You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
'Provision' CRD support custom device plugin resource which can be ignored when matching ec2 instances
Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?
in my case, pod need mount /dev/fuse. Now we build a device plugin for /dev/fuse, and register example/fuse resource to kubelet, and already deploy a daemonSet for this resource. this case is same like “nvidia.com/gpu” . But karpenter can't identify "example/fuse", and won't privisioning new k8s node.
if karpenter can ignore this custom resource ,and still provisioning new node , the "example/fuse" daemonSet will depoly soon, and pod can be scheduled immediately.
so, i think if karpenter Provisioner CRD can support custom device plugin resource, and ignore this resource when trying to match available ec2 instances.
Are you currently working around this issue?
Now i have to create a leading pod without this "example/fuse" resource. And force karpenter to provisioning new node. Then delete this leading pod when it's running, this means i already got enough resource for real pod. And the real pod which need "example/fuse" can be scheduled soon. this solution seems some kind of weird .
So If karpetner native support custom device plugin resource woluld be best, users just define custom Provisioner for his needed resource.
Additional Context
No response
Attachments
No response
Community Note
Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
If you are interested in working on this issue or have submitted a pull request, please leave a comment
The text was updated successfully, but these errors were encountered:
Tell us about your request
'Provision' CRD support custom device plugin resource which can be ignored when matching ec2 instances
Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?
in my case, pod need mount /dev/fuse. Now we build a device plugin for /dev/fuse, and register
example/fuse
resource to kubelet, and already deploy a daemonSet for this resource. this case is same like “nvidia.com/gpu” . But karpenter can't identify "example/fuse", and won't privisioning new k8s node.if karpenter can ignore this custom resource ,and still provisioning new node , the "example/fuse" daemonSet will depoly soon, and pod can be scheduled immediately.
so, i think if karpenter
Provisioner
CRD can support custom device plugin resource, and ignore this resource when trying to match available ec2 instances.Are you currently working around this issue?
Now i have to create a leading pod without this "example/fuse" resource. And force karpenter to provisioning new node. Then delete this leading pod when it's running, this means i already got enough resource for real pod. And the real pod which need "example/fuse" can be scheduled soon. this solution seems some kind of weird .
So If karpetner native support custom device plugin resource woluld be best, users just define custom Provisioner for his needed resource.
Additional Context
No response
Attachments
No response
Community Note
The text was updated successfully, but these errors were encountered: