Alcun Atirutan BBS

clicking through to someone else's pleroma/akkoma and it takes a minute to load the page experience.

@deutrino there's a website that tunes your postgres settings based on hardware specs and that is 99% of your way there.

@Moon @deutrino https://pgtune.leopard.in.ua/

@i @deutrino @Moon I found that switching the busy-wait off according to Pleroma docs improved performance. I have to say though, I'm on a critically low instance for $3/month with only 1 CPU. So I guess the CPU time previously wasted by BEAM went to Postgres. Not likely to do anything to anyone running a real server.
https://docs-develop.pleroma.social/backend/configuration/optimizing_beam/

@kazriko @i @deutrino @pro we have 128GB of RAM and normally we run about 16GB max, but sometimes under heavy load we would get OOMkilled too.

@kazriko @i @deutrino @pro it's heavily a function of how many connections your server has, I can't tell you exactly but people do run pleroma successfully on low ram but yeah you ight just get an oom sometimes when load jumps up. it hasn't been remotely as bad lately as during the last big twitter migration here, we were getting errors very frequently and I haven seen one for over a month now that you mention it. anecdotally, fediverse interest has dropped off tremendously since that influx.

@kazriko @deutrino @pro @Moon some swap on a write-enduring volume ought to help

@kazriko @i @deutrino @Moon I have this on the instance that runs Mu:

[root@vps221183 ~]# free
total used free shared buff/cache available
Mem: 1967900 612476 527060 152892 828364 1037024
Swap: 2064380 495120 1569260
[root@vps221183 ~]# uptime
20:04:21 up 147 days, 19:25, 2 users, load average: 0.06, 0.09, 0.03
[root@vps221183 ~]# df .
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/vda1 41193916 19738056 19719980 51% /
[root@vps221183 ~]#