docker-compose.yaml: Restart on failure #360
Conversation
In kubernetes it is expected containers might fail and quit, so we can duplicate this behavior in docker as well. While failure is not normal and we should avoid it, but we should not assume it is critical issue in container environment. Signed-off-by: Denys Fedoryshchenko <denys.f@collabora.com>
There was a problem hiding this comment.
I think the reason why this wasn't enabled initially is because the docker-compose deployment is mostly meant to be used by developers, and it's better to not automatically restart services when they're failing while working on the code as it just gets in the way.
Whatever works best for developers. If this was to address production use-cases then I don't think it's a good idea. If it's because developers generally find it useful then it probably makes more sense.
@RCN What do you think?
|
Personally, while debugging I prefer if things don't restart automatically after they crash so I can notice and go through the stack trace. Maybe we can keep these changes if they're useful for the general public and split the current version (without the |
|
I think I agree with the collective and reasoned opinion. The only problem, like recently during staging issues, when the server was overloaded, all the containers “died” due to the fact that containers couldn't contact the API server or due to problems with DNS resolving. On the other hand, they should not “crash” if they experience temporary problems with the network... |
In kubernetes it is expected containers might fail and quit, so we can duplicate this behavior in docker as well. While failure is not normal and we should avoid it, but we should not assume it is critical issue in container environment.
Failure might be expected in situations like this: #358
Or for example if API service was temporary unreachable.