[09:23:48] <elukey>	 https://github.com/knative/serving/blob/main/docs/scaling/SYSTEM.md is very nice
[09:24:06] <elukey>	 so the queue-proxy sidecar can buffer/limit concurrency to revisions, that is nice to know
[09:24:42] <elukey>	 we currently don't collect metrics about it, I'll need to add something to the kserve-inference chart
[10:31:37] <elukey>	 first version of the knative ingress policies!
[10:31:38] <elukey>	 https://gerrit.wikimedia.org/r/c/operations/deployment-charts/+/737875 
[16:15:14] <elukey>	 I found the problem with the network policy, but I am still not sure why :D
[17:19:44] <accraze>	 o/
[17:20:08] <accraze>	 elukey: sounds like you are having fun :)
[17:30:05] <elukey>	 accraze: o/ I think I am still not getting some nuances of network policies
[17:38:08] <accraze>	 for knative?
[17:58:27] <elukey>	 exactly yes
[17:58:43] <elukey>	 also its behavior varies a lot between versions
[17:59:00] <elukey>	 so setting exactly how traffic needs to be filtered is tricky
[22:47:26] <accraze>	 cool! got minikube running on the new ml-sandbox -- now trying to use our wmf images to run the kserve stack
[22:51:04] <accraze>	 ah shoot i forgot we're running k8s v1.16 on ml-serve, the new sandbox should probably match that for now
[23:31:32] <accraze>	 hah modifying the kserve quickinstall script to use our wmf images is more involved that i thought :D