Closed simsicon closed 10 years ago
Hi,
PumaWorkerKiller doesn't seem to be able to kill my Puma workers:
PumaWorkerKiller: Out of memory. 1 workers consuming total: 522.10546875 mb out of max: 506.88 mb. Sending TERM to #<Puma::Cluster::Worker:0x007f00a761a6f8 @pid=8, @phase=0, @stage=:booted> consuming 506.47265625 mb.
I have this line in my logs since 25 minutes.
The configuration in my Procfile is the following:
web: bundle exec puma -p $PORT -w 2 -t 0:9
Looks like I was using a version of Puma which is incompatible with puma_worker_killer ; after upgrading from 2.5.1 to 2.8.1, it works as expected. Maybe the gemspec for puma_worker_killer could be updated.
sorry for the delay here, github doesn't watch repos you create automatically for me, so i'm just now getting this. Could someone give me a PR to rev the gemspec, I would :heart: you so much, then we can close this issue. We can do ~> 2.7 as I originally wrote and tested this against 2.7 until i realized my version was behind. We should also update https://github.com/schneems/puma_auto_tune
I've reved-the spec and it's on master now, releasing a new gem soon.
Hi, thanks for providing this Gem.
I am current using puma 2.8.0 and rails 3.2.13, running 4 workers on cluster mode. I added
puma_worker_killer
to my app, put those configs andPumaWorkerKiller.start
in myproduction.rb
, but processes went beyond 512 mb which was the RAM value I set.I took look into the code, realized this gem tries to term processes first, or kill them if failed, I don't know why my processes were so "vital". I guess there's a memory leak in my code, I can't find it right now, because I can't reproduce the problem in my staging or development env. Any ideas? What info I can put here would help?
Thanks.