kube-ops-view
kube-ops-view copied to clipboard
hjacobs
Separate nodes by availability zone
It would be nice to have nodes separated into failure zones so I can clearly see where each nodes/pods live
AZ-s are nice, but easier would be split that per regions.
OpenShift doesn't use master=true label (I've added it manually right now to test isMaster that actually works). OpenShift uses region label. Maybe the best idea is to make it configurable by which label it should group that?
There are also regions in pure Kube: failure-domain.beta.kubernetes.io/region
@morsik honestly "region" does not sound so useful as multi-region clusters should be the exception and not the default (because of latencies). I don't know many people running multi-region Kubernetes (and the cloud providers also don't have this as a managed service AFAIK).
@hjacobs: I don't know about basic Kube, but region in OpenShift doesn't necessary means long distance. In OO, there is region=infra for infrastructure nodes, and region=default for default nodes. This is very useful to group by in Ops View GUI.
@morsik ok, this was a misunderstanding. You mentioned failure-domain.beta.kubernetes.io/region which is the AWS region when running on AWS, so I assumed it's similar for OpenShift.
@hjacobs: oh right. This is my bad with example.
failure-domain.beta.kubernetes.io/region is indeed used by Cloud Providers in Kubernetes. I forgot about it - you're right.
But making it fully configurable ("By which label GUI should group nodes?") would be useful and address all of those problems. For me grouping by region is ok, for @puco (requestor) something about availability zone is ok - so user can choose.