OpenResty
@OpenResty
Memory fragmentation is a common problem in computer systems.
In this article we use #openrestyxray to analyze and visualize the internals of unmodified #openresty or #nginx servers and applications. blog.openresty.com/en/nginx-shm-frag/ -
OpenResty
@OpenResty
#openresty and #nginx servers are often configured with shared memory zones which can hold data that is shared among all their worker processes. In this article, we will explore how these shared memory zones consume physical memory (or RAM).
blog.openresty.com/en/how-nginx-shm-consume-ram/ -
OpenResty
@OpenResty
For those interested in #openrestyxray, this gif will help you review what you can achieve with it!
"With OpenResty XRay we were able to troubleshoot mysterious software issues in production instantly." Zoom
"OpenResty XRay is easy to install, noninvasive and intuitive." Yundun -
OpenResty
@OpenResty
We want to share with you our post where we demonstrate how to setup the simplest reverse proxy and load balancer in #openresty Edge.
blog.openresty.com/en/edge-new-proxy/
Check it out and let us know what you think 🤩
#openrestyedge #reverseproxy #loadbalancer -
OpenResty
@OpenResty
If you're wondering how to upload your own SSL certificates to your HTTPS applications in #openresty Edge; just check this post where we show you how to do it 😉
blog.openresty.com/en/edge-upload-ssl-certs/
#ssl #https #openrestyedge -
OpenResty
@OpenResty
Check how #openrestyxray resolved a significant website crash of #bilibili, one of the busiest #streaming websites in Asia. 😱🤩
Read the full story, here: blog.openresty.com/en/bilibili-xray-incident/ -
OpenResty
@OpenResty
When the client requests hit the #cache on the Edge Node server, no requests need to be sent to the backend server. This reduces response latency and saves network bandwidth.
Check the full post on how to enable HTTP response caching in #openrestyedge: blog.openresty.com/en/edge-cache-rules/ -
OpenResty
@OpenResty
Learn how to purge #openresty Edge’s HTTP cache across the whole gateway network in real time! 🤓
Check the story here👇
blog.openresty.com/en/edge-cache-purge/
#openrestyedge -
OpenResty
@OpenResty
Sometimes the clients send requests too fast, like under a denial of service attack. In such cases, we should limit the request rate to protect both the gateway servers and origin servers. #openrestyedge helps you with this!🤩 Check the full post here👇
blog.openresty.com/en/edge-rate-limit/ -
OpenResty
@OpenResty
Don't miss our post about Lua-Land CPU Flame Graphs in #openresty XRay
blog.openresty.com/en/lua-cpu-flame-graph/
This article will introduce the idea of Lua-land CPU flame graphs and use #openrestyxray to produce real flame graphs for several small and standalone Lua examples. 😉📉 -
OpenResty
@OpenResty
🚨On Saturday, August 20, our founder & creator, @agentzh, will host the class "Why the Lua language and its implementation can super-charge OpenResty and Nginx" at the course Building a Programming Language by Roberto Ierusalimschy.🤩
Check the info here👉clas.pt/7vb4 -
OpenResty
@OpenResty
#dynamictracing allows software engineers to answer some tricky problems about software systems, such as high CPU or memory usage, high disk usage, long latency, or program crashes.
Check the full post here 👇
medium.com/@openresty/the-wonderland-of-dynamic-tracing-p… -
OpenResty
@OpenResty
Advanced #observability built for #opensource tech! We solved a major incident in production that caused Bilibili’s website crash, by using our OpenResty XRay’s dynamic tracing technology.
Check the full post here: 👇
link.medium.com/xqkwzIDRSrb -
OpenResty
@OpenResty
Is now live! 👀
Our new post talks about Tracing the Slowest PCRE Regular Expressions in OpenResty or Nginx Processes and how OpenResty XRay provides dynamic-tracing tools to analyze processes in real-time and quickly locate those vulnerable and slow RE!
👉blog.openresty.com/en/ngx-slowest-pcre-regexes/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We can now compile a Lua-compatible language, YLua, directly down to stap+/ebpf/gdb/odb dynamic tracing tools, for inspecting unmodified Lua and OpenResty applications.
OpenResty
@OpenResty
🤓 In the new post we have on our blog, we talk about a straightforward use case of dumping out the names of all loaded Lua modules in a running OpenResty and LuaJIT application.
Go check it out👇
blog.openresty.com/en/lj-dump-loaded-mods/ -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I used to talk to @brendangregg about using a SQL language to describe complex dynamic tracing tool logic. Now YSQL makes it a reality and it's indeed very flexible and powerful. It emits both systemtap and ebpf tools as demonstrated in the blog post below.
OpenResty
@OpenResty
Check our new blog post👉 Analyzing the Most CPU-Consuming Requests in #openresty or #nginx.
blog.openresty.com/en/ngx-cpu-hottest-reqs/
Here, we show how to use the dynamic-tracing tools in OpenResty #xray to analyze unmodified OpenResty and Nginx web servers for such statistics in real-time.🔍📊🕑 -
OpenResty
@OpenResty
🤓 In the new post we have on our blog, we talk about a straightforward use case of dumping out the names of all loaded Lua modules in a running OpenResty and LuaJIT application.
Go check it out👇
blog.openresty.com/en/lj-dump-loaded-mods/ -
OpenResty
@OpenResty
Check our new blog post👉 Analyzing the Most CPU-Consuming Requests in #openresty or #nginx.
blog.openresty.com/en/ngx-cpu-hottest-reqs/
Here, we show how to use the dynamic-tracing tools in OpenResty #xray to analyze unmodified OpenResty and Nginx web servers for such statistics in real-time.🔍📊🕑 -
OpenResty reposted
Classpert
@classpert
📣It's coming! On July 2nd, we're kicking off Building a Programming Language with Roberto Ierusalimschy, the creator of Lua.
Register Now clas.pt/7vb4
#programminglanguages #coding #softwareengineering #softwaredevelopment #hackernewsThis media couldn't be displayed. See it on the original post. -
OpenResty
@OpenResty
#openresty Edge is a powerful, efficient and scalable solution that protects valuable information by configuring HTTP basic Authentication in gateway applications.🔐 Our config changes don't require server reload, restart, or binary upgrade. Full post👇
blog.openresty.com/en/edge-basic-auth/ -
OpenResty
@OpenResty
With #openresty Edge, you can have real-time network-wide statistics. 📊
In addition to the traditional metrics run during regular activity, you can also set up your dynamic metrics with SQL language to have specific database visibility and monitoring. ✅
blog.openresty.com/en/edge-custom-dynamic-metrics/ -
OpenResty
@OpenResty
For everyone following and interested in OpenResty, we invite you to join our community Slack. You'll find updates, information, answers, and support on using OpenResty if you need it. Feel free to join and welcome in advance! 🤩🙌
openresty-community.slack.com/join/shared_invite/zt-19959… -
OpenResty
@OpenResty
🚀 We're glad to announce that we recently released version 0.30 of Test::Nginx, a test scaffold used during OpenResty development.
You can check the full release note, change logs, and more, here👇
openresty.org/en/ann-test-nginx-030.html -
OpenResty
@OpenResty
OpenResty 1.21.4.1 was just released. See the full release announcement: openresty.org/en/ann-1021004001.html Download: openresty.org/en/download.html
-
OpenResty
@OpenResty
Today’s new video tutorial: "Limit Request Rate by Custom Keys in OpenResty Edge": youtu.be/YKP091Y7sU0
Text version: blog.openresty.com/en/edge-rate-limit/
The video provides subtitles in multiple languages. OpenResty Edge: openresty.com/en/edge/ -
OpenResty
@OpenResty
We just kicked out OpenResty 1.21.4.1 RC1: openresty.org/cn/ann-1021004001rc1.html This is the first version that includes the Nginx 1.21.4 core. It also includes the new feature we OpenResty Inc. sponsored Mike Pall to do in LuaJIT: JIT compiling hash table iteration loops in Lua. Thanks!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We just published the final part, Part 4, of the series "Ylang: Universal Language for eBPF, Stap+, GDB, and More":
blog.openresty.com/en/ylang-intro-part4/
Now this series is complete. I also updated the previous 3 parts with more materials as well. -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I just published Part 3 of the series "Ylang: Universal Language for eBPF, Stap+, GDB, and More":
blog.openresty.com/en/ylang-intro-part3/
The last part, Part 4, is coming soon! -
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I just published Part 2 of the series "Ylang: Universal Language for eBPF, Stap+, GDB, and More":
blog.openresty.com/en/ylang-intro-part2/
This series is expected to have 4 or 5 parts in total. -
OpenResty
@OpenResty
The beginning of a new series of articles on our universal dynamic tracing language, ylang, offered by OpenResty XRay.
Yichun Zhang (agentzh)
@agentzh
I just published a new blog post "Ylang: Universal Language for eBPF, Stap+, GDB, and More (Part 1)": blog.openresty.com/en/ylang-intro-part1/ It's the first part of a long series. I must thank @brendangregg, Frank Ch. Eigler, and many others for years of inspiration.
-
OpenResty
@OpenResty
Today’s new video tutorial: “Gateway Config's Version Control & Release Management in OpenResty Edge": youtu.be/XkPlx9If6sI
Text version: blog.openresty.com/en/edge-config-release/
The video has subtitles in multiple languages.
Try out OpenResty Edge for free: openresty.com/en/edge/ -
OpenResty
@OpenResty
Today’s new video tutorial: youtu.be/i_n53z_zE3A
Text version: blog.openresty.com/en/showman-google/
As always, we provide subtitles in multiple languages. The video itself was also generated by OpenResty Showman.
Try out OpenResty Showman for free: openresty.com/en/showman/ -
OpenResty
@OpenResty
Today’s new video tutorial: “Set Different Host Headers for Upstream Requests in OpenResty Edge": youtu.be/NIlOrqi-I8U
Text version: blog.openresty.com/en/edge-upstream-host/
As always, we provide subtitles in multiple languages. It's also auto generated by OpenResty Showman. -
OpenResty
@OpenResty
We just upgraded openssl to 1.1.1l in our official binary packages (rpm/deb/apk/homebrew/etc). This includes important security fixes. All the users of these packages should upgrade ASAP.
-
OpenResty
@OpenResty
We just published the new pre-built binary package repos for Debian 11 Bullseye (for both x86_64 and aarch64):
openresty.org/en/linux-packages.html
Enjoy and feedback welcome! -
OpenResty
@OpenResty
Today’s new video tutorial: “Redirect to different URIs Based on Clients' Language Settings (OpenResty Edge)": youtu.be/Z7zeYz2zDbQ Text version: blog.openresty.com/en/edge-lang-redirect/ As always, we provide subtitles in multiple languages. It's also auto generated by OpenResty Showman. Enjoy
-
OpenResty
@OpenResty
Our OpenResty Edge product now has built-in integration support for Kubernetes (K8s) clusters, for either upstream clusters, gateway server clusters, or DNS authoritative server clusters, or all of them at the same time: openresty.com/en/edge/
-
OpenResty
@OpenResty
OpenResty 1.19.9.1 is now released: openresty.org/en/ann-1019009001.html This is the first OpenResty version based on the Nginx 1.19.9 core.
-
OpenResty
@OpenResty
Today’s new video tutorial: “Enforce the use of SSL in websites via OpenResty Edge": youtu.be/H1T-019IK2k Text version: blog.openresty.com/en/edge-http-to-https/ As always, we provide subtitles in multiple languages. It's also auto generated by OpenResty Showman.
-
OpenResty
@OpenResty
Today’s video tutorial: “Issue Free SSL Certificates via Let's Encrypt in OpenResty Edge ": youtu.be/OBrJlaAPv1k Text version: blog.openresty.com/en/edge-lets-encrypt/ As always, we provide subtitles in multiple languages. It's also auto generated by OpenResty Showman.
-
OpenResty
@OpenResty
OpenResty 1.19.9.1 RC1 is now available for testing: openresty.org/en/ann-1019009001rc1.html It is the first version based on Nginx core 1.19.9. Feedback welcome!
-
OpenResty
@OpenResty
OpenResty 1.19.3.2 was released last week with the latest Nginx security fixes: openresty.org/en/ann-1019009001rc1.html All the users should upgrade to this version.
-
OpenResty
@OpenResty
Today’s video tutorial: “Upload SSL Certificates for HTTPS Sites in OpenResty Edge": youtu.be/DV-uPZ-I5og Text version: blog.openresty.com/en/edge-upload-ssl-certs/ As always, we provide subtitles in multiple languages.
-
OpenResty
@OpenResty
We now provide official OpenResty Yum/Apt pre-built package repos for the latest Fedora 34 and Alpine 3.13: openresty.org/en/linux-packages.html Feedback welcome!
-
OpenResty
@OpenResty
Today’s video tutorial generated by our OpenResty Showman product: “Purge HTTP Cache in OpenResty Edge": vimeo.com/536099261 Text version: blog.openresty.com/en/edge-cache-purge/ As always, we provide subtitles in multiple languages.
-
OpenResty
@OpenResty
Today’s video tutorial generated by our OpenResty Showman product: "Enable HTTP Cache in OpenResty Edge": vimeo.com/532136845 Text version: blog.openresty.com/en/edge-cache-rules/ As always, we provide subtitles in multiple languages.
-
OpenResty
@OpenResty
Today’s short video tutorial: "Setup a Simplest Reverse Proxy and Load Blancer with OpenResty Edge": youtu.be/fgtiOYrdsE4 Text version: blog.openresty.com/en/edge-new-proxy/ As always, we provide subtitles in multiple languages.
-
OpenResty
@OpenResty
Today’s short video tutorial: "Install OpenResty on Ubuntu 20.04 via apt-get": youtu.be/Hu68mQDKXCo Text version: blog.openresty.com/en/ubuntu20-or-install/ As always, we provide subtitles in multiple languages.
-
OpenResty
@OpenResty
Now we provide official OpenResty Aarch64/ARM64 pkg repos for Ubuntu 18.04/20.04, Debian 9/10, CentOS/RHEL 7/8, Fedora 32/33. See openresty.org/en/linux-packages.html for more details.
-
OpenResty
@OpenResty
Today’s short video tutorial: "How We Generate Videos from Screenplay Files“: youtu.be/oew5RYUPMYY Text version: blog.openresty.com/en/ordemo/ As always, we provide subtitles in multiple languages.
-
OpenResty
@OpenResty
Today’s short video tutorial: "Streaming HTTP Response Output in OpenResty": youtu.be/4VQP7eKq4oA Text version: blog.openresty.com/en/stream-resp/ As always, we provide subtitles in multiple languages.
-
OpenResty
@OpenResty
We now provide official OpenResty package repositories for Fedora 33 and SUSE Linux Enterprise Server/Desktop 12.x/15.x: openresty.org/en/linux-packages.html
-
OpenResty
@OpenResty
We just upgraded our official OpenResty binary packages (rpm/deb/apk/etc packages) to use the latest OpenSSL 1.1.1i version to include important security fixes. All the users of these packages should upgrade their openresty-openssl111 package to the latest 1.1.1i version.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Check out our OpenResty Edge product: openresty.com/en/edge/ The secret of our Edge's performance is massive inter-rule optimizations: doc.openresty.com/en/edge-introduction/
OpenResty
@OpenResty
How fast is OpenResty Edge's WAF? The following chart shows req/sec in the y-axis. So the higher, the better.
-
OpenResty
@OpenResty
How fast is OpenResty Edge's WAF? The following chart shows req/sec in the y-axis. So the higher, the better.
-
OpenResty
@OpenResty
Today’s short video tutorial: "Precompile Lua Modules into LuaJIT Bytecode to Speedup OpenResty Startup": youtu.be/EP7c0BM2yNo Text version: blog.openresty.com/en/luajit-bytecode/ As always, we provide subtitles in 13 languages.
-
OpenResty
@OpenResty
We just published a new short video tutorial today: “Measure Execution Time of Lua Code Correctly in OpenResty”: youtu.be/VkRYW_qLoME Text version: blog.openresty.com/en/time-lua/ We provide subtitles in 13 languages.
-
OpenResty
@OpenResty
Today's new tutorial video, "Lookup OpenResty Documentation on Terminal with restydoc": youtu.be/_PjpdUEeKeM The text version: blog.openresty.com/en/restydoc/ Enjoy!
-
OpenResty
@OpenResty
We just published a new tutorial video, "Share Data Across Requests Served by OpenResty": youtu.be/AVR5Ft6FXTo The text version: stage.blog.openresty.com/en/data-share/ Enjoy!
-
OpenResty
@OpenResty
Washington Post last week had a server side misconfiguration which let out our default home page shipped with the OpenResty open source software. We got a lot of angry users of theirs writing to us complaining about phishing, hijacking, and etc., which is just nonexistent.
-
OpenResty
@OpenResty
The text version of this tutorial is published on our blog site here: blog.openresty.com/en/centos8-or-install/?src=tt
OpenResty
@OpenResty
Just published new video tutorial "Install OpenResty on CentOS 8 via dnf" youtu.be/B5lfAZunxfc Enjoy!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
This video also demonstrates some of the power of navlang and opslang. All the web browser and web page operations were automated by navlang, which is used by the OpenResty Demo tool chain. The terminal operations were performed by opslang.
OpenResty
@OpenResty
Just published new video tutorial "Install OpenResty on CentOS 8 via dnf" youtu.be/B5lfAZunxfc Enjoy!
-
OpenResty
@OpenResty
Just published new video tutorial "Install OpenResty on CentOS 8 via dnf" youtu.be/B5lfAZunxfc Enjoy!
-
OpenResty
@OpenResty
Just published new tutorial video "The resty Command-Line Utility Demo": youtu.be/L1c7aw4mSOo The text version is on our blog site: blog.openresty.com/en/resty-cmd/?src=tt
-
OpenResty
@OpenResty
New short video tutorial: "Write Your Own Lua Modules for OpenResty Applications": youtu.be/vfYxOMl5LVY Its text version: blog.openresty.com/en/or-lua-module/
-
OpenResty
@OpenResty
This video was automatically generated by our OpenResty Demo tool chain from this simple screenplay document: gist.github.com/agentzh/8565f5c42bba1af88c88cde1fba5ae37 Input this doc, and output a read-to-upload mp4 file :)
OpenResty
@OpenResty
We just uploaded a new short video in the OpenResty Tutorial series: "Hello World HTTP Interface in OpenResty". Enjoy! youtu.be/eSfYLvVQMxw
-
OpenResty
@OpenResty
We just uploaded a new short video in the OpenResty Tutorial series: "Hello World HTTP Interface in OpenResty". Enjoy! youtu.be/eSfYLvVQMxw
-
OpenResty
@OpenResty
OpenResty 1.19.3.1 formal version is just released: openresty.org/en/ann-1019003001.html
-
OpenResty
@OpenResty
We just published new blog post "Introduction to Lua-Land CPU Flame Graphs": blog.openresty.com/en/lua-cpu-flame-graph/ Learn how to profile OpenResty/Nginx Lua apps online like a pro. Thanks
@brendangregg
for inventing flame graphs in the first place :) -
OpenResty
@OpenResty
We just published a new blog post, "Memory Fragmentation in OpenResty and Nginx's Shared Memory Zones": blog.openresty.com/en/nginx-shm-frag/
-
OpenResty
@OpenResty
New blog post "How OpenResty and Nginx Shared Memory Zones Consume RAM": blog.openresty.com/en/how-nginx-shm-consume-ram/
-
OpenResty
@OpenResty
OpenResty 1.17.8.2 is just out: openresty.org/en/ann-1017008002.html Fixed a regression in ngx_http_perl_module in the last release.
-
OpenResty
@OpenResty
Test::Nginx 0.27 is released: openresty.org/en/ann-test-nginx-027.html
-
OpenResty
@OpenResty
OpenResty 1.17.8.1 formal version is just released: openresty.org/en/ann-1017008001.html
-
OpenResty
@OpenResty
We just added new official pre-built package repos for Fedora 32, Ubuntu 20.04 Focal and Alpine 3.7 ~ 3.11: openresty.org/en/linux-packages.html
-
OpenResty
@OpenResty
OpenResty 1.15.8.3 formal release is just out with important security fixes: openresty.org/en/ann-1015008003.html All OpenResty users are strongly encouraged to upgrade to this version.
-
OpenResty
@OpenResty
Just kicked out OpenResty 1.17.8.1 RC1 for the community to try out. See the full changes and download links here: openresty.org/en/ann-1017008001rc1.html Lots of bug fixes and new features. Feedback welcome!
-
OpenResty
@OpenResty
OpenResty now provides official yum/apt repos for Fedora 31, CentOS 8, and Ubuntu 19.10: openresty.org/en/linux-packages.html
-
OpenResty
@OpenResty
OpenResty new preview verison 1.17.8.1 RC0 is just out: openresty.org/download/openresty-1.17.8.1rc0.tar.gz Feedback welcome! The next version 1.17.8.1 RC1 is already around the corner and will be out in the next few days.
-
OpenResty
@OpenResty
We just published the new blog post "How OpenResty and Nginx Allocate and Manage Memory": blog.openresty.com/en/how-or-alloc-mem/ We use OpenResty XRay to demonstrate real data and
graphs for real OpenResty applications in the wild. Enjoy! -
OpenResty
@OpenResty
OpenResty now provides new official binary package repos for Fedora 30, Debian 10, OpenSUSE Leap 15.1, Amazon Linux 2, and RHEL 8: openresty.org/en/ann-1015008002.html More to come :) Our new build farm and test farm are powered by KVM + Opslang.
-
OpenResty
@OpenResty
OpenResty 1.15.8.2 security release was out today: openresty.org/en/ann-1015008002.html With nginx's new http2 security fixes.
-
OpenResty
@OpenResty
Our ylang and ylua higher-level compilers from OpenResty Trace thus have equally good support for floating-point numbers as well when targeting systemtap (their GDB backends always have such support since GDB is pure userland anyway).
Yichun Zhang (agentzh)
@agentzh
We recently added rather complete floating-point number support to @systemtap's kernel runtime. Reading, addition, subtraction, multiplication, division, IEEE modulo, C-style fmod, sqrt, negation, and abs for 32-bit and 64-bit floating point numbers are all supported. See
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The LuaJIT GC64 mode is *required* on Aarch64 (ARM64) anyway.
OpenResty
@OpenResty
The latest OpenResty 1.15.8.1 version enables LuaJIT's GC64 mode by default on x86_64, effectively removing the original 2G memory limit per worker process. We just wrote this detailed blog post to explain the GC64 mode and the old memory limit: blog.openresty.com/en/luajit-gc64-mode/
-
OpenResty
@OpenResty
The latest OpenResty 1.15.8.1 version enables LuaJIT's GC64 mode by default on x86_64, effectively removing the original 2G memory limit per worker process. We just wrote this detailed blog post to explain the GC64 mode and the old memory limit: blog.openresty.com/en/luajit-gc64-mode/
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Actually the ylua compiler actually emits the ylang code and then our ylang compiler compiles it further down to (bloody) stap scripts or gdb python scripts. No cooperation or plugins are needed on the side of target processes of LuaJIT or OpenResty. Ylang is a superset of C ;)
OpenResty
@OpenResty
We spent 2 weeks and developed a new compiler called ylua, which can compile a super set of the Lua language down to dynamic tracing and debugging tools in @systemtap and gdb python. Using Lua to trace Lua apps is a lot of fun. Ylua will be part of our OpenResty Trace product.
-
OpenResty
@OpenResty
We spent 2 weeks and developed a new compiler called ylua, which can compile a super set of the Lua language down to dynamic tracing and debugging tools in @systemtap and gdb python. Using Lua to trace Lua apps is a lot of fun. Ylua will be part of our OpenResty Trace product.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We're also porting our commercial products like OpenResty Edge and OpenResty Trace to aarch64 (arm64) servers for our customers :)
OpenResty
@OpenResty
OpenResty now provides official CentOS 7 Yum repository for pre-built OpenResty rpm packages targeting the Aarch64 (ARM64) architecture: openresty.org/en/linux-packages.html More aarch64 binary packages for more Linux distributions are also coming :)
-
OpenResty
@OpenResty
OpenResty now provides official CentOS 7 Yum repository for pre-built OpenResty rpm packages targeting the Aarch64 (ARM64) architecture: openresty.org/en/linux-packages.html More aarch64 binary packages for more Linux distributions are also coming :)
-
OpenResty
@OpenResty
Starting from the latest 1.15.8.1 release, OpenResty now provides official APT deb package repositories for the latest Ubuntu 19.04 Disco version: openresty.org/en/linux-packages.html We also upgraded all our openresty-openssl* rpm/deb packages to OpenSSL 1.1.0j. Enjoy!
-
OpenResty
@OpenResty
OpenResty 1.15.8.1 formal release is just out: openresty.org/en/ann-1015008001.html One full year's hard work of all our developers and contributors. Full change log: openresty.org/en/changelog-1015008.html
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The formal version 1.15.8.1 will be out in a week if nothing serious is reported in RC2.
OpenResty
@OpenResty
OpenResty 1.15.8.1 RC2 is just out: openresty.org/en/ann-1015008001rc2.html The lua-resty-core is now enabled by default in OpenResty. There are also several important bug fixes since RC1. Feedback welcome!
-
OpenResty
@OpenResty
OpenResty 1.15.8.1 RC2 is just out: openresty.org/en/ann-1015008001rc2.html The lua-resty-core is now enabled by default in OpenResty. There are also several important bug fixes since RC1. Feedback welcome!
-
OpenResty
@OpenResty
OpenResty 1.15.8.1 RC1 is just out for community testing: openresty.org/en/ann-1015008001rc1.html It is a very big version :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I successfully used Mozilla rr to quickly track down a subtle bug in @LuaJIT's GC which was reported by an OpenResty open source user. A patch for the fix has already been submitted to the LuaJIT mailing list, along with the full rr debugging session: www.freelists.org/post/luajit/Patch-to-fix-a-GC-write-bar…
-
OpenResty
@OpenResty
OpenResty now provides its official lua-resty-shell library which is based on the nonblocking ngx.pipe API: github.com/openresty/lua-resty-shell The ngx.pipe API is much more flexible upon which one can easily implement a tcl/expect-like engine and even more powerful.
-
OpenResty
@OpenResty
We prepared a formal document for our OpenResty EC2 Test Cluster toolchain: openresty.org/en/ec2-test-cluster.html Over the years it helped us find countless bugs before reaching our users' production, even including bugs in official NGINX and LuaJIT cores.
-
OpenResty
@OpenResty
We've prepared the OpenResty C Coding Style Guide document: openresty.org/en/c-coding-style-guide.html It also applies to general NGINX core and NGINX C module development since this is essentially the NGINX C Coding Style Guide :) Feedback welcome!
-
OpenResty
@OpenResty
OpenResty now provides official Yum/Apt repos with binary OpenResty related packages for Fedora 28/29 & Ubuntu 18.10 Cosmic: openresty.org/en/linux-packages.html Enjoy!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I finished my talk "New development in the OpenResty open source world" today at OpenResty Con 2018. The slides are here: openresty.org/slides/openresty-con-2018-oss/ (in Chinese tho). Will do another talk on "OpenResty technical support & OpenResty Trace platform" tomorrow.
-
OpenResty
@OpenResty
OpenResty 1.15.6.1 RC0 review version is just out: openresty.org/download/openresty-1.15.6.1rc0.tar.gz It includes the nginx 1.15.6 core with the latest http2 and mp4 security fixes. Alas. they do not provide hot fix patches for older nginx core versions like 1.13.6.
-
OpenResty reposted
OpenResty
@OpenResty
Published preview version 1.15.5.1 RC0 for OpenResty: openresty.org/download/openresty-1.15.5.1rc0.tar.gz This is the first version of OpenResty based on the latest Nginx 1.15.5 core. Also the first with official ARM64 support. The final version will get even more new features like ngx.pipe & sock:peek API.
-
OpenResty
@OpenResty
Published preview version 1.15.5.1 RC0 for OpenResty: openresty.org/download/openresty-1.15.5.1rc0.tar.gz This is the first version of OpenResty based on the latest Nginx 1.15.5 core. Also the first with official ARM64 support. The final version will get even more new features like ngx.pipe & sock:peek API.
-
OpenResty
@OpenResty
OpenResty's ngx_lua module, now drops support for the std Lua 5.1 interpreter so that we can remove a ton of compatibility code. It's recommended to use OpenResty's own LuaJIT branch versions for the best speed though the official LuaJIT 2.x releases will still be supported.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The video recording for my @nginxconf 2018 talk on dynamic tracing Nginx/OpenResty/Pg/Perl/Redis/Python/etc is now on YouTube: youtu.be/NR8U1qYHGqk The slides can be browsed here: openresty.org/slides/nginx-conf-2018/ (with some new updates). 35min was never enough for this talk. Alas.
-
OpenResty
@OpenResty
Thanks @kong for kindly sponsoring the OpenResty Inc team to do the ssl_ceritificate_by_lua* feature and ngx.ssl module in OpenResty's ngx_stream_lua_module. It's the best time to do dynamic TLS handshakes and TLS certificate lazy loading in your TCP servers atop OpenResty!
-
OpenResty
@OpenResty
More nice things about the arm64 port of OpenResty: we now have FFI implementations of the ndk.* Lua API in lua-resty-core as well. Also have the pure FFI versions of ngx.re.* API (and others) in the nginx stream subsystem. User apps using these will be JIT'd.
-
OpenResty
@OpenResty
We'll soon have the nonblocking ngx.pipe API in the open source OpenResty. It'll be trivial to do tcl/expect style scripting in OpenResty while preserving 100% nonblocking on Linux, Mac, and probably other operating systems. Better than tcl/expect: we do stderr/stdout separately.
-
OpenResty
@OpenResty
We're currently busy merging the arm64 patches to OpenResty's mainline repos. It also speeds up other architectures significantly for the simplest lua http/stream handlers (~10%). Thanks @Cloudflare again for the sponsorship.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Just published all the slides for my dynamic tracing talk's on @nginxconf 2018 on the web: openresty.org/slides/nginx-conf-2018/ It can now be viewed directly without downloading a tarball :) Also fixed some typos and made the images much smaller.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Just finished my dynamic tracing talk on #nginxconf 2018. The slides are in this tarball: drive.google.com/file/d/1R7VXRL6a1qv2M1bUbhvBKn5o7w07jFI2… Just decompress it and open the index.html file in the root directory with your favorite web browser (use arrow keys to switch pages).
-
OpenResty
@OpenResty
OpenResty Conference 2018 will be held in Hangzhou, China in mid-November. You're very welcome to submit a talk proposal and share your stories with OpenResty! See groups.google.com/d/msg/openresty-en/Py2FXeFYNfo/fzU43T5Q…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I'll give a presentation titled "Tracing and Troubleshooting NGINX, OpenResty, and Their Backends" on #nginxconf 2018 next month: www.nginx.com/nginxconf/2018/agenda/ … See you there! :)
-
OpenResty
@OpenResty
One nice side product of the arm64 work is that OpenResty is also more than 10% overall faster in ALL architectures for the simplest Lua handlers for nginx requests, including x86_64 ;)
-
OpenResty
@OpenResty
Cloudflare kindly sponsored OpenResty Inc to implement the arm64 support in the open source OpenResty. It will get included in the next open source release of OpenResty.
-
OpenResty
@OpenResty
OpenResty's resty command-line utility now supports the --gdb-opts=OPTS command-line option. It can be used to automate gdb with the underlying nginx process. Below is an example.
-
OpenResty
@OpenResty
Just documented our own changes in our OpenResty branch of LuaJIT: github.com/openresty/luajit2 We'll keep this list up to date.
-
OpenResty
@OpenResty
OpenResty 1.13.6.2 formal is now released. Announcement: groups.google.com/d/msg/openresty-en/s3p1kdcyKHk/o3rewdwY… Download: openresty.org/en/download.html Changes: openresty.org/en/changelog-1013006.html This is the first version which supports OpenSSL 1.1.0 and UDP server scripting. Also offer binary packages for Win64 & Ubuntu 18.04
-
OpenResty
@OpenResty
Starting from OpenResty 1.13.6.2 RC1, we provide official 64-bit Windows (Win64) native binary packages for OpenResty :) Try this: openresty.org/download/openresty-1.13.6.2rc1-win64.zip
-
OpenResty
@OpenResty
The open source OpenResty now supports UDP downstream cosocket Lua API in ngx_stream_lua_module. Time to code up nifty and fast UDP web apps with Lua atop OpenResty! :)
-
OpenResty
@OpenResty
OpenResty 1.13.6.2 RC1 is now available for testing. Release note: groups.google.com/d/msg/openresty-en/bTfqTjshcX4/uP8qF0Ep… Feedback welcome!
-
OpenResty
@OpenResty
Special thanks go to the Smartian company for donating more than 150K USD to the OpenResty open source project: openresty.org/en/donors.html
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Well, just comment out its stats_correct() C function calls, then you are good :) Despite this jaw-dropping flaw, wrk is fine.
-
OpenResty
@OpenResty
OpenResty Inc recently sponsored Shuxin Yang to add a new table.clone builtin API to OpenResty's LuaJIT branch. The docs is here: github.com/openresty/luajit2/commit/617f1183 It is much faster than a hand-crafted Lua loop and can also be 100% JIT compiled. It makes our fandang's Lua output 7% faster.
-
OpenResty
@OpenResty
Our team's Datong has added the local=on and local=/path/to/etc/resolv.conf parameters to OpenResty's nginx core's standard "resolver" config directive. This way nginx's own nonblocking DNS resolver can follow OS's name server settings: github.com/openresty/openresty/
-
OpenResty
@OpenResty
We recently noted that @nginx's --with-file-aio option on Linux would cause horrible startup time (an extra 20ms) on a Linux system. Without it, the startup time is only ~4ms. So we've removed this option from our official binary builds of OpenResty. Linux's AIO is lame anyway.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
One of our commercial customers complain about the long tail in request latency reported by wrk. So we dig it up and found the latency numbers reported by wrk is more than 10x larger than reality. And its author admits wrk adds noises to the data to emulate a real Internet. Gosh!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
With the new resty tool, our Perl 6 dialect compiler, fanlang's test suite has sped up by total 44%. It's because fanlang's test scaffold invokes resty to run the Lua code generated by the compiler. The *NIX culture loves tool-based programming and tools' startup time is the key.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I recently optimized @OpenResty's resty command-line utility's startup time from ~40ms to ~10ms. Already faster than python3 (14ms), v8's d8 (13ms), ruby (~40ms), or nodejs (~50ms). The resty tool is still written in Perl 5 and the nginx binary takes ~4ms.
-
OpenResty
@OpenResty
Our team's spacewander just added the ngx.resp Lua module to lua-resty-core, which supports the add_header() Lua API function that never overrides existing resp headers: github.com/openresty/lua-resty-core/blob/master/lib/ngx/r…
-
OpenResty
@OpenResty
ngx_stream_lua_module now provides the lua_add_variable config directive in nginx.conf's stream {} to define your own nginx variables inside a TCP/UDP server and read/write them via the ngx.var API in Lua: github.com/openresty/stream-lua-nginx-module Thanks Datong for adding this feature.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Wrote a small mini-tt2.pl tool that checks if our C code TT2 templates in ngx_meta_lua module follows the nginx coding style and also makes sure the generated C source for ngx_http_lua and ngx_stream_lua modules also follows the nginx coding style: github.com/openresty/meta-lua-nginx-module/blob/master/ut…
-
OpenResty
@OpenResty
Support for OpenSSL 1.1.0 has landed in OpenResty's most components' GitHub repos master. The next OpenResty release should see full compatibility with OpenSSL 1.1.0. Thanks Alessandro Ghedini, space wander, and dejiang for the contribution!
-
OpenResty
@OpenResty
OpenResty now provides official Yum repo for Fedora 27: openresty.org/en/linux-packages.html And we retired the old one for the EOF Fedora 25.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OpenResty's ngx.get_phase() API function can now be fully JIT compiled with LuaJIT when using the latest lua-resty-core. Thanks Robert for the patch!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Test::Nginx::Socket now supports the --- access_log section for checking nginx access log contents, like the existing --- error_log section: github.com/openresty/test-nginx/commit/f29f2f6f2c Thanks spacewander for the patch!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I just documented the ngx-build tool which we've been using for years for everyday nginx C module development: github.com/openresty/openresty-devel-utils It's quite powerful.
-
OpenResty
@OpenResty
OpenResty 1.13.6.1 formal release was just out: groups.google.com/d/msg/openresty-en/CoWdvOVmZhQ/VEFJItme… It's based on the latest NGINX mainline release 1.13.6 and includes ngx_stream_lua_module for generic TCP server scripting.
-
OpenResty reposted
Cindy Sridharan
@copyconstruct
*Fantastic* talk from @GOTOcph @GOTOcon Shopify's architecture in scaling to 80K RPS. Shoutout to @OpenResty deemed "the most impactful addition to Shopify's stack in recent memory". www.youtube.com/watch?v=N8NWDHgWA28
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I'm glad that our OpenResty community is posting flame graphs when talking about optimizations: github.com/openresty/lua-resty-mysql/pull/68 Blind optimizations are often dangerous and ineffective.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The formal version 1.13.6.1 will be out if there's nothing alarming in the community feedback for this RC version in the coming week.
OpenResty
@OpenResty
OpenResty 1.13.6.1 RC1 is now available for testing: groups.google.com/d/msg/openresty-en/pVPxLB5jfl4/QOnUnXnt… It's based on nginx 1.13.6 and includes ngx_stream_lua_module.
-
OpenResty
@OpenResty
It's already been 10 years since OpenResty's birth on Q4 2007. Happy 10 years anniversary!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Video recording of my talk "OpenResty Commercialization & DSL Era" on OpenResty Con 2017. It's in Chinese though: youtu.be/ZhP5jb1RZt8
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
All the talks' video recordings on OpenResty Con 2017 can be found here: www.itdks.com/dakashuo/playback/1439 We have one English talk :)
OpenResty
@OpenResty
OpenResty Con 2017 was a great success! Thanks all our organizers, sponsors & speakers! See all the talks' slides: con.openresty.org/cn/2017/
-
OpenResty
@OpenResty
OpenResty Con 2017 was a great success! Thanks all our organizers, sponsors & speakers! See all the talks' slides: con.openresty.org/cn/2017/
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I gave the talk "OpenResty Commercialization & DSL era" at this year's OpenResty Con conference. The slides are here:con.openresty.org/cn/2017/books/OpenResty Commercializati…
-
OpenResty
@OpenResty
OpenResty 1.13.6.1 RC1 is now available for testing: groups.google.com/d/msg/openresty-en/pVPxLB5jfl4/QOnUnXnt… It's based on nginx 1.13.6 and includes ngx_stream_lua_module.
-
OpenResty
@OpenResty
Just upgraded our openresty-pcre* packages to PCRE 8.41 in our official Linux pre-built package repos: openresty.org/en/linux-packages.html
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The code is here: github.com/openresty/stream-lua-nginx-module
OpenResty
@OpenResty
A LuaJIT internal bug in its GC64 mode cought by OpenResty's test suite: www.freelists.org/post/luajit/Consistent-SEGV-on-x64-with… Mike Pall fixed it in 2 hours since my mail
-
OpenResty
@OpenResty
ngx_stream_lua_module supports balancer_by_lua* and preread_by_lua* in master. Thanks @mashape for sponsoring OpenResty Inc. to do these.
-
OpenResty reposted
SoftwareMill
@softwaremill
Welcome to the world of @OpenResty. @marcinstozek will present this #webplatform at @JDD_Krakow jdd.org.pl/lecture.html #nginx #lua
-
OpenResty
@OpenResty
A LuaJIT internal bug in its GC64 mode cought by OpenResty's test suite: www.freelists.org/post/luajit/Consistent-SEGV-on-x64-with… Mike Pall fixed it in 2 hours since my mail
-
OpenResty
@OpenResty
OpenResty Inc. now offers the commercial software product, OpenResty Edge 2, for distributed traffic management: openresty.com
-
OpenResty
@OpenResty
Added 4 more new regions to our mini CDN network: London, Frankfurt, Singapore, and New York. *openresty.org should be faster now.
-
OpenResty
@OpenResty
We're busy preparing for the OpenResty Con 2017 event on Oct 21 in Beijing: con.openresty.org/cn/2017/ Calling for papers: talk@openresty.org
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The docs are here: github.com/openresty/lua-nginx-module & github.com/openresty/lua-nginx-module Thanks Thibault Charbonnier for the patch! resty.limit.count uses this.
OpenResty
@OpenResty
ngx_lua module's shm dictionaries (shdicts) now has got the expire() method for setting exptime on existing keys and ttl() for fetching TTL.
-
OpenResty
@OpenResty
ngx_lua module's shm dictionaries (shdicts) now has got the expire() method for setting exptime on existing keys and ttl() for fetching TTL.
-
OpenResty
@OpenResty
lua-resty-limit-traffic library got a new resty.limit.count module for doing github API style req count limiting: github.com/openresty/lua-resty-limit-traffic/blob/master/…
-
OpenResty
@OpenResty
OpenResty's "resty" command line utility now has new CLI options like `-j v`, `-j dump`, `-j off`, and `-l lua_module`, stolen from luajit.
-
OpenResty
@OpenResty
You are welcome to try that patch out on your side. Feedback welcome!
Yichun Zhang (agentzh)
@agentzh
Adding HTTP/2 support to our Test::Nginx::Socket test scaffold via curl: github.com/openresty/test-nginx/pull/67
-
OpenResty
@OpenResty
How to run OpenResty's grand test suite with various test modes on your side: github.com/openresty/openresty.org/issues/70
-
OpenResty
@OpenResty
The new deb pkg openresty_1.11.2.5-2 with the conf files fix has just been published to our Ubuntu APT repos as well.
-
OpenResty
@OpenResty
Our official APT repos for Debian 8+ now support multiple versions of the same deb packages. We upgraded reprepro to the latest version.
-
OpenResty
@OpenResty
Just published deb pkg openresty_1.11.2.5-2 to our official APT repos for Debian. Fixed the issue of overriding/removing nginx config files.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The new ngx_stream_lua module was actually generated from ngx_meta_lua_module: github.com/openresty/meta-lua-nginx-module/ which emits ngx_http_lua module too.
OpenResty
@OpenResty
OpenResty's ngx_stream_lua_module now supports nginx 1.13.3+: github.com/openresty/stream-lua-nginx-module/ Thanks @mashape for sponsoring OpenResty Inc to do it.
-
OpenResty
@OpenResty
OpenResty's ngx_stream_lua_module now supports nginx 1.13.3+: github.com/openresty/stream-lua-nginx-module/ Thanks @mashape for sponsoring OpenResty Inc to do it.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
NGINX Inc wrote down the 2nd part of my DSL talk at nginx.conf 2016 as a blog post, with my original slides: www.nginx.com/blog/building-business-systems-with-domain-…
-
OpenResty
@OpenResty
Ingy döt Net has been using our own Perl 6 dialect, fanlang, to write our own DSL compilers for OpenResty Inc's commercial gateway products.
Yichun Zhang (agentzh)
@agentzh
Ingy explains his work at OpenResty Inc:"I write new programming languages for a company that only uses their own new programming language."
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Thanks all our developers and contributors! Next release will be based on NGINX 1.13.4+ if nothing special comes up :) Enjoy!
OpenResty
@OpenResty
OpenResty 1.11.2.5 formal is just out. ANN: groups.google.com/d/msg/openresty-en/x64vkgEKKZU/nifrF5ti… Download: openresty.org/en/download.html Changes: openresty.org/en/changelog-1011002.html
-
OpenResty
@OpenResty
OpenResty 1.11.2.5 formal is just out. ANN: groups.google.com/d/msg/openresty-en/x64vkgEKKZU/nifrF5ti… Download: openresty.org/en/download.html Changes: openresty.org/en/changelog-1011002.html
-
OpenResty reposted
NGINX Open Source Community
@nginxorg
Domain-Specific Languages for #nginx & @OpenResty, Part 1 via @agentzh from @OpenResty www.nginx.com/blog/building-business-systems-with-domain-…
-
OpenResty
@OpenResty
OpenResty 1.11.2.5 RC1 is just out for testing: groups.google.com/d/msg/openresty-en/Mw5B11AQwqM/qY5MltXm… Lots of bug fixes and new features. Feedback welcome!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OpenResty now maintains its own fork of the ngx_postgres module here: github.com/openresty/ngx_postgres
-
OpenResty
@OpenResty
We now provide the openresty-asan RPM pkgs in OpenResty's official yum repos: openresty.org/en/rpm-packages.html openresty-asan Deb pkg will also come.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Wrote a general guideline on debugging openresty/nginx memory issues when gdb bt is not helpful: openresty.org/en/debugging.html
-
OpenResty
@OpenResty
OpenResty 1.11.2.4 is just out to include the latest security fix from upstream nginx.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OpenResty now maintains its own branch of LuaJIT test suite based on Mike Pall and others' tests: github.com/openresty/luajit2-test-suite
-
OpenResty
@OpenResty
We've just migrated OpenResty's official APT repo for Ubuntu from Launchpad PPA to our openresty.org site: openresty.org/en/linux-packages.html
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We now also provide OpenResty's official APT repository for Debian 9.x stretch amd64. Same installation steps as jessie. Feedback welcome!
Yichun Zhang (agentzh)
@agentzh
OpenResty now provides official binary deb packages and APT sources for Debian 7.x wheezy and 8.x jessie: openresty.org/en/linux-packages.html Enjoy!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OpenResty now provides official binary deb packages and APT sources for Debian 7.x wheezy and 8.x jessie: openresty.org/en/linux-packages.html Enjoy!
-
OpenResty
@OpenResty
OpenResty now offers official APT repositories /w binary .deb packages targeting Ubuntu 14.04 trusty ~ 17.10 artful: openresty.org/en/linux-packages.html
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The git master of lua-resty-mysql now supports the "charset" option in connect(): github.com/openresty/lua-resty-mysql Thanks Wilhelm Liao for the patch!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Dejiang contributed a lj-releng static code analyzer for OpenResty/LuaJIT's Lua applications: github.com/openresty/openresty-devel-utils/blob/master/lj… It uses OR's LuaJIT.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We are dropping support for i386 in our new OpenResty repositories though.
OpenResty
@OpenResty
We moved away from Fedora COPR to our own VM build farm atop VMware. Please update your local yum repos per this doc:openresty.org/en/linux-packages.html
-
OpenResty
@OpenResty
We moved away from Fedora COPR to our own VM build farm atop VMware. Please update your local yum repos per this doc:openresty.org/en/linux-packages.html
-
OpenResty
@OpenResty
Now OpenResty provides official yum repository for the latest Amazon Linux x86_64: openresty.org/en/linux-packages.html
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Documented our `lbc` gdb command for dumping LuaJIT bytecode in a specified PC range: github.com/openresty/openresty-gdb-utils It does NOT invoke target code
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Basically, the yielded Lua coroutine will not get resumed until the next event cycle in the nginx event loop /w *0* timeout in epoll() etc.
OpenResty
@OpenResty
Merged Datong's patch that makes ngx.sleep(0) voluntarily yield control to the nginx event loop & other concurrent reqs, with *0* I/O delay.
-
OpenResty
@OpenResty
Merged Datong's patch that makes ngx.sleep(0) voluntarily yield control to the nginx event loop & other concurrent reqs, with *0* I/O delay.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I've just added a Lua API to initiate graceful shutdown of the current nginx process (no UNIX signal is involved): github.com/openresty/lua-resty-core/blob/master/lib/ngx/p…
-
OpenResty
@OpenResty
Just merged Dejiang Zhu's patch for ngx.timer.every() Lua API for easily setting up recurring timers in OpenResty: github.com/openresty/lua-nginx-module/
-
OpenResty
@OpenResty
Th 32KB PCRE JIT stack size limit was hit when running the OR/Lua code generated by our fanlang compiler compiling our edgelang compiler.
OpenResty
@OpenResty
Andreas Lübbe contributed the Lua API for tuning the PCRE JIT stack size in OpenResty: github.com/openresty/lua-resty-core/blob/master/lib/ngx/r… The default 32KB is too small
-
OpenResty
@OpenResty
Andreas Lübbe contributed the Lua API for tuning the PCRE JIT stack size in OpenResty: github.com/openresty/lua-resty-core/blob/master/lib/ngx/r… The default 32KB is too small
-
OpenResty
@OpenResty
It is now possible for OpenResty to capture & analyze nginx error logs by Lua right inside the same process w/o IO: github.com/openresty/lua-resty-core/blob/master/lib/ngx/e…
-
OpenResty
@OpenResty
OpenResty now supports enabling a privileged agent process under nginx master: github.com/openresty/lua-resty-core/blob/master/lib/ngx/p… It's handy to control nginx itself.
-
OpenResty reposted
DarkGod
@darkgodone
Spent a day porting my server code from custom socket+libev code to @OpenResty ! Pleasure to work with the lua stream interface!
-
OpenResty
@OpenResty
OpenResty® 1.11.2.3 formal is just out: groups.google.com/d/msg/openresty-en/tvPorn8kZXE/n6wrxGp1… Every user should upgrade to this version. Thanks to all our contributors!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Oh my: status.fedoraproject.org All red. Fortunately we have an OR yum repository mirror on openresty.org.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Test::Nginx 0.26 is just released to CPAN: groups.google.com/d/msg/openresty-en/SO2s9VD8hDA/GzirJigY… Multi-core parallel testing support, new Dgram server testing support...
-
OpenResty
@OpenResty
OpenResty 1.11.2.3 RC1 is just out for testing: groups.google.com/d/msg/openresty-en/rrCLmpwismY/p6T16szN… Accumulated a lot of new improvements as always ;) Feedback welcome!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Our Perl 6 dialect impl, fanlang, is now mature enough to be used to rewrite our edgelang compiler, which itself can be run on OpenResty :D
-
OpenResty
@OpenResty
For Mac users, it is now recommended to install OpenResty via homebrew: brew install homebrew/nginx/openresty We're actively maintaining it.
-
OpenResty
@OpenResty
OpenResty's official yum repo updates: just upgraded openssl, zlib, and pcre to their latest official releases.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We need deb packaging expert to help us finish the official OpenResty deb packages: github.com/happyaron/openresty-packaging Anyone wants to help us? We pay!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Busy setting up the OpenResty Inc. company in the US. We'd like to sell highly efficient and highly flexible enterprise solutions atop OR :D
-
OpenResty reposted
NGINX
@nginx
.@ShopifyDevs share their secret weapon for handling massive traffic - an #nginx & @OpenResty #lua edge tier shopify.engineering/232808527-from-super-bowl-to-kylie-co…
-
OpenResty reposted
powerdns
@powerdns
PowerDNS is proud to support @OpenResty development! openresty.org/en/donors.html #lua
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Implemented a common subset of Perl 6 OO features in my fanlang compiler targeting OpenResty/LuaJIT. An example from fanlang's test suite.
-
OpenResty
@OpenResty
You are welcome to participate the discussion on migrating over to lua-resty-core in ngx_lua: github.com/openresty/lua-nginx-module/issues/949
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I'd like to permanently switch ngx_lua's CFunction-based Lua API impl over to the FFI-based ones in lua-resty-core. Any objections?
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The video recording for my DSLs + OpenResty talk on nginx.conf 2016 is out: www.youtube.com/watch?v=w2RAZLptE-U Slides are here: openresty.org/slides/Building-Business-Systems-with-DSLs-…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OpenResty's official RPM packages now enable nginx's standard ngx_http_geoip module by default. See openresty.org/en/linux-packages.html for details.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The resty command line utility now supports the --shdict "NAME SIZE" option. Thanks Thibault Charbonnier for the patch!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Slides used at my talk at OpenResty Con 2016 yesterday are here: openresty.org/slides/New-development-of-OpenResty-in-2016… A good summary of OpenResty for the past year :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
My talk at OpenResty Con 2016 (in Chinese):youtu.be/_4fezJVmlhM?a(来自@YouTube)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Video recording for my recent talk, "Writing optimal Lua code for LuaJIT and OpenResty", at Lua Workshop 2016 SF: www.youtube.com/watch?v=FfhEdF40nhQ
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I'll still live in the bay area of US but as a freelancer in the next few months. I will continue leading the @OpenResty project, of course.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Today was my last day at @Cloudflare. Thanks @Cloudflare for the great support of my work on the OpenResty project in the last 4.5 years!
-
OpenResty
@OpenResty
OpenResty 1.11.2.2 formal is just out: www.freelists.org/post/luajit/OpenResty-11122-released This is the first formal version containing opm & lua-resty-limit-traffic :)
-
OpenResty
@OpenResty
This year's OpenResty Con 2016 conference will be held in Shenzhen, China on 10 December: con.openresty.org/cn/2016/ Welcome to join us!
-
OpenResty
@OpenResty
OpenResty 1.11.2.2 RC2 is just out for community testing: groups.google.com/d/msg/openresty-en/pWAJyx4YSXc/0Voaud5j… We include some new fixes since RC1. Feedback welcome!
-
OpenResty
@OpenResty
Our OpenResty Software Foundation based in Hongkong has just got tax exemptions: www.ird.gov.hk/cgi-bin/irdnew/ach/search.cgi?lang=e&id=91… You are welcome to make a donation!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The next version of OpenResty will turn on LuaJIT's Lua 5.2 compat mode by default: luajit.org/extensions.html Those new 5.2 features are nice!
-
OpenResty
@OpenResty
OpenResty 1.11.2.2 RC1 now available for testing: groups.google.com/d/msg/openresty-en/cNeiiH_Qhl4/Ses9XCd6… Includes opm, lua-resty-limit-traffic, and etc. Feedback welcome!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
opm now supports the "opm clean dist" command, which can clean up the temp files and directories generated by "opm build". zhuanyenan++
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
ngx_http_lua module's ssl_session_fetch_by_lua* & ssl_session_store_by_lua* directives are now only allowed in http {} context of nginx.conf
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I've just merged a patch from Thibault Charbonnier, which implements the ngx.re.split() API function for OpenResty: github.com/openresty/lua-resty-core/blob/master/lib/ngx/r…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Our ngx_http_lua module now has a C API for 3rd-party nginx modules to add custom shm stores like shdict: github.com/openresty/lua-nginx-module/commit/73bd64dbc helloyi++
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We also just renamed our nginx-systemtap-toolkit to openresty-systemtap-toolkit since it's much more than nginx: github.com/openresty/openresty-systemtap-toolkit :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We just renamed our nginx-gdb-utils project to openresty-gdb-utils to better reflect its strength and use cases: github.com/openresty/openresty-gdb-utils
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Our ngx_replace_filter module can now be compiled as a dynamic module with nginx 1.9.11+: github.com/openresty/replace-filter-nginx-module shmilee++
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The official deb packages for @OpenResty are still under construction: github.com/happyaron/openresty-packaging Anyway, PR welcome! :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
TCP cosockets in @OpenResty now have new method `settimeouts(conn_timeout, send_timeout, read_timeout)`: github.com/openresty/lua-nginx-module Dejiang++
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We @Cloudflare just opensource TLS session ticket key auto rotation in NGINX or OpenResty: github.com/openresty/lua-ssl-nginx-module It supports global sync.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We're currently working on the ngx.re.split API in ngx_http_lua_module. Please help us review: github.com/openresty/lua-resty-core/pull/66 Thanks thibaultcha!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Just open-sourced lib lua-resty-memcached-shdict that I created for @Cloudflare's Lua CDN: github.com/openresty/lua-resty-memcached-shdict Thanks @Cloudflare!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Just open-sourced the lua-resty-shdict-simple library that I created for @Cloudflare's Lua CDN: github.com/openresty/lua-resty-shdict-simple Thanks @Cloudflare.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
ngx_http_lua_module now has lua_malloc_trim directive that can periodically trim memory footprint of nginx workers: github.com/openresty/lua-nginx-module
-
OpenResty reposted
Siftery-Insights
@sifteryinsights
Woohoo! @OpenResty You've added 376 customers in the past 2 weeks!🚀🚀 siftery.com/openresty
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Just open-sourced a naive multipart data parser lib, lua-resty-multipart-parser, for @OpenResty: github.com/agentzh/lua-resty-multipart-parser Thanks @Cloudflare!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Published the lua-tablepool lib originally created for @Cloudflare's Lua CDN. Thanks @Cloudflare for the support! github.com/agentzh/lua-tablepool
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
No, no, no. Lua is not a DSL. It's not convenient enough. @OpenResty will offer a proper DSL (targeting Lua) for WAF and API/web gateways.
Ivan Ristic
@ivanristic
I actually find it surprising that, with Lua being part of ModSecurity for many years, people don’t actually use it. // cc @agentzh
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Here's the slides used at my "OpenResty/LuaJIT app optimizations" talk at Lua Workshop 2016: openresty.org/slides/Optimizing-Lua-Applications-for-LuaJ… Enjoy!
-
OpenResty
@OpenResty
Our very first community pkg contributor to opm.openresty.org is Quan Zhao. We send him ¥500 (or $75) to express our appreciation :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We're actively working with a deb packaging expert on @OpenResty's official deb packages for Ubuntu and Debian. Stay tuned!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I will give a talk on writing optimal Lua code for @OpenResty and @LuaJIT at Lua Workshop 2016 on Oct 14 in SF: www.luasf2016.org
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The opm.openresty.org pkg server itself is 100% in OpenResty/Lua, backed by PostgreSQL. It's also open source: github.com/openresty/opm/tree/master/web
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
OPM package server already has 42 uploads across 40 distinct packages from 12 contributors. Come on, OPM authors! opm.openresty.org :D
-
OpenResty reposted
Jérémy Lal
@kapouer
github.com/kapouer/upcache allows easy cache configuration/invalidation from expressjs using surrogate keys and jwt scopes, thx to @OpenResty
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Our opm.openresty.org web server is alive! Welcome to upload your first OR package via the opm toolchain to our central package server
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I've been busy hacking on "opm", the official OpenResty package management system: github.com/openresty/opm Feedback welcome! :)
-
OpenResty
@OpenResty
We just upgraded the openresty-openssl* packages in our official yum repos to the latest OpenSSL 1.0.2i release. Please do "yum update" :)
-
OpenResty reposted
Lua Conf
@LuaConf
Intelligent load balancers, by @xldenis www.youtube.com/watch?v=WSUEjWgLk30 #luaconf #lualang @nginxorg @OpenResty [English]
-
OpenResty reposted
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I just gave a talk on DSL+OpenResty at @nginx conf 2016. Nice big conference room! I was very excited. Slides here: openresty.org/slides/Building-Business-Systems-with-DSLs-…
-
OpenResty
@OpenResty
OpenResty's official RPM packages now use the dual number mode in the LuaJIT builds since it's usually more efficient for web applications.
-
OpenResty
@OpenResty
OpenResty 1.11.2.1 formal is just out: groups.google.com/d/msg/openresty-en/WS6KHG6flbw/cEIUV-wT… It's a big milestone release. Special thanks go to all our contributors!
-
OpenResty
@OpenResty
We just refactored the official OpenResty RPM packages and added the openresty-zlib and openresty-pcre packages. Thanks CJ Ess for the patch
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I'll give the talk "Building Business Systems with Domain-Specific Languages for @OpenResty" on #nginxconf 2016 in Sep. See you there! :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Hacked in UDP server support in ngx_stream_lua and a simple Lua DNS server in NGINX gets 110K r/s /w a single worker
-
OpenResty
@OpenResty
OpenResty 1.11.2.1 RC1 is now available for testing. Many new features and bug fixes. Announcement:groups.google.com/d/msg/openresty-en/cNeiiH_Qhl4/Ses9XCd6… Feedback welcome!
-
OpenResty
@OpenResty
We already have a production user that uses OpenResty's shm lists and ngx.semaphore to run a push system doing 1400K+ conns per box on avg.
-
OpenResty
@OpenResty
It's great for building efficient C1000K pub/sub servers atop OpenResty with the ngx.semaphore API. Thanks Dejiang!
Yichun Zhang (agentzh)
@agentzh
The ngx_http_lua module now provides Lua API for shared-memory-based lists: github.com/openresty/lua-nginx-module Enjoy shm queues in @OpenResty web apps!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The shm dict objects' incr() method in ngx_lua now accepts an optional "init" arg for initializing nonexistent keys: github.com/openresty/lua-nginx-module
-
OpenResty
@OpenResty
The OpenResty official site now utilizes our new logo: openresty.org Thanks Yuansheng Wang for the patch.
-
OpenResty
@OpenResty
These will get included in the next OpenResty release, 1.11.2.1.
Yichun Zhang (agentzh)
@agentzh
ssl_session_fetch_by_lua* & ssl_session_store_by_lua* have just been merged to master, so have the ngx.ssl.session module of lua-resty-core.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
ngx_lua_upstream module now has a Lua API func to fetch the upstream target name for the cur req: github.com/openresty/lua-upstream-nginx-module thanks Justin Li.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Now lua-resty-redis supports lazily generating Lua methods for *any* Redis commands the user attempts to use, even for future Redis servers.
-
OpenResty
@OpenResty
Daidong Zhang has contributed a logo design to OpenResty. Thanks a lot!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Now almost all the OpenResty components' GitHub repo are added to Travis CI for regression: travis-ci.org/openresty thanks Ilya Shipitsin!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Rendered docs for the new nonblocking distributed SSL session caching feature in NGINX: github.com/openresty/lua-nginx-module/tree/ssl-session & github.com/openresty/lua-resty-core/blob/ssl-session/lib/…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Opensourced the ssl_session_fetch_by_lua* and ssl_session_store_by_lua* feature for distributed SSL session caching: github.com/openresty/lua-nginx-module/pull/822
-
OpenResty
@OpenResty
OpenResty now provides official pre-built Linux packages and repositories for RHEL/CentOS/Fedora: openresty.org/en/linux-packages.html Feedback welcome!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Lua scripts that send emails via Gmail's SMTP server (port 465 and 587) w/ cosockets only: gist.github.com/agentzh/d9288e4dd88941fc685598b98f74766a & gist.github.com/agentzh/d3b1a640f0b4d1fdb45497ccb00a8763
-
OpenResty
@OpenResty
Starting from 1.9.15.1, OpenResty formal releases are also uploaded to the GitHub openresty/openresty repo: github.com/openresty/openresty/releases/tag/v1.9.15.1
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Wrote a lengthy book section to explain the "Valgrind test mode" of my Test::Nginx: openresty.gitbooks.io/programming-openresty/content/testi… Using a buggy example w/ LuaJIT.
-
OpenResty
@OpenResty
The openresty.org official site now supports small screen devices. Thanks kindy for contributing the responsive design change.
-
OpenResty
@OpenResty
OpenResty 1.9.15.1 is out. Download: openresty.org/en/download.html Changes: openresty.org/en/changelog-1009015.html Announcement: groups.google.com/d/msg/openresty-en/JUShMoBzW1c/8MCJ0VvV… Enjoy!
-
OpenResty
@OpenResty
OpenResty 1.9.7.5 is out. Download: openresty.org/en/download.html Announcement: groups.google.com/d/msg/openresty-en/TqbgNJkzgaU/6t0da2HO… It includes the latest NGINX security fix.
-
OpenResty
@OpenResty
OpenResty 1.9.15.1 RC2 now available for testing. See the announcement for details: groups.google.com/d/msg/openresty-en/GcGLtmKdUE4/suhcr0k_…
-
OpenResty
@OpenResty
OpenResty 1.9.15.1 RC1 now available for testing: groups.google.com/d/msg/openresty-en/xGB24OSkdUI/hXY2B4ir… New NGINX 1.9.15 core, new LuaJIT core, & more. Feedback welcome!
-
OpenResty
@OpenResty
Join us at the Guangzhou OpenResty Meetup in China this Sunday! techparty.org/guangzhou/2016/04/14/techparty-openresty-ap…
-
OpenResty reposted
DreamHost
@DreamHost
We're now building our #nginx services on a new platform that gives you extra web security + development features: www.dreamhost.com/blog/2016/04/07/introducing-extra-web-s…
-
OpenResty reposted
Jesse Greathouse
@jessegreathouse
"The 5 Best technologies you've never heard of" www.linkedin.com/pulse/5-best-technologies-youve-never-he… @LuaJIT @OpenResty @typescriptlang #lxc @unity3d
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Now that NGINX's stream subsystem supports UDP, we can update our ngx_stream_lua_module to allow coding up generic UDP servers with Lua :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The new restydoc utility now works. Below is the screenshot for the command "restydoc -s lua_shared_dict ngx_lua" :)
-
OpenResty
@OpenResty
The new version of openresty.org is now alive! Built entirely upon the OpenResty stack backed by Postgres. Code is opensourced too
-
OpenResty
@OpenResty
OpenResty 1.9.7.4 formal is just out. Download: openresty.org Changes: openresty.org Thanks to all our contributors! :)
-
OpenResty
@OpenResty
OpenResty 1.9.7.4 RC1 now available for testing: groups.google.com/d/msg/openresty-en/4bYJRjC9UFI/UMAjgOxW… It has lots of bug fixes and small new features. Feedback welcome!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Just added details (slides, videos, and etc) of my OpenResty related presentations in 2015 to this page: openresty.org Enjoy! :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
We'll have our 1st Bay Area OpenResty meetup tomorrow! www.meetup.com/Bay-Area-OpenResty-Meetup/events/228411221/ Hear the stories of ppl from CloudFlare, Mashape, and Adobe.
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Planning to add an optional DSL layer to @OpenResty so we can program on a much higher level. Lua remains some kind of "VM machine code" :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I just opensourced the openresty-survey app built completely on OpenResty (& MySQL): github.com/agentzh/openresty-survey Still some rough corners though
-
OpenResty reposted
Ingmar Burguersson
@iburgueno
SeeRedis. Webapp for @OpenResty to see the data in a #redis instance, written in @LuaLang --> github.com/ignacio/seeredis /cc @kernelp4nic
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Okay, now the ngx_set_misc and ngx_http_lua modules can be compiled as "dynamic modules" with NGINX 1.9.11+ as well :) Testers welcome!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Now our NGINX modules ngx_echo, ngx_headers_more, and OpenResty's fork of ngx_eval can be built as "dynamic modules" with NGINX 1.9.11+ :D
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
@CloudFlare will host the first Bay Area OpenResty meetup at 5:30pm ~ 6:30pm on 9 March in its office. Details here: mailman.nginx.org/pipermail/nginx/2016-February/049774.html
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
ngx_stream_lua_module is now quite usable & passed all the relevant tests from ngx_http_lua_module. Feedback welcome:github.com/openresty/stream-lua-nginx-module
-
OpenResty
@OpenResty
OpenResty 1.9.7.3 is now released with the latest security fixes from the mainline NGINX core. See the ANN here: groups.google.com/d/msg/openresty-en/3Xyk0XzkI6Y/s53tN2ZL…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Just created the Bay Area OpenResty Meetup group on meetup.com. Let's meet each other in @CloudFlare's office next month!
-
OpenResty
@OpenResty
OpenResty 1.9.7.2 formal is just out. Download: openresty.org Changes: openresty.org ANN: groups.google.com/d/msg/openresty-en/U1PMBJbdVoA/DXUvdHlB… Enjoy!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
ngx_stream_lua_module can now do "hello world" TCP servers. Many more features are coming in the next few days :D
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
This is the first version of OpenResty that includes ngx.semaphore, balancer_by_lua*, and ssl_certificate_by_lua* :D
OpenResty
@OpenResty
OpenResty 1.9.7.2 RC1 now available for testing. Src: openresty.org/download/ngx_openresty-1.9.7.2rc1.tar.gz Win32: openresty.org/download/ngx_openresty-1.9.7.2rc1-win32.zip ANN: groups.google.com/d/msg/openresty-en/SZsbhGJc6D0/vJ0eHSIb…
-
OpenResty
@OpenResty
OpenResty 1.9.7.2 RC1 now available for testing. Src: openresty.org/download/ngx_openresty-1.9.7.2rc1.tar.gz Win32: openresty.org/download/ngx_openresty-1.9.7.2rc1-win32.zip ANN: groups.google.com/d/msg/openresty-en/SZsbhGJc6D0/vJ0eHSIb…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Hacking on the lemplate compiler that generates standalone Lua scripts for @OpenResty from Perl's TT2 templates: github.com/openresty/lemplate
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
I wrote a quick nginx.conf snippet to demonstrate the upcoming distributed SSL session caching feature in OpenResty.
-
OpenResty
@OpenResty
This will get included in the next OpenResty formal release, 1.9.7.2.
Yichun Zhang (agentzh)
@agentzh
Now ngx_lua's ssl_certificate_by_lua* directives have official documentation: github.com/openresty/lua-nginx-module/ & ngx.ssl too: github.com/openresty/lua-resty-core/blob/master/lib/ngx/s…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Just merged ngx_lua's ssl_certificate_by_lua* feature and the ngx.ssl & ngx.ocsp Lua modules into their git masters.
-
OpenResty
@OpenResty
The Chinese OpenResty community is setting up the OpenResty Software Foundation to support & reward OpenResty contributors worldwide.
-
OpenResty
@OpenResty
Chinese companies Smartisan Technology and Qingting.fm recently announced funding to support OpenResty: it.people.com.cn/n1/2016/0101/c223607-28003381.html Happy 2016!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Official documentation for the new balancer_by_lua* directive: github.com/openresty/lua-nginx-module and for ngx.balancer: github.com/openresty/lua-resty-core/blob/master/lib/ngx/b… Enjoy!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Merged balancer_by_lua* & ngx.balancer to git masters. Time to build dynamic load balancers for NGINX with Lua! ;)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The documentation for ngx.semaphore is here: github.com/openresty/lua-resty-core/blob/master/lib/ngx/s… Thanks Kugou Inc. for contributing this feature
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The ngx.semaphore API for synchronizing "light threads" across (request) contexts in the ngx_lua module has just been merged to git master.
-
OpenResty
@OpenResty
The only English talk on OpenResty Con 2015 Beijing was from Aapo Talvensaari. Video:www.youtube.com/watch?v=VqBt5icKCI8 Slides: iresty.com/download/ebook/2015_con/aapo.pdf
-
OpenResty
@OpenResty
OpenResty 1.9.7.1 formal release just out. Download: openresty.org Changes: openresty.org ANN: groups.google.com/d/msg/openresty-en/d79n1my9-Uc/NPA8Lu7m…
-
OpenResty
@OpenResty
Ambitious roadmap of OpenResty for the new year, 2016: groups.google.com/d/topic/openresty-en/1EfiZTjhGas/discus… We need help and support from all of you to make it real :)
-
OpenResty
@OpenResty
OpenResty 1.9.7.1 RC2 now available for testing. Source tarball: openresty.org/download/ngx_openresty-1.9.7.1rc2.tar.gz Announce: groups.google.com/d/topic/openresty-en/lpAVpbOEJn0/discus… Feedback welcome!
-
OpenResty
@OpenResty
Great! Enjoy OpenResty!
Christophe Dolivet @cdolivet
Fun week with first introduction of #nginx +#lua in production. Thx @OpenResty for the great package !
-
OpenResty
@OpenResty
OpenResty 1.9.7.1 RC1 now available for testing. Download: openresty.org/download/ngx_openresty-1.9.7.1rc1.tar.gz Announcement: groups.google.com/d/topic/openresty-en/jpJigSx1owo/discus… Feedback welcome!
-
OpenResty
@OpenResty
Welcome to join the OpenResty user survey: openresty.org/survey/ The OpenResty team wants to know more about our user community. Thanks!
-
OpenResty
@OpenResty
According to the current plan, the next OpenResty release will be 1.9.7.1, which is based on the NGINX 1.9.7 core. The ETA is Christmas! :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
NGINX 1.9.5+ removes the (200) limit on the # of subrequests per main request. Time to start 1000 parallel subreqs!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
balancer_by_lua* + keepalive in action :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
ngx_lua's balancer_by_lua* (on the balancer-by-lua github branch) can now work with the std connection pool. thanks doujiang for the patch.
-
OpenResty
@OpenResty
Now we have the openresty-devel mailing list for the OpenResty development team: groups.google.com/group/openresty-devel GitHub is not enough for us :)
-
OpenResty
@OpenResty
OpenResty 1.9.3.2 is just out. Download: openresty.org Changes: openresty.org Announce: groups.google.com/d/topic/openresty-en/O5P_0Bmu-gI/discus… Enjoy!
-
OpenResty
@OpenResty
OpenResty 1.9.3.2 RC2 now available for testing. Src: openresty.org/download/ngx_openresty-1.9.3.2rc2.tar.gz Win32 bin: openresty.org/download/ngx_openresty-1.9.3.2rc2-win32.zip Ann:groups.google.com/d/topic/openresty-en/nKV1IOcH34Y/discus…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
After days' hacking, @OpenResty's build system now works with the MinGW compiler toolchain to produce Win32 binaries:github.com/openresty/ngx_openresty/blob/master/doc/README…
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
ngx_lua module's balancer_by_lua* is opensourced here: github.com/openresty/lua-nginx-module/tree/balancer-by-lua tests as docs:github.com/openresty/lua-nginx-module/tree/balancer-by-lu…
-
OpenResty
@OpenResty
OpenResty 1.9.3.2 RC1 now available for testing. Download: openresty.org/download/ngx_openresty-1.9.3.2rc1.tar.gz Announcement: groups.google.com/d/topic/openresty-en/4dW-3dNM_mM/discus… Feedback welcome!
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
The OpenResty China community is busy preparing the 1st OpenResty conference in Beijing next month: www.iresty.com I'll be there :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
My first C module based on NGINX's new "stream" subsystem, ngx_stream_echo: github.com/agentzh/stream-echo-nginx-module Set up a silly TCP server in seconds! :)
-
OpenResty reposted
Yichun Zhang (agentzh)
@agentzh
Wrote a lua-resty-limit-traffic library for more flexible req rate and concurrency limiting & adjustment in NGINX: github.com/openresty/lua-resty-limit-traffic
-
OpenResty
@OpenResty
OpenResty formal release 1.9.3.1 just out: groups.google.com/d/msg/openresty-en/mWwzwh5wFe0/NnlorTSR… Download: openresty.org Changes: openresty.org Enjoy!
-
OpenResty
@OpenResty
OpenResty 1.9.3.1 RC1 now available for testing. Download: openresty.org/download/ngx_openresty-1.9.3.1rc1.tar.gz Announcement: groups.google.com/d/topic/openresty-en/zEiZXK97x9s/discus… Feedback welcome!
-
OpenResty
@OpenResty
The folks at Qihoo 360 have been working on a (free!) book titled "OpenResty Best Practices": www.gitbook.com/book/moonbingbing/openresty-best-practice… It's in Chinese though
-
OpenResty
@OpenResty
OpenResty 1.7.10.2 is just out. Download: openresty.org Changes: openresty.org Thanks to all our contributors! Enjoy!

























































































































































