[08:19:09] swift rebalance still going AFAICS [09:34:03] mmhh yeah a bit suspicious dispersion hasn't increased in a while, I'll take a look [09:41:37] ok one known problem is that swift uid/gid are not fixed in general across the fleet, so reimage isn't fully unattended [09:41:55] i.e. the filesystems had an old uid for swift [09:45:52] the right thing to do is this I think https://phabricator.wikimedia.org/T123918#7318655 [10:09:12] Oh, I didn't know that :-/ [10:11:12] thanks for fixing :) [10:11:35] yeah as you can see from the task it's been a long standing issue, but the detection/fixing of uid should be automated so at least reimages of existing hosts are unattended [10:11:39] for sure, no problem [13:52:33] Am I right that gerrit requires single-commit changes? It doesn't have facility for "this change makes more sense as a series of commits"? [13:53:54] Emperor: you can chain change-requests, each of which is a single commit [13:54:00] you can give it a stack of commits, it will make each a separate change and each can be reviewed/merged separately [13:54:07] majavah said it better [13:54:39] OK gitlab's MR UI is nicer for this [13:55:21] it's hard to imagine a UI which would be worse. ;) [13:56:19] porting workflows powered by gerrit relation chains was a very long convo somewhere on a gitlab related mw.o talk page, let me try to find that :P [13:57:13] https://www.mediawiki.org/wiki/Topic:Vpbonspnaaafjdwq