Auto-merge retry mechanism on errors #5967
Description
Community Note
- Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request. Searching for pre-existing feature requests helps us consolidate datapoints for identical requirements into a single place, thank you!
- Please do not leave "+1" or other comments that do not add relevant new information or questions, they generate extra noise for issue followers and do not help prioritize the request.
- If you are interested in working on this issue or have submitted a pull request, please leave a comment.
- I'd be willing to implement this feature (contributing guide)
Describe the user story
GitHub's API isn't always perfect in terms of when a merge can happen. It's not uncommon for us to run into HTTP 403 Resource not accessible by integration errors after an atlantis apply. It's relatively rare, but you still notice this at scale in an organisation.
Describe the solution you'd like
When automerging is used on atlantis apply commands, have a mechanism to retry (e.g. up to 3 times) on some common/configurable errors. This can be opt-in.
Describe the drawbacks of your solution
- Not all errors are transient: sometimes, an error like 403 really isn't transient, and it truly cannot be merged. These cases can result in N auto-merge retried requests.
- Not everyone wants retries on their Atlantis deployments.
Describe alternatives you've considered
We can also implement an automatic GitHub Action workflow that runs on PR comments from Atlantis when it comments with an error like this. This is definitely useful, but it isn't all too generic, and is a patch to a case that can often be solved by just trying once more after some 5 seconds.