fix(nginx) wait for nginx reload to avoid zombies #6
Conversation
If a unix process is spawned, the parent process must wait for its pid or the process will become a zombie after exit. This fixes the problem that the "/opt/router/sbin/router" binary leaves behing a defunct nginx process each time the config is reloaded by spawning "nginx -s reload".
|
@felixbuenemann : Thank you for the PR! We are going to merge this in now because we trust that it has been tested on your side and it is a very small isolated change. By the way, are you using Workflow in production? Are you interested in joining the fork as a maintainer? |
Yes, we are currently running 2 Staging and 2 Production clusters with Deis Workflow on AWS.
I'm already way too busy and we might switch to vanilla k8s or possibly convox in the future – but it is nice to have this fork as an alternative. In the mean time I will send pull requests with fixes to hephy, if I have any. |
Okay, no problem. Nice of you to keep the fork in mind.
Okay, thank you for your help! 👍 |
If a unix process is spawned, the parent process must wait for its pid or the process will become a zombie after exit. This fixes the problem that the /opt/router/sbin/router binary leaves behing a defunct nginx process each time the config is reloaded by spawning "nginx -s reload".
Copied from deis/router#356 by request of @Cryptophobia to fix deis/router#331.