[08:52:47] <wm-bb>	 <Poslovitch> Hey there, is there a way to install apt packages in a kube pod (run with `webservice —backend=kubernetes node16`) ? I am wishing to use djvulibre-bin.
[08:52:54] <wm-bb>	 <Poslovitch> (on Toolforge)
[09:40:58] <vgutierrez>	 as a heads up, I'm merging https://gerrit.wikimedia.org/r/c/operations/puppet/+/888652, it could impact acme-chief instances deployed in several WMCS projects
[09:53:25] <arturo>	 vgutierrez: ACK
[09:58:26] <vgutierrez>	 looking good in deployment-prep and traffic WMCS environments
[10:18:58] <taavi>	 vgutierrez: https://gerrit.wikimedia.org/r/c/operations/puppet/+/893410/
[10:19:43] <vgutierrez>	 taavi: nice catch
[10:22:11] <taavi>	 can you merge? I don't have access on that repo
[10:26:30] <vgutierrez>	 taavi: my fault, I keep forgetting
[10:27:13] <vgutierrez>	 (done)
[10:51:50] <taavi>	 vgutierrez: I've updated all usees in WMCS to set passive as an array
[10:55:58] <vgutierrez>	 taavi: awesome
[15:46:09] <wm-bb>	 <bd808> Not today, no. There is on going work towards support for custom containers, but things are not quite ready for even alpha testers yet. (re @Poslovitch: Hey there, is there a way to install apt packages in a kube pod (run with webservice —backend=kubernetes node16) ? I am wishing ...)
[16:07:59] <TheresNoTime>	 Has there been any talk on adding (v)GPU flavours to WMCS? (I assume currently we don't have any hypervisors with GPUs?)
[16:09:51] <TheresNoTime>	 I'm sure I saw *something* about the machine learning team having/getting GPU hosts, be interesting to have that sort of capability available within WMCS..
[16:14:44] <arturo>	 TheresNoTime: yes, we don't have GPUs (as of today)
[16:19:12] <TheresNoTime>	 just chuck a RTX 4090 in one, sorted /s
[16:20:35] <Lucas_WMDE>	 TheresNoTime: you trying to mine crypto on cloud VPS? :P
[16:20:53] <TheresNoTime>	 :D
[16:21:46] <TheresNoTime>	 my pipeline ideas all revolve around "LLM detection models"
[16:21:54] <Lucas_WMDE>	 ahh, I see
[16:22:17] <Lucas_WMDE>	 (I was bracing myself for an :AngeryCat: response tbh ^^)
[16:22:46] <TheresNoTime>	 a generalised LLM detection model just wouldn't be feasible, but a model for detecting LLM use in the context of Wikipedia articles... maybe?
[16:22:57] <TheresNoTime>	 Lucas_WMDE: not on IRC :p
[16:38:03] <arturo>	 TheresNoTime: perhaps, if you can make a case for it in a phab ticket, we could inform our budget next year :-P
[16:41:08] <TheresNoTime>	 hmmmmm :P
[16:41:20] <bd808>	 TheresNoTime: the biggest challenge with GPUs is finding ones that are not 100% non-free software. The next biggest is price.
[16:41:35] <bd808>	 https://wikitech.wikimedia.org/wiki/Analytics/Cluster/AMD_GPU
[16:41:58] <bd808>	 ^ that's what I know has been experimented with in the past
[16:42:23] * TheresNoTime looks..
[16:42:34] <TheresNoTime>	 and hadn't thought about the non-free bit..
[16:43:20] <bd808>	 having Open principals makes it even harder to get into cryptocurrency and LLMs ;)
[17:09:01] <wm-bb>	 <Poslovitch> Alright. Maybe I can circumvent that by switching to a Cloud VPS-hosted tool? (re @bd808: Not today, no. There is on going work towards support for custom containers, but things are not quite ready for even alpha teste...)
[17:09:59] <bd808>	 @Poslovitch: yes, Cloud VPS project admins have full root access to their instances.
[18:33:24] <wm-bot>	 !log tools.stewardbots <maurelio> Update stewardbots to 2dca7cd
[18:33:28] <stashbot>	 Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools.stewardbots/SAL
[18:38:39] <wm-bot>	 !log tools.stewardbots-legacy <tools.stewardbots> Update stewardbots-legacy to 2dca7cd
[18:38:42] <stashbot>	 Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools.stewardbots-legacy/SAL
[23:00:19] <wm-bb>	 <Poslovitch> @bd808 Alright, thank you. I'll investigate that then. :)