Apologies if this duplicates an existing task
Currently toolforge:heritage is out of sync with our master branch. The difference is that our master branch uses pymysql whereas toolforge was rolled back 2017-11-17 as part of T180833: Monuments database dropped to 10% of its contents.
We should figure out which one to use and then make sure the two are in sync again.
Mentioned in SAL (#wikimedia-cloud) [2018-07-20T17:16:07Z] <Lokal_Profil> Manually resolving conflicts and creating a local commit. Resolving the difference to master is tracked at https://phabricator.wikimedia.org/T200101
Which is why we switched over. I believe the revert was caused by this causing some component to fail cascading into the whole harvest failing.
Yeah, so this was me panicking about the drop in the monuments DB. As I had changed both the library and to the replicas at the same time, it was unclear which one was the issue (or both). As outlined in T180833 I tried reverting only the replicas, which did not helped ; only reverting pymysql helped.
Now, either we move back to pymysql, or we revert in the master as well (which I should have done back then).
Back then, we were running pymysql=0.7.11, and it’s now at 0.9.1, so things might be better now.
Maybe bump pymysql and then keep an eye on it for a few days. If we do it now then there is plenty of time to revert before WLM starts.
Mentioned in SAL (#wikimedia-cloud) [2018-07-31T09:08:49Z] <Lokal_Profil> Manually resolving conflicts and creating a local commit due to T200101