[10:40:06] I'm about to stop the circular replication in prod sections [10:41:00] ack, lmk if I can be of any assistance [10:46:18] volans: dry run confirmed it but to triple check, is the dcs correct here? sudo cookbook sre.switchdc.databases.finalize -t T374972 eqiad codfw [10:46:19] T374972: Output test logs of production testing of the pre switchover tasks related to databases - https://phabricator.wikimedia.org/T374972 [10:47:01] yes, it's the "finalize" of the move from eqiad to codfw. As j.aime already pointed out it can be confusing [10:47:23] so we can rename the cookbook or better clarify the options, I tried to expand a bit the help message IIRC [10:48:03] I think it should figure it out automatically [10:48:31] if we trust it so much sure :) [10:48:40] (no "wide APIs" principle) [10:48:48] it can ask "Am I doing it right?" :D [10:48:55] s1 done [10:49:06] it's also an easy change as it's easy to get the current primary in multiple ways :) [10:49:30] yay [10:49:37] I'm tailing the logs [10:49:44] s2 fine too [10:54:27] about to do x1, it's RBR and caused a mess during prep [10:54:57] I wonder if this causes a mess too [10:55:08] nah, it wasn't a mess, just it failed to succeed [10:55:18] the clean up was fun :D [10:55:43] so far it looks okay [10:55:52] we do delete where here so should be fine no? [10:56:27] yeah [10:57:25] they look fine [10:57:29] everything done now [10:57:35] let's go crazy [10:57:53] lol [13:09:39] Amir1: I'll be testing dbctl support in spicerack after the meetings. Is there a preferred DB host I can use for it? Apart from dry-run tests I'd also like to change some things like its weight and pooled % (even by just 1 digit) at least. The more I can modify it the better :) [13:10:18] I don't think we have test sections in dbctl [13:10:22] but you can go with s5 or s6 [13:10:30] they are low risk sections [13:10:32] eqiad may be safer (?) [13:10:34] a replica in eqiad with no special role? [13:10:53] any of s5/s6 in eqiad should be fine, yeah [13:11:33] (don't go with sanitarium masters: https://orchestrator.wikimedia.org/web/cluster/alias/s5) or the ones in vslow/dump groups [13:11:38] you should be fine [13:13:12] got it, db1185 seems a possible one