Does anyone who knows stuff about Akkoma/Pleroma have an hour or two of their time to help me fix issues?
I think the issues are IO/Postgres/OBAN queue related.
Conversation
Notices
-
Embed this notice
SuperDicq (superdicq@minidisc.tokyo)'s status on Saturday, 13-Jul-2024 23:14:31 JST SuperDicq -
Embed this notice
SuperDicq (superdicq@minidisc.tokyo)'s status on Saturday, 13-Jul-2024 23:22:34 JST SuperDicq In general. IO usage is very high, OBAN queue is not being processed, DBConnection.ConnectionErrors being spammed in the logs.
-
Embed this notice
SuperDicq (superdicq@minidisc.tokyo)'s status on Saturday, 13-Jul-2024 23:30:10 JST SuperDicq @mr64bit@p.mr64.net I've been been playing with it trying to find a good amount, currently it is at default which is 50 I believe.
I've been playing with setting trying to increase/decrease like trying 20 or 80 to no avail. My max_connections in postgres is currently set to 50, but I also tried different settings like 20 or 80.
The only other thing using postgresql is this Misskey instance, which runs fine. -
Embed this notice
mr64bit (mr64bit@p.mr64.net)'s status on Saturday, 13-Jul-2024 23:30:11 JST mr64bit @SuperDicq How big is the database pool size in your pleroma config vs max_connections in your postgresql.conf?
(and is there anything else connecting to postgresql?) -
Embed this notice
SuperDicq (superdicq@minidisc.tokyo)'s status on Saturday, 13-Jul-2024 23:34:34 JST SuperDicq @mr64bit@p.mr64.net I don't know if a higher or lower setting is better considering my server specs: 16 CPU cores, 32 GB RAM, HDD storage (aka plenty of cpu and ram, but slow storage).
-
Embed this notice
SuperDicq (superdicq@minidisc.tokyo)'s status on Saturday, 13-Jul-2024 23:43:51 JST SuperDicq @mr64bit@p.mr64.net Yes, it's probably a timing out issue.
-
Embed this notice
mr64bit (mr64bit@p.mr64.net)'s status on Saturday, 13-Jul-2024 23:43:52 JST mr64bit @SuperDicq Got it. I was thinking maybe your `DBConnection.ConnectionError`s were from a lack of available connections on the database, but sounds like that's not it. From what you're describing, sounds more like queries timing out, but I'd expect a different error in that case.
-
Embed this notice