[09:17:13] <elukey>	 really nice https://journal.arrikto.com/gpu-virtualization-in-k8s-challenges-and-state-of-the-art-a1cafbcdd12b
[11:22:45] <elukey>	 finished  the deployment of the ml-staging-codfw layers (knative, kserve..)
[11:22:50] <elukey>	 all works :) (at least so far)
[11:23:18] <jayme>	 wow! 🎉 congrats
[11:24:35] <elukey>	 well congrats to you for your work on 1.23 :)
[11:31:00] <jayme>	 don't jinx it :D
[17:16:44] <jhathaway>	 anyone care to weigh in on this commit to skip library charts, before I commit it? https://gerrit.wikimedia.org/r/c/operations/deployment-charts/+/888275
[17:17:43] <akosiaris>	 what's a chart library ? 
[17:18:00] * akosiaris hasn't followed last weeks developments, if any
[17:18:32] <jhathaway>	 help templates that are not rendered, but included by other charts, https://helm.sh/docs/topics/library_charts/
[17:18:42] <jhathaway>	 the jaeger upstream chart uses one
[17:18:57] <jhathaway>	 which broke our CI
[17:19:52] <akosiaris>	 ah, that helm 3 new thing. First time we use it I guess
[17:20:04] <akosiaris>	 yeah, makes sense to ignore them, let me +1
[17:20:41] <jhathaway>	 thanks
[23:07:18] <jhathaway>	 The jaeger chart uses Ingress & CronJob resources which are not available in k8s 1.16. However, our CI runs kubeconform over all rendered templates with k8s 1.16 and 1.23. It was not obvious to me how to exclude testing the jaeger chart with 1.16. I dug into the CI rake scripts a bit, but a way forward eluded me, anyone have any thoughts?