Wei comments

Results 149 comments of

Wei

Mutating web hook for identifying arch / platform supported by container images of pod

I am willing to implement a webhook to handle everything like this.

node selection: One supper large node with many small size nodes

Can you help take a look at this? @jonathan-innis @ellistarn

node selection: One supper large node with many small size nodes

The problem is here: https://github.com/kubernetes-sigs/karpenter/blob/c4be45e04079f0d9313150b3ae7b5313132b0e36/pkg/controllers/provisioning/scheduling/queue.go#L76

node selection: One supper large node with many small size nodes

I can help fix this issue if you think it is a real problem.

node selection: One supper large node with many small size nodes

> This is a pretty major change. Would you be interested in writing an RFC? https://github.com/kubernetes-sigs/karpenter/tree/main/designs https://karpenter.sh/docs/contributing/design-guide/ Sure, I will do it.

fix: fix image digest parsing

cc @njtran

fix: do not reschedules the workload to the same unhealthy cluster when application failover enabled

/cc @XiShanYongYe-Chang @chaunceyjiang @chaosi-zju

fix: do not reschedules the workload to the same unhealthy cluster when application failover enabled

/cc @RainbowMango I don't see the purpose of `Immediately`. Should we consider deprecating it?

fix: fix occasional e2e failure

By the way, should we re-enqueue the item if an unexpected error happens, like a broken Internet connection? For example, the following code: https://github.com/kubernetes-sigs/karpenter/blob/14f12dd82b6c8dfc9bff634df0527ae718faa748/pkg/controllers/nodeclaim/lifecycle/registration.go#L59 I prefer to re-enqueue them, make...

fix: when all replicas of a deployment are on one node, restart the deployment instead of evicting it

I like this idea