Draft: Retry latest MR, if gets immediatelly reassigned
Challenge
Assume 1hr, 250 jobs pipeline, and one job it's a bit slower or fails
due to infrastructure issues (cannot be workarounded by GitLab retry:
).
When the person who assigned Marge watches this, it's mentally depriving.
Why not wait for other MRs to merge and retry? Because the pipeline
cannot be reused anymore, but when it's without changes to the main
repository, the pipeline CAN be reused.
Current Marge-bot approach
Currently, Marge-bot just takes MR one-by-one in order as they are assigned, so it is impossible to re-run the same pipeline if it's not only one in the queue (= empty queue before the MR failing).
Solution
What could be done, that as Marge-bot reports failure, the author has 1 minute (TBH, author or CI admins has often MR open) to reassign the MR and Marge-bot prioritizes it so the pipeline can be reused.
TODO
- not pseudocode
- make the feature optional
- optional definable 1 minute delay after failing MR