[01:12:15] * AaronSchulz can't access https://wikitech.wikimedia.org/wiki/Special:Preferences, seems to be an es* cluster timeout due to fetching some messages from MessageCache . I was trying to look at my ssh key. [01:14:09] es1021/es1022 [05:25:01] !log tools.masto-collab Updated from 0631e5f to cfa007b [05:25:04] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools.masto-collab/SAL [07:15:29] !status wikitech issues T339079 [07:15:29] T339079: wikitech intermittent DB errors - https://phabricator.wikimedia.org/T339079 [08:10:06] !status Ok [08:18:29] taavi: what is that magic? [08:19:08] dcaro: it's a wmopbot feature! [08:19:24] awesome \o/! [09:46:15] #wikipedia-it-ot [09:47:36] ops sorry [18:11:17] Hey, I'm looking at improving video2commons workers. One idea is to move them from CloudVPS into toolforge. Is this still a bad idea due to NFS speeds? [18:52:47] in The Future there will be a way to build containers without NFS [18:58:10] @chicocvenancio: I guess the answer depends a lot on what your NFS usage looks like. I'm not immediately sure how Toolforge would be a better runtime for v2c conversion and upload nodes than Cloud VPS, but I'm not deeply familiar with the current implementation. What would you hope to gain from the migration? [20:53:43] I'm looking at something to orchestrate deployments, health checks, and a concentrated way to check logs. (re @wmtelegram_bot: @chicocvenancio: I guess the answer depends a lot on what your NFS usage looks like. I'm not immediately sure how Toolfo...) [20:54:03] In theory k8s should be enough for all of it. [21:09:41] @chicocvenancio: *nod* We do not have nice abstractions for any of those things in the Toolforge Kubernetes cluster today. That makes me think you might be better served with an isolated k8s cluster. There is a very much alpha/pre-alpha thing described at https://wikitech.wikimedia.org/wiki/Portal:Cloud_VPS/Admin/Magnum that I think you could convince us to let you try. [21:11:17] magnum's currently unusable for anyone without openstack-level admin access unfortunately [21:14:58] bd808: Do you know if profile::acme_chief::cloud::designate_sync_password (as set on beta cluster puppetmaster, to communicate with some part of openstack designate it looks like?) is important to keep semi private? I.e. is it abusable from outside WMCS in some way? If so, I'd move it along with various other semi secrets to Horizon for beta cluster (which syncs to public git), but I'm keeping this as-is until I know more about it. [21:15:22] Of course even now it's semi public given how liberally we hand out beta cluster access, but double checking in case it's something we'd like to keep there [21:21:21] Krinkle: it is the password for a Developer account that can modify OpenStack configuration for the deployment-prep project, so yes it should not be made public. [21:22:01] Alright, I'll add a scary warning to it for future ref :) [21:24:50] taavi: oh. I guess I knew that it would require use of the API as we are not ready to turn on a Horizon UI for it, but I did not understand that to also mean that you needed admin level credentials. :/ [21:25:58] Chico had tools admin before, but I don't think he was a cloud admin too at that time. [21:49:41] I mean yeah, Magnum would certainly solve it. Toolforge k8s seems better than puppet or the 3 nodes VPS nodes there are currently. (re @wmtelegram_bot: @chicocvenancio: *nod* We do not have nice abstractions for any of those things in the Toolforge Kubernetes cluster toda...) [21:53:40] (re cloud admin) 5 years is a bit long for my memory to be accurate, but that sounds right [21:55:15] @chicocvenancio: ok. NFS throughput may or may not be an issue. I guess you'd need to figure out a way to check that based on some understanding of the IOPS needs of v2c. You would certainly want a much larger k8s quota for CPU, RAM, and pods than default, but we have ways to accommodate that. [21:58:04] @chicocvenancio: I do think you are close to the ideal alpha tester for Magnum too. So if that sounds like a way you would be interested in going it might be worth a larger discussion in WMCS about where we are with that project. I do know we aren't really close to ready to make it general availability. [22:07:24] ack. For today I've created one worker(out of 7) in v2c toolforge tool, I want to see how it behaves with a few tasks and then we can discuss the next steps. [22:15:39] !log tools.toolinfo-scraper Updated to d041d0f [22:15:41] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools.toolinfo-scraper/SAL