Conversation

Notices

1. Embed this notice
   :suya: (newt@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:25 JST :suya:
   My plan right now is to fetch the newly made copy of our database home and try to reindex it. If it takes under a day, I'm willing to do the same on the main server, taking it offline.
   • Embed this notice
     lainy (lain@lain.com)'s status on Sunday, 12-Feb-2023 05:10:14 JST lainy

     in reply to

     • karolat
     @karolat @newt i recommend to leave it at the default 10 connections.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:16 JST karolat

     in reply to
     @newt I had Postgres at 25 and pleroma at 20 I thought. I guess it depends on which queue it’s better to have requests sitting in, I wonder if having so many connections open can degrade query performance further
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:17 JST :suya:

     in reply to

     • karolat
     @karolat yup. But it was more than what Postgres allowed still.
     
     Anyhow, I’m thinking setting it back to 50. Should be better.
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:18 JST :suya:

     in reply to

     • karolat
     @karolat you also changed database settings. I actually had to fix after you, because you decreased the number of allowed connections in postgres but didn't do the same in pleroma settings.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:18 JST karolat

     in reply to
     @newt I did, I experimented with 10 and 20 and settled on 20 for pleroma. That’s the max pool, right?
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:19 JST karolat

     in reply to
     @newt what’s strange is that after I did the analyze we constantly get 500s when viewing profiles. I don’t thin this was happening before, or at least I didn’t notice it
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:20 JST :suya:

     in reply to

     • karolat
     @karolat nope. Now you need an index for an index lol.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:21 JST karolat

     in reply to
     @newt with all those indexes you think the queries would be pretty fast :blobthinkglare:
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:22 JST :suya:

     in reply to

     • karolat
     @karolat indexes for it are bigger than the table itself. If reindexing solves it, then we're all good and can repack the rest.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:23 JST karolat

     in reply to
     @newt maybe we can just take it offline to vacuum full the activities, and then unpack the rest online :blobcatcomfthink:
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Sunday, 12-Feb-2023 05:10:24 JST :suya:

     in reply to

     • karolat
     cc @karolat
   • Embed this notice
     lainy (lain@lain.com)'s status on Sunday, 12-Feb-2023 06:11:38 JST lainy

     in reply to

     • karolat
     @karolat @newt you don't need more than 10. setting it higher won't help, that's why there you have to override it with a setting to even make it start.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Sunday, 12-Feb-2023 06:11:39 JST karolat

     in reply to

     • lainy
     @newt @lain yes, newt set it to 100 I think, so 20 was a compromise :comfyxdsweat:
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Sunday, 12-Feb-2023 06:11:40 JST :suya:

     in reply to

     • lainy
     • karolat
     @lain @karolat we tried that some long ago and it didn't work properly. Setting it to 50 did the trick.