Health Check

[vbaidak]

Hi Team,

Is there any health check command?

[m90]

What's the failure scenario you would like to check for with such a health check? Right now, as long as the container did not exit, things are assumed running as configured (which obviously might not be true).

However, if this image would implement a health check, I wouldn't even know which conditions to check for. Do you have something specific in mind?

[vbaidak]

Yeah, that's why I'm asking. I'm ok with just having unless-stopped condition and asking only for the case if there is any process inside a container that is running and could potentially fail without container fail, so that we are not getting backups and do not know about this

[m90]

The process in the container works kind of like crond and is designed to never ever fail and instead just run the scheduled jobs, no matter if these invocations fail or not. This means the container being up does not tell you anything about whether backups have successfully been performed or not, the only guarantee you get is that it has been tried. I.e. you could have a container that is up for over a year but hasn't performed a single backup as you have accidentally provided bad credentials for your storage backend.

If you want job-level visibility instead, you can either monitor your logs or use the notification feature, which can notify you about either all runs or just failed ones. Docs are here https://offen.github.io/docker-volume-backup/how-tos/set-up-notifications.html

There are also ideas about exposing metrics via HTTP (e.g. Prometheus), but I don't know how / if this will ever happen: #64

[vbaidak]

Thanks for the explanation!
Well, I've already configured email notifications and they work well, so my backup failures will be covered.
What about the process itself? Maybe we can at least check if it is running via ps or does it have PID 1?

[m90]

What about the process itself?

This should be the responsibility of Docker or whatever tool runs the container in that case. If the process exits, the container does so as well, and in this case whoever is supervising the container should be taking care of whether it wants to restart the container or not.

[vbaidak]

I see