Conversation

Notices

1. Embed this notice
   Alex Gleason (alex@gleasonator.com)'s status on Thursday, 02-Feb-2023 08:20:36 JST Alex Gleason

   in reply to

   • anime graf mays ?️?
   • Matty
   • Matty
   • matty

   @graf @matty @matty @matty Is the theory that workers are overloading resources so Postgres can’t do a query?

   I’d love to see htop not overloaded and see what happens.

   • Embed this notice
     anime graf mays ?️? (graf@poa.st)'s status on Thursday, 02-Feb-2023 08:20:37 JST anime graf mays ?️?

     in reply to

     • Matty
     • Matty
     • matty
     @matty @matty @matty @alex definitely something not right with the workers if thats the case. it took me like a week of messing with queues to get the delicate balance we needed. i think i ended up on 30 for notifications and 100/100 for federator incoming/outgoing. i wouldnt suggest that at all for you, ours is that high due to traffic + postgresql being on its own epyc server
   • Embed this notice
     matty (matty@poa.st)'s status on Thursday, 02-Feb-2023 08:20:40 JST matty

     in reply to

     • anime graf mays ?️?
     • Matty
     • Matty
     @graf @matty @matty @alex No, I mean it takes a LONG time for the notifications page to load. But, with the lower connection count, it doesn't load notifications at all. I've just reset most of the configuration to default including rate limiters and workers, then set the maximum connections to 50 and the timeout to 60_000 to see if we can get federation going again.
   • Embed this notice
     anime graf mays ?️? (graf@poa.st)'s status on Thursday, 02-Feb-2023 08:20:41 JST anime graf mays ?️?

     • Matty
     • Matty
     • matty
     @matty @matty @matty not sure which account you're on right now but i see a delay here for our notifications on soapbox FE. if I use tootdeck.poa.st or husky they come in instantly but they seem to be polled and updated every 3-5 seconds. is this what you mean by delayed? @alex