Marge times out waiting for the MR to merge instead of detecting CI failure.

[alatiera]

Marge times out waiting for the MR to merge instead of detecting CI failure.

[jcpetruzza]

At that point Marge is polling GitLab, waiting for the merge-request state to become "merged" or something known. The timeout is currently hard-coded to five minutes. I don't expect your GitLab instance to take 5' to process the merge, or does it? Otherwise, do you know in which state the merge-request ended afterwards? Was it still shown as "open"?

[alatiera]

I think here is what happened

• We opened a new MR with the CI intentionally failling
• While the CI was running we assigned Marge to the MR
• Marge set's it to merge when succeed
• CI fails as expected, but Marge just timeouts
• MR stays open with auto-merge on succeed enabled

We are using the docker image v0.6.1 I think

[nanonyme]

This actually also has its good sides. On a multi-axis project you might have a transient issue that disappears with retry. Since Marge doesn't immediately mark the build as failed, you have a chance to kick retry. That can sometimes save hours worth of rebuilding.