Allow Ci::Minutes::BatchResetService to continue with next batch in case of failues
Problem
From !29684 (comment 329847627):
-
Right now if one of the "sub-batch" (of 1000) fails the whole batch of 100,000 blows up. This is better than the previous version when the whole reset was interrupted but it's still bad.
-
We also print the first 10 IDs in the batch that failed. This is not really useful as it does not identify the ID range that failed
Proposal
-
I think we should change this approach to a "best effort" by logging an error in Sentry when a transaction is rolled back and continue with the next batch. This way we know exactly the impact of the failure.
-
In doing this we could also log the first and last IDs of the batch, rather than the first 10 IDs. This would allow us to know which batch we can manually retry.