SideKiq "stopping"


(Michael - DiscourseHosting.com) #1

Ok, last report for today :smile:

Nothing urgent here but just wanted to let you know that we sometimes still see sidekiq eating up some CPU. This always concerns old sidekiq processes which should have been stopped. Their command line always says “stopping”.

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                                                                                                                    
22336 www-data  20   0 1438m 106m 1136 S   25  5.3   3535:16 ruby1.9.1                                                                                                                  
 6012 root      20   0 17332 1272  920 R    1  0.1   0:00.04 top                                                                                                                        
14680 www-data  20   0  513m 193m 8520 S    1  9.7   1:13.18 ruby1.9.1                                                                                                                  
14716 www-data  20   0  513m 192m 8520 S    1  9.6   1:11.32 ruby1.9.1      

root@basic218:~# ps ax|grep side
 6023 pts/0    S+     0:00 grep --color=auto side
14526 ?        S      0:00 sudo -u www-data RUBY_GC_MALLOC_LIMIT=90000000 RAILS_ENV=production nohup bundle exec sidekiq
14530 ?        Sl     6:39 sidekiq 2.13.0 discourse [0 of 25 busy]                          
22333 ?        S      0:00 sudo -u www-data RUBY_GC_MALLOC_LIMIT=90000000 RAILS_ENV=production nohup bundle exec sidekiq
22336 ?        Sl   3535:29 sidekiq 2.13.0 discourse [0 of 25 busy] stopping

So 14530 is the new sidekiq, 22336 is the old one.
We have tried to stop it by sending a kill signal to the PID, which obviously arrived (since the “stopping” text appeared), but causes it to loop somewhere, somehow.

This happens about 1/10 of the time sidekiq is stopped.


(Jeff Atwood) #2

Still an issue? We are switching to Unicorn across the board now with Rails 4. And we want to kill off bluepill but the other alternatives are kind of crap too.


(Michael - DiscourseHosting.com) #3

Somehow, no. The Sidekiq problems disappeared about a week ago, probably with v0.9.7.6, although I’m unable to locate a change that could have solved this…

Will the move to Unicorn also affect Sidekiq then?

Bluepill is crap indeed. We’re using plain old shell scripts for process control.


(Sam Saffron) #4