[17:07:56] akosiaris OK if I merge your change re: wikikube ? [17:10:06] akosiaris looked at the changes and went ahead and merged [17:37:04] am I right in thinking scap is being phased out as we move over to kubernetes/helm? [17:37:46] dduvall: ^ [17:38:48] inflatador: yes, thanks. [17:39:45] zip: it's dropping some duties and gaining some others, but it's not being phased out [17:40:44] gotcha [17:40:56] just trying to wrap my head around how to deploy a non-mediawiki service [17:54:27] Give some input and someone on this channel should be able to help. Alternatively, phab task. [18:00:12] zip: I would suggest strongly that new non-MediaWiki things be deployed as containers via helm, but there may be reasons that is not possible. Scap3 (the non-MediaWiki part of scap) feels like a dying need at this point. [18:01:50] Even without k8s & helm andrewbogott and I have moved services like Striker and Horizon into containers as a scap3 replacement. I can't speak for andrewbogott, but for me that change made deployments much easier. [18:48:24] * andrewbogott agrees [19:02:10] yeah, i agree as well. scap3 still works for things that cannot be containerized, but we're not actively making improvements to it. in contrast, a lot of investment has gone into making things easier for folks targeting k8s (CI capacity for image builds, testing of built container images as services in gitlab, etc.) [20:38:04] sukhe you mentioned there was a cumin alias for finding hosts w/GPUs? Do you remember what it is? I'm not finding it ;( [20:41:57] inflatador: basically (I think): P:amd_gpu should cover everything that includes the GPU related stuff [20:43:14] sukhe thanks...looks like that one is a bit too broad (includes stat hosts that don't have GPUs for example)...will keep checking though [20:46:02] inflatador: oh interesting. how about: C:amd_rocm [20:46:29] # Deploy AMD's ROCm suite and GPU drivers. [20:47:19] do check with analytics once but I would expect this will include the base minimum (returns six hosts?) [20:47:51] you could run "lshw -C display | grep product" on * and get the different sets? [20:48:03] sukhe excellent, that looks much better [20:48:34] mutante yeah, I'll probably do that to double-check. Context is updating docs to be a bit more dynamic [20:48:56] * mutante nods [20:49:08] like instead of saying "xyz hosts have GPUs," say "here's the puppet code for GPUs" or "run this cumin query" [20:49:15] makes sense [20:54:02] suk-he's cumin query matches what I see when I look up GPU-specific prom metrics (`amd_rocm_gpu_usage_percent`) [21:08:11] ah, I found a couple of newer GPU hosts in Puppet... klaus-man is holding out on us ;P [21:26:59] akosiaris: sure thing! I just wanna get the lay of the land so nobody has to teach me stuff that's in the wiki