Lei retweeted
Heading to KubeCon + CloudNativeCon North America in Salt Lake City? Come find us at Booth 262, Nov 10–12.
We'll be running a live demo there: an OpenTelemetry pipeline sends metrics, logs, and traces over OTLP into one GreptimeDB instance, and dashboards and AI-assisted investigation query that same data.
Our CEO @killme20082 , CTO @Sunng , and engineer @ratuthomm will be there. Bring your observability storage bill and your migration questions.
#KubeCon
greptime.com/blogs/2026-10-2…
Lei retweeted
A lock-free MPSC ring buffer in Rust that the compiler was perfectly happy with.
It passed its full test suite, a Quint model check, loom's exhaustive interleaving search, and cargo Miri test, and yet it had been undefined behaviour (UB) from the very beginning: a genuine data race, in code where every slot each thread touched was disjoint and every atomic was correct.
Here is how Miri finally caught it, and why the borrow checker was never going to. Here's a Sunday evening blog post that gets into the depth of the reasoning -
How Miri found a data race in a lock-free ring buffer that the compiler had no way to see ..
debasishg.github.io/blog/dat…
买了个中兴 F50pro，到货才发现虽然名叫 pro 但是比起 F50 USB3.0 阉割成了 2.0，tethering 协议也改了没法直接 TypeC 转 RJ45 共享给其他路由器。后来 hack 了一下走无线桥接，路由器上再接树莓派做旁路由，用送的电信流量卡勉强能跑到 240Mbps。
我在芝加哥、京都、莫斯科、格拉纳达都逛过当地的旧货市场。看着父母辈当年视若珍宝的徽章、照片、日记被当作废纸和垃圾一样打包出售是我对“拥有自己的后代”这个事情的抗拒感的最初来源。既不能指望他/她从精神上理解你也能不指望物质上赡养你，那还是先过好自己的生活更重要。
Looks like the "zero-cost abstract" magic has both mental and runtime overhead😅
🦀 Async Rust has some really subtle behaviors you need to understand.
One nasty edge case: future cancellation + tokio::select! can quietly land you in a deadlock that’s brutal to debug.
This write-up is a great deep dive: rfd.shared.oxide.computer/rf…
#rust #rustlang
the best time to learn these principles was three years ago, and the second-best is now
After 3.5 years building GreptimeDB (@Greptime ), here's what I wish I knew before starting:
The harsh reality: We jumped into execution without proper business validation. Our idea wasn't wrong, but we skipped the unglamorous work of talking to 50-100 potential customers first.
What this cost us:
1. Chased SaaS/PLG strategies that looked good on paper but didn't match market reality for data infrastructure
2. Built features for "someday" instead of "today's customers"
3. Underestimated what it takes to build a database - it's a 5-year minimum commitment
What actually worked:
1. Finding co-founders with resilience, not just skills, couldn’t have asked for a better partner than @Sunng
2. Treating everyone as adults - transparency over "protecting the team"
3. Listening to the harshest critics first
The biggest lesson: Find 5-10 design partners who will run with you before writing code. The fear of "custom work" is pointless when you haven't proven PMF yet. That’s like worrying about scaling before you even have users.
Building infrastructure is a long game. Snowflake spent 8 years before their IPO. MongoDB built community for 9 years before cloud revenue mattered.
For data infra founders: the path is longer and messier than it looks. But find the right people, stay honest about what's working and what isn't, and keep shipping. ✊
#startups #databases #opensource #productmarketfit #founders #datainfrastructure #lessonslearned
Lei retweeted
After 3.5 years building GreptimeDB (@Greptime ), here's what I wish I knew before starting:
The harsh reality: We jumped into execution without proper business validation. Our idea wasn't wrong, but we skipped the unglamorous work of talking to 50-100 potential customers first.
What this cost us:
1. Chased SaaS/PLG strategies that looked good on paper but didn't match market reality for data infrastructure
2. Built features for "someday" instead of "today's customers"
3. Underestimated what it takes to build a database - it's a 5-year minimum commitment
What actually worked:
1. Finding co-founders with resilience, not just skills, couldn’t have asked for a better partner than @Sunng
2. Treating everyone as adults - transparency over "protecting the team"
3. Listening to the harshest critics first
The biggest lesson: Find 5-10 design partners who will run with you before writing code. The fear of "custom work" is pointless when you haven't proven PMF yet. That’s like worrying about scaling before you even have users.
Building infrastructure is a long game. Snowflake spent 8 years before their IPO. MongoDB built community for 9 years before cloud revenue mattered.
For data infra founders: the path is longer and messier than it looks. But find the right people, stay honest about what's working and what isn't, and keep shipping. ✊
#startups #databases #opensource #productmarketfit #founders #datainfrastructure #lessonslearned
每次看到同事摸鱼他都说他在等编译，这么看 Ruast 真是打工人的天命之选，既能以 target 目录太大为借口要求老板买高配 MacBook，又能在 cargo build 的时候打开屎黄色网站刷两把，只有老板受伤的世界达成了。