-
Notifications
You must be signed in to change notification settings - Fork 855
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Karpenter takes time to provision a node #2250
Comments
In this case, Karpenter thought that the pod would schedule against a node already in your cluster. kube-scheduler thought that it wouldn't schedule due to one of these reasons:
In newer versions of Karpenter, you can run a Can you provide the pod spec of the pod that wouldn't schedule? |
@tzneal We notice that scheduling logic has been changed on Karpenter in v0.13.2 here https://karpenter.sh/v0.13.2/concepts/#scheduling Can this newer version solve this issue where |
It's possible, and if it does re-occur you'll be able to describe the pod to determine which node we thought it would schedule to. |
what does it happen when Karpenter thinks that a pod should be scheduled against existing capacity, while the kube-scheduler marked it unschedulable? What schedules the pod in the end? |
@cebernardi If that occurs, it's a bug and you can report it. We rely on kube-scheduler to perform the scheduling. |
Labeled for closure due to inactivity in 10 days. |
Any solution? |
Version
Karpenter: v0.10.1
Kubernetes: v1.21
Expected Behavior
Karpenter should launch a new worker node in case of unscheduled pods when there is no capacity on the cluster
Actual Behavior
Sometime Karpenter takes 10-30 minutes to provision a new node, till then the pods remain in unscheduled state.
Steps to Reproduce the Problem
This issue is intermittent. We are also not able to reproduce it explicitly.
Resource Specs and Logs
Application logs -
Karpenter Controller logs at the same time -
Note - There was no affinity/antiaffinity or nodeSelector on the pods which Karpenter shouldn't have satisfied.
The text was updated successfully, but these errors were encountered: