[06:44:48] _joe_: atm I have tcpdump with -W and -C, plus ifstat in tail (grepping for high tx bandwidth usage). When I see high values I'll just check the pcap, in this way I don't cause cpu usage for mc1027 :( [07:08:40] <_joe_> elukey: my point is - in the moment of congestion, it's ok to inspect with memkeys [07:16:19] _joe_ sure but usually it last few seconds, it is difficult to get it if not sustained [07:17:05] <_joe_> well as long as it's a few seconds and not sustained, it's less of an emergency and that should be fixed by better telemetry from MediaWiki imho [07:27:30] I am not saying it is an emergency, just trying to reduce the wtp* servers tko, I agree that we should have a better telemetry [07:27:50] but things are not really moving on this front :) [09:24:12] so the key that causes some problems should be [09:24:13] WANCache:v:nlwiki:preprocess-hash:6b35ce98ba6b117cf3eeb7807471739e:1 [09:24:26] in slab 154, mc1027 (~200K) [09:37:36] created https://phabricator.wikimedia.org/T248962 [15:01:04] jbond42 (and everyone), when you add a new required param to a profile, please add a default to cloud.yaml. e.g. https://gerrit.wikimedia.org/r/#/c/operations/puppet/+/584951/ [15:06:01] andrewbogott: sorry, i have created https://phabricator.wikimedia.org/T248994 to see if we can check for this in CI [15:06:18] that would be great! [15:16:16] jbond42: while we're at it… can you have a go at getting puppet to run on jbond-stretch.puppet.eqiad.wmflabs? I'm trying to get catalogs to compile everywhere so that the puppet/facter version upgrades can take [15:17:53] andrewbogott: one sec i think i can just delete that [15:18:00] better yet! [15:19:23] andrewbogott: have deleted all my instances in the puppet project [15:19:35] great, thanks! [15:46:00] o/ [15:47:55] 👋 [16:00:44] Is there an easy way/established pattern for giving access to grafana and nothing/very little else for a NDAed user? (A contractor on CPT in this case) [16:02:24] hnowlan: usually the NDA group, see https://wikitech.wikimedia.org/wiki/LDAP/Groups [16:02:35] AFAIK there isn't a smaller one at this time [16:03:34] hnowlan: I assume write access to grafana, as RO is for everyone [16:04:20] Just read access in this case [16:04:45] then just a browser :-P [16:05:36] They don't have an account at all right now, heh. [16:05:42] But that sounds suitable [16:05:44] not needed [16:06:51] oh, sorry, logstash is needed too [16:07:01] then NDA [16:07:17] cool, thanks! [16:07:33] follow the instructions at the top of the page