feat: ignore removal error due to non-existing containers

[nothub]

This PR allows watchtower to ignore a failure to remove containers that doesn't exist. This can be useful when the start of the containers is managed by docker compose or an external system such as systemd.

Watchtower causes an issue for me in combination with the --abort-on-container-exit flag for dockers compose up subcommand.

Docker compose removed the container on its own after the container was stopped. This leads to an error when watchtower attempts to remove the container.

Error: No such container: 425173...

The error is thrown by this ContainerRemove call.

Fixes #1480

[github-actions]

Congratulations on opening your first pull request! We'll get back to you as soon as possible. In the meantime, please make sure you've updated the documentation to reflect your changes and have added test automation as needed. Thanks! 🙏🏼

[codecov]

Codecov Report

Base: 64.76% // Head: 66.19% // Increases project coverage by +1.42% 🎉

Coverage data is based on head (0c9030d) compared to base (a4d00bf).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1481      +/-   ##
==========================================
+ Coverage   64.76%   66.19%   +1.42%     
==========================================
  Files          24       24              
  Lines        2310     2313       +3     
==========================================
+ Hits         1496     1531      +35     
+ Misses        718      682      -36     
- Partials       96      100       +4     

Impacted Files	Coverage Δ
pkg/container/client.go	41.29% <100.00%> (+9.45%)	⬆️
pkg/container/container.go	51.76% <0.00%> (+1.17%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[piksel]

Just skipping all container removal sounds a bit inflexible, and only fixes the problem if all containers are set to auto-remove.
It's probably better to ignore the error and try to create the new container anyway.
Perhaps we can even detect that the container is set to auto-remove (should be the same as docker run --rm.

[nothub]

Yes you are right, checking for each container is a better choice.
I pushed a change with that behaviour.
The error is now skipped if it is of type ErrNotFound.

[nothub]

Is there anything else you want me to change for the PR to be accepted?

[piksel]

I mostly haven't had time. Sorry. I will get to reviewing it as soon as I get some spare time.

[piksel]

Yeah, watchtower already checks for AutoRemove, but since this is just the docker-compose process manually stopping and removing the containers it's a special case. Ignoring that the container was not found when trying to remove it should be fine in general though.

[piksel]

This change seems to be the most sane course to take regardless of whether it's caused by the issue this PR attempts to fix.
I added a test for the "normal" case (container still existing after being stopped), as well as one for when the container could not be found.

[nothub]

Awesome :) Thanks a lot for your work on this project!

[piksel]

@all-contributors add @nothub for code

[allcontributors]

@piksel

I've put up a pull request to add @nothub! 🎉