Jelte Fennema-Nio

were you able to get some answers to the 3 questions in my previous item? Also another follow up question: What OS are you using? Because maybe it's related to...

@brnck it sounds like that could be the same issue as what @jarojasm95 reported, but it might also be a separate separate issue. Since you made a bigger version jump,...

No worries and sounds good. FYI I would be extremely surprised if v1.20.1 makes a difference, it has a small behavioral change if you have `options` in `ignore_startup_parameters` (which you...

@brnck > Looks like the scale is too small or I don't hit some certain condition to reproduce the issue. What do you mean with this? If I look at...

@brnck @jarojasm95 I'd still like to get to the bottom of this regression. Have either of you been able to try out this branch to see if it solves your...

@lili2311 Could you try one or both of these two things, to help debug the issue: > Two things that would help getting to the bottom of what is causing...

@stafot thanks for trying out the branch, too bad it didn't help. Could you share your pgbouncer config file? That might help me find the reason. @pablogamboa @lili2311 could you...

Okay then, that's probably not it. Still full (anonymized) config would be useful. Any chance you could try this to get a sense of where the extra CPU is spent:...

> @JelteF we tracked down the performance regression to `scram-sha-256`, it's not 1.20 specific as we reproduced it on 1.19.1 as well. We had `scram-sha-256` and our passwords in `userlist.txt`...

FYI pgbouncer uses TCP_NODELAY. I'm not sure if you meant disabling NOdelay or disabling THE delay. https://github.com/pgbouncer/pgbouncer/blob/e6ce619785c93392794976c1e936e8c9f589a5ad/src/util.c#L229