Fix/optimize Resolve the performance overhead of calling large clusters by AutuSnow · Pull Request #54 · kubernetes-sigs/dra-driver-cpu

AutuSnow · 2026-02-05T14:06:31Z

Resolve the performance overhead of calling large clusters

Note: RemoveResourceClaimAllocate() requires rebuilding the cache during deletion, with a complexity of O (n). This is because cpuset does not support efficient differencing operations (one claim may have allocated CPUs that overlap with other claims). However, due to relatively few deletion operations, and the use of Get Shared CPUs ()
Very frequent calls (called every time a container is created/deleted)

Fixes:#53

k8s-ci-robot · 2026-02-05T14:06:41Z

Welcome @AutuSnow!

It looks like this is your first PR to kubernetes-sigs/dra-driver-cpu 🎉. Please refer to our pull request process documentation to help your PR have a smooth ride to approval.

You will be prompted by a bot to use commands during the review process. Do not be afraid to follow the prompts! It is okay to experiment. Here is the bot commands documentation.

You can also check if kubernetes-sigs/dra-driver-cpu has its own contribution guidelines.

You may want to refer to our testing guide if you run into trouble with your tests not passing.

If you are having difficulty getting your pull request seen, please follow the recommended escalation practices. Also, for tips and tricks in the contribution process you may want to read the Kubernetes contributor cheat sheet. We want to make sure your contribution gets all the attention it needs!

Thank you, and welcome to Kubernetes. 😃

AutuSnow · 2026-02-05T14:07:34Z

/assign @ffromani

ffromani

thanks for the fix, this was indeed one of our TODOs.
Please verify this change is covered by tests, and feel free to add more coverage if not.

I wonder, in general, if and how we can assess the performance gains/changes of PRs.

ffromani · 2026-02-05T14:10:38Z

-	configs map[types.UID]PodCPUAssignments
+	mu                  sync.RWMutex
+	configs             map[types.UID]PodCPUAssignments
+	sharedCPUContainers map[types.UID]struct{}


this is used as set, so it should be used a set (k8s.io/apimachinery/pkg/util/sets)

I have completed the modifications and added benchmark tests

AutuSnow · 2026-02-09T13:58:55Z

@ffromani Can I take another look

ffromani

thanks for the PR and sorry it took me so long to review again. Except micro-benchmarks, I'm not sure about the actual gain when operating the plugins, but the code becomes IMO a bit nicer, so we should move forward and merge. Please rebase.

I'm inclined into accept into 0.1.0, pending final review

ffromani · 2026-02-25T06:26:10Z

+		s.allocatedCPUs = cpuset.New()
+		for _, cpus := range s.resourceClaimAllocations {
+			s.allocatedCPUs = s.allocatedCPUs.Union(cpus)
+		}


perhaps use s.allocatedCPUs = s.allocatedCPUs.Difference(cpus) like you do below in GetSharedCPUs?

ffromani · 2026-02-25T06:29:12Z

@@ -61,10 +64,23 @@ func (s *PodConfigStore) SetContainerState(podUID types.UID, state *ContainerSta
 	s.mu.Lock()
 	defer s.mu.Unlock()

+	// Check if there's an existing state for this container that we need to clean up from cache.


why can't we just override? what's the risk here?

why can't we just override? what's the risk here?

When a container restarts, SetContainerState is called again with the same containerName but a new containerUID. If we just override without cleanup, the old containerUID would remain as a stale entry in the sharedCPUContainers set, causing GetContainersWithSharedCPUs() to return UIDs for containers that no longer exist. The cleanup ensures we remove the old UID before inserting the new one.

thanks, fully makes sense. Please make sure we have a unit test which demonstrates this.

thanks, fully makes sense. Please make sure we have a unit test which demonstrates this.

I will do it this way

linux-foundation-easycla · 2026-02-25T07:37:43Z

The committers listed above are authorized under a signed CLA.

✅ login: AutuSnow / name: Snow (2655227)

ffromani

please rebase when you resubmit

ffromani · 2026-02-25T07:59:03Z

@@ -61,10 +64,23 @@ func (s *PodConfigStore) SetContainerState(podUID types.UID, state *ContainerSta
 	s.mu.Lock()
 	defer s.mu.Unlock()

+	// Check if there's an existing state for this container that we need to clean up from cache.


thanks, fully makes sense. Please make sure we have a unit test which demonstrates this.

ffromani · 2026-02-27T12:25:40Z

+// TODO: CoreID is only unique within a socket. On multi-socket systems where CoreIDs repeat
+// across sockets, this method may return CPUs from multiple sockets for the same CoreID.
+// Consider refactoring core-level methods to use CoreKey{SocketID, CoreID} for correct
+// disambiguation, similar to CPUTopology.CPUsByCore.


correct, but this in inherited from cpumanager code. We need to decide if and how drift from cpumanager. It's a post-0.1.0 decision

ffromani

/approve
/lgtm

this is a TODO we had. The fix is nice, contained and was in review since long time. I'm doing an exception and merging before 0.1.0.

k8s-ci-robot · 2026-02-27T12:27:03Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: AutuSnow, ffromani

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~OWNERS~~ [ffromani]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

k8s-ci-robot requested review from ffromani and pohly February 5, 2026 14:06

k8s-ci-robot added size/M Denotes a PR that changes 30-99 lines, ignoring generated files. cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. labels Feb 5, 2026

AutuSnow force-pushed the fix/optimize branch from a5dd6bd to f99b7a4 Compare February 5, 2026 14:07

k8s-ci-robot assigned ffromani Feb 5, 2026

ffromani reviewed Feb 5, 2026

View reviewed changes

AutuSnow force-pushed the fix/optimize branch from f99b7a4 to a283610 Compare February 5, 2026 14:50

k8s-ci-robot added size/L Denotes a PR that changes 100-499 lines, ignoring generated files. needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. and removed size/M Denotes a PR that changes 30-99 lines, ignoring generated files. labels Feb 5, 2026

ffromani reviewed Feb 25, 2026

View reviewed changes

AutuSnow force-pushed the fix/optimize branch from a283610 to ca351f7 Compare February 25, 2026 07:37

k8s-ci-robot added cncf-cla: no Indicates the PR's author has not signed the CNCF CLA. and removed needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. labels Feb 25, 2026

AutuSnow force-pushed the fix/optimize branch from ca351f7 to 01ae7a9 Compare February 25, 2026 07:40

k8s-ci-robot added cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. and removed cncf-cla: no Indicates the PR's author has not signed the CNCF CLA. labels Feb 25, 2026

ffromani reviewed Feb 25, 2026

View reviewed changes

Fix/optimize Resolve the performance overhead of calling large clusters

2655227

AutuSnow force-pushed the fix/optimize branch from 01ae7a9 to 2655227 Compare February 25, 2026 08:27

ffromani reviewed Feb 27, 2026

View reviewed changes

k8s-ci-robot added the lgtm "Looks good to me", indicates that a PR is ready to be merged. label Feb 27, 2026

k8s-ci-robot added the approved Indicates a PR has been approved by an approver from all required OWNERS files. label Feb 27, 2026

k8s-ci-robot merged commit c12dee2 into kubernetes-sigs:main Feb 27, 2026
2 checks passed

AutuSnow deleted the fix/optimize branch February 27, 2026 13:04

AutuSnow mentioned this pull request Mar 10, 2026

REQUEST: New membership for @AutuSnow kubernetes/org#6198

Closed

11 tasks

Conversation

AutuSnow commented Feb 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

k8s-ci-robot commented Feb 5, 2026

Uh oh!

AutuSnow commented Feb 5, 2026

Uh oh!

ffromani left a comment

Choose a reason for hiding this comment

Uh oh!

ffromani Feb 5, 2026

Choose a reason for hiding this comment

Uh oh!

AutuSnow Feb 5, 2026

Choose a reason for hiding this comment

Uh oh!

AutuSnow commented Feb 9, 2026

Uh oh!

ffromani left a comment

Choose a reason for hiding this comment

Uh oh!

ffromani Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

ffromani Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

AutuSnow Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

ffromani Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

AutuSnow Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

linux-foundation-easycla Bot commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ffromani left a comment

Choose a reason for hiding this comment

Uh oh!

ffromani Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ffromani Feb 27, 2026

Choose a reason for hiding this comment

Uh oh!

ffromani left a comment

Choose a reason for hiding this comment

Uh oh!

k8s-ci-robot commented Feb 27, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

AutuSnow commented Feb 5, 2026 •

edited

Loading

ffromani Feb 25, 2026 •

edited

Loading

linux-foundation-easycla Bot commented Feb 25, 2026 •

edited

Loading

ffromani Feb 25, 2026 •

edited

Loading