[10:37:36] FYI: mr1-codfw is currently unreachable hence we can't currently access the mgmt network in codfw. I would suggest to refrain from performing reimages there or any other operation that might require access to the mgmt consoles. [cross-posting] [11:04:50] dunno if this already got shared, but it's nice ! https://people.wikimedia.org/~nshahquinn-wmf/new-data-center-impact/ [14:13:27] will grafana become unavailable for a while? [14:13:51] or not really, if everything goes well [14:30:15] router rebooted? [14:30:21] yes [14:30:24] see dcops [14:30:33] thanks, I wasn't on that channel [16:42:03] cwhite: (git blame) I'm looking at T346053, and as such, I'm trying to evaluate the impact of a temporary downtime of an apifeatureusage host. It seems that the only thing we run on these boxes is a logstash process writing to kafka. [16:42:04] T346053: Migrate apifeatureusage hosts to Bullseye or later - https://phabricator.wikimedia.org/T346053 [16:43:14] if we can't tolerate downtime, I'll probably spin new hosts running bullseye or bookworm, and replace the other hosts. [16:44:10] brouberol: you're correct. the stream is also backed by kafka. a temporary downtime should be tolerable :) [16:47:51] nice! So if I take the host down for 40 minutes (usually the time it takes to run a full reimage), no-one will send pikes and torches my way? [16:59:48] pikes and torches seem unlikely as the feature is used, but IIRC not frequently. [17:26:26] thanks, noted!