In an effort to update our Forum I did a fresh VPS install + Restore on the weekend. This was going to solve multiple issues for us: Renew outdated ubuntu Update Discourse Upgrade to Postgres 15 While generally things went fine I saw issues coming up afterwards with Postgres processes going wild…

Right now it’s only one sidekiq task running with “Jobs::BackfillBadge” but still 7 postgres processes are seemingly blocking 100% cpu constantly. Really curious what’s going on there. [Screenshot 2025-02-17 @ 14.59.12@2x]

[圖片] Helmi: I did a fresh VPS install + Restore on the weekend. After such moves it is a good idea to run a vacuum for database stats.

How much ram and cpu do you have? How much memory do you give postgres?

That test serer is on 32GB, 16 core, config is set to 64MB work mem. EDIT: Shared buffers are at 8GB Currently doing a vaccuum which looks stuck [Screenshot 2025-02-17 @ 17.00.32@2x] Not sure if it’s doing something but it’s been there for 30+ minutes already. I did put the forum to read only …

You really don’t want to run a vacuum full. All you need to get performance back is a VACUUM VERBOSE ANALYZE. You can’t run a FULL in a running site.

I’m not expert at huge databases, but I’d make the buffers two or three times that. I’m sure you have indexes that are 8gb

[圖片] Jay Pfaffman: I’m not expert at huge databases, but I’d make the buffers two or three times that. :thinking: Postgres recommends never setting shared_buffers over 40% of the internal memory? That said, [圖片] Helmi: our database is quite huge - we’re at around 150gb after the restore…

[圖片] Richard - Communiteq: Postgres recommends never setting shared_buffers over 40% of the internal memory? Aha! Sensible advice from an expert! So maybe I was right that 8GB/25% isn’t enough, and though 16GB is more than 40%, it might still be a good suggestion because . . . . [圖片] Rich…

Guys. as said this is a test server - there is no traffic on it. This server definitely isn’t good enough for production use but that’s not the issue here. The question is why we do see postgres processes sticking like that (with 100% CPU usage) and slowing down things drastically. We were running t…

重装/恢复后Postgres进程无限运行和性能差

支持

helmi (Helmi) 2025年二月17日 14:00 2

现在只有一个 sidekiq 任务在运行“Jobs::BackfillBadge”，但仍有 7 个 postgres 进程似乎一直在阻塞 100% 的 CPU。真想知道那里发生了什么。

話題		回覆	觀看
Slow Sidekiq + Postmaster using 95%+ CPU (32 cores) after Postgresql Version Upgrade Self-hosting server-resources	22	3513	2020年五月29日
Postgres using 100% of CPU Self-hosting server-resources	16	20180	2018年八月25日
Post 3.5.1 update - CPU maxed out 100% Support	4	198	2025年二月4日
Primary Postgres database process (postmaster) eating all CPU Self-hosting server-resources	13	4414	2020年五月28日
Server load issues (after rebuild hanged) Self-hosting	4	276	2024年五月19日

重装/恢复后Postgres进程无限运行和性能差

相關主題