Richard Schneeman
Richard Schneeman
The warning that puma emits links to this issue. We should keep this open until people can safely ignore that warning.
Thanks! Can someone validate the original repro with puma is comparable to running without the env set? The warning message links to this issue. We should scope it to not...
That is very unexpected. Maybe we change the warning from “turn it off” to “turn it on” in 3.5/4.0?
This PR touches on some related changes, I need to review https://github.com/puma/puma/pull/3582
> Warning[:performance] = true TIL https://alchemists.io/articles/ruby_warnings
I was able to get this working via `brew install rsync` (on a mac) which fixes the problem. I'm unsure how to install the latest version of openrsync on my...
At the time I was frustrated having to handle the no queue case for fixing keepalives. I am still interested in where in the internals we could simplify things by...
Worth noting that Puma 7 removed `wait_until_not_full` behavior. So the semantics have changed, yet again. We handled the no queue case for keepalives by closing them, but didn't consider the...
The reason given is to reduce memory fragmentation. We're mitigating that in glibc via setting this environment variable by default https://github.com/heroku/heroku-buildpack-ruby/blob/633a1d170155c279e4a5803fc8352d93108ec1b9/changelogs/v205/malloc.md?plain=1. I'm possibly interested in exploring doing something with jemalloc,...
Similar issue from earlier https://github.com/puma/puma/issues/2625 h/t @slizco