Conversation
Notices
-
@keithzg @feld does postgres die? 1gb is very tight...
-
@feld @lain At this point I think I can confirm that RAM was the issue, as I have not seen this recur since I bumped the VM up to 2GB from 1GB.
This does make my leading theory (though not with anything approaching absolute certainty) that the current version of #Pleroma is more demanding in this regard than the ancient version I was on, because this certainly never was an issue before my upgrade.
It's possible that increased Fediverse traffic is to blame instead, but it was several months into this latest surge that I performed this upgrade, so the second-place suspect for me is just Ubuntu itself. I upgraded to Ubuntu Server 22.04 before upgrading Pleroma and didn't observe any problems until the Pleroma upgrade (at which point they started occurring within the hour at the time), but that could certainly have been just a(n un)lucky coincidence.
-
@lain @feld Unfortunately I can't give a very specific version where this might have been introduced, if indeed it's actually a bug, because I was upgrading from a very old version (mid-2019).
-
@lain @feld I don't thiiink so, I swear I've checked that at times when I've been getting the hard failure and when there's soft failures where some requests fail and I see popups in the web interface along the lines of
> Error fetching timeline: Unexpected token '<', "<html> <h"... is not valid JSON
That being said, in the logs for the Pleroma service itself I do at times see:
> (DBConnection.ConnectionError) connection not available and request was dropped from queue after 2999ms. This means requests are coming in and your connection pool cannot serve them fast enough.
So maybe that's indeed the problem. Although it certainly wasn't a problem *before* I upgraded, and it immediately started happening after I did.
LS
keithzg
All ０７６萌ＳＮＳ content and data are available under the