[09:43:18] just noticed, on eqiad only: [09:43:29] Last snapshot for s5 at eqiad (db1150.eqiad.wmnet:3315) taken on 2022-02-14 00:08:58 is 683 GB, but previous one was 647 GB, a change of 5.7% [09:44:15] there's a big schema change on cebwiki.templatelinks, so probably it was that [09:44:16] I can reasearch further if you belive it is worth it [09:44:20] ok [09:44:22] nah, it is that for sure [09:44:36] the table is around 180GB, so it probably got a few GB back after the re-creation [11:19:46] Amir1: I know it is not canonical data, but in the past we had meetings to back it up, or prepare a way to recover it faster than rebuild all [11:20:55] jynus: yeah but this is on small set of categories only [11:21:13] if it was even 10% of categories, I'd say let's recover [11:21:14] sorry, talking about cassandra, not category count/links [11:21:21] aah ok [11:21:26] cassandra/restbase [11:22:07] I think (and correct me if I'm wrong), it has several usages, some are important and those were moved out of restbase [11:22:33] the current one is basically caching (again, the details are blurry) [11:22:42] yeah, but I also belive it is the backbone of mobile frontend [11:23:21] so I was told to work with the owners at some point a recovery scenario (not necesarilly backups) [11:23:58] nothing concrete happened after that [11:24:23] but I also want to be on top of new stuff happening so we tell them to have backups into account [11:27:35] I suggest talking to Luke [11:27:57] thanks [12:41:46] I'm looking for a DBA to deploy the four lines in https://gerrit.wikimedia.org/r/c/operations/puppet/+/759357 . Followup https://gerrit.wikimedia.org/r/c/operations/puppet/+/759359 is already deployed. [12:42:24] andre: I will take care of that [12:42:30] thank you [12:45:07] andre: all done [12:49:10] marostegui, thanks a lot! [14:38:03] * Emperor sends swift upstream https://review.opendev.org/c/openstack/swift/+/829053 [14:38:48] I suspect they may hate it a bit, but it does at least mean (barring a couple of bizarro edge-cases that you can't fix without a not-python wrapper) that error conditions will result in exit status of 2 as documented [14:39:43] (it's suprisingly hard writing about unusual cases in exception handling without using "the exception is" or equivalent) [14:40:12] :) [14:42:31] * Emperor had to resist "The exception is the SystemExit exception, which exceptionally we ..."