[WIP]Add separate plugin for DRA(Dynamic Resource Allocation). #3577
Conversation
volcano-sh-bot
added
the
needs-rebase
|
@Subhasish-Behera: PR needs rebase. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
| ListenAddress string | ||
| EnablePriorityClass bool | ||
| EnableCSIStorage bool | ||
| EnableDynamicResourcesAllocation bool |
There was a problem hiding this comment.
We should set a separate feature struct like https://github.com/volcano-sh/volcano/blob/33a9eb5b96cddc3c076a439ed134d41a0d095220/pkg/scheduler/plugins/predicates/predicates.go#L291，which is imported from k8s instead of putting it in ServerOption, cause it's a plugin switch.
| // The state is initialized in PreFilter phase. Because we save the pointer in | ||
| // framework.CycleState, in the later phases we don't need to call Write method | ||
| // to update the value | ||
| type stateData struct { |
There was a problem hiding this comment.
Wr should put all the codes copyed from k8s in a separate file and give a comment that these codes are copyed from k8s，once dra has been published as a public struct we should remove them.
| podSchedulingContextLister resourcev1alpha2listers.PodSchedulingContextLister | ||
| } | ||
|
|
||
| //func (pl *dynamicResources) OnSessionClose(ssn *framework.Session) { |
There was a problem hiding this comment.
All the commented codes should be removed: )
|
I think we should also add an e2e test to cover this: ) |
|
Please also sign off your commit with |
|
This PR may need to close, this PR will follow up to finish introducing DRA: #3799 |
This WIP PR is for DRA(Dynamic resource allocation) is the new way of requesting access to resources available in Kubernetes 1.26+.
Recent discussions in the Volcano community address the need to implement a way to DRA. It would extend
From an end user’s perspective, It moves away from the limited "countable" interface for requesting access to resources (e.g. "nvidia.com/gpu: 2"), providing an API more akin to that of persistent volumes.
Issue Reference: #3143
Design Details: #3487