load average went from 8.00 to 0.21 with this one simple trick 🐧
Discussion
You can’t post this and not share the "simple trick" 🤣. It’s against the Geeky Linux User Constitution.
this one might be too valuable, gotta make a living somehow 😂⚡
Linux Gulag for you mate. You will be sent to the Stallman mountains to mine AGPL until you repent 🤣.
I have my own trick to decrease server load. It involves the "kill - 9" command and serving a cached static website 😅 instead of my usual stuff.
😂😂😂 it will all be included in the gplv3 relay tools, don't worry.. 😂
ok, since u zapped.. i turned off read ahead in strfry, which i had already thought was off. 🐧prob not a surprise to anyone who read the config and didnt get their booleans confused like i did 😂
Life-changing "V2V" interaction right here! Am I doing the "Nostr influencer" thing right?
Lol, LMDB strikes again! I assume you're talking about LMDB, right? I don't run strfry, but LMDB readahead and mapsize caused some fun bugs in Khatru relays too. Well done, regardless! I'm not operating at the scale of relay.tools or nostr1.com, but I know the feelingnwhen a struggling server "comes back to life" with a bit of tuning.
My last moment like this was when I figured out how to drop incoming packets from a specific Cloudflare outbound IPv6 (used by all cross-region Cloudflare Worker requests). A few of my relays went from being DDoSed into oblivion (to the point where even the log writer couldn’t keep up with all the logs about rate-limited requests) to a smooth 7% or less CPU usage. It feels good. A kind of good that folks outside DevOps probably won’t ever experience. It’s a very lonely "feel-good" moment.
