Closed maingoh closed 5 years ago
Any advise ? Is it something you would be interested in ?
I don't have a strong opinion on this. I'm not sure Circus is the best tool for managing several processes in a container, especially with Kubernetes. But I can see the appeal.
Maybe it could be implemented with a hook? (https://circus.readthedocs.io/en/latest/for-devs/writing-hooks/)
Thank you, I will take a look at the hooks !
I don't know any alternative except supervisord but I guess it will be the same issue ? It is not always easy to have only one process per container.
So I tried all hooks but unfortunately none of them allow me to check if a process crashed with a bad exit code.
Would you be interested If I provide a on_failure
or any other hook name ?
I added a new type of hooks in #1104 to solve my use case. This allows to know when a process is reaped. Could you take a look ?
Done thanks to after_reap
hook
I am using circus to handle multiple processes inside a container, but when used with kubernetes it actually hide the crash to kubernetes and we don't know that the container is failing until we watch the logs. How complicated would it be to add an option to circus endpoint or each watcher to quit circus when the watchers crash ? So that we can let kubernetes restarting the pod.
I would be ready to implement such behavior, Thank you,