Skip to content

Conversation

@toVersus
Copy link
Contributor

Fixes #2325

Description

This PR updates Karpenter to support Pod Level Resources. The Pod Level Resources feature is planned to graduate to beta in Kubernetes 1.34. Calculating pod resource requests and limits is complex, and to allow components like the scheduler and Node Autoscalers to share a common logic, the necessary functions have been published in k8s.io/component-helpers/resources. While Karpenter currently implements its own custom logic, this PR transitions it to use the shared library.

There is one difference between the shared library and the existing custom logic. The current custom logic includes behavior that merges limits into requests when only limits are specified. However, this merging is actually handled by the API server's defaulting logic. Some existing tests rely on this merging logic, and without the second commit, those tests fail. Personally, I think Karpenter doesn't need to implement this logic explicitly, but WDYT?

How was this change tested?

make presubmit

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

@k8s-ci-robot k8s-ci-robot added the cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. label Jul 18, 2025
@k8s-ci-robot
Copy link
Contributor

Welcome @toVersus!

It looks like this is your first PR to kubernetes-sigs/karpenter 🎉. 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/karpenter 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. 😃

@k8s-ci-robot k8s-ci-robot added the needs-ok-to-test Indicates a PR that requires an org member to verify it is safe to test. label Jul 18, 2025
@k8s-ci-robot
Copy link
Contributor

Hi @toVersus. Thanks for your PR.

I'm waiting for a kubernetes-sigs member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

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-sigs/prow repository.

@k8s-ci-robot k8s-ci-robot added the size/L Denotes a PR that changes 100-499 lines, ignoring generated files. label Jul 18, 2025
@coveralls
Copy link

coveralls commented Jul 18, 2025

Pull Request Test Coverage Report for Build 17568207446

Details

  • 3 of 3 (100.0%) changed or added relevant lines in 2 files are covered.
  • 10 unchanged lines in 3 files lost coverage.
  • Overall coverage decreased (-0.1%) to 81.572%

Files with Coverage Reduction New Missed Lines %
pkg/utils/resources/resources.go 1 91.09%
pkg/controllers/disruption/drift.go 2 87.76%
pkg/controllers/provisioning/scheduling/preferences.go 7 88.76%
Totals Coverage Status
Change from base Build 17559484392: -0.1%
Covered Lines: 10677
Relevant Lines: 13089

💛 - Coveralls

@k8s-ci-robot k8s-ci-robot added the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Aug 14, 2025
@toVersus toVersus force-pushed the feat/pod-level-resources branch from 1b939b2 to c3399c4 Compare August 28, 2025 00:23
@k8s-ci-robot k8s-ci-robot removed the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Aug 28, 2025
@toVersus toVersus force-pushed the feat/pod-level-resources branch from c3399c4 to cfcdce2 Compare August 28, 2025 00:30
@k8s-ci-robot k8s-ci-robot added the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Sep 9, 2025
@toVersus toVersus force-pushed the feat/pod-level-resources branch from cfcdce2 to 3fdaaa8 Compare September 9, 2025 00:21
@k8s-ci-robot k8s-ci-robot removed the needs-rebase Indicates a PR cannot be merged because it has merge conflicts with HEAD. label Sep 9, 2025
Replaced Karpenter’s custom Pod resources calculation logic
with k8s.io/component-helpers/resources shared package.

Signed-off-by: Tsubasa Nagasawa <[email protected]>
@toVersus toVersus force-pushed the feat/pod-level-resources branch from 3fdaaa8 to f9eb4d7 Compare September 9, 2025 00:44
@jmdeal
Copy link
Member

jmdeal commented Sep 17, 2025

Personally, I think Karpenter doesn't need to implement this logic explicitly, but WDYT?

So the issue is that the defaulting applies to the pods created for the daemonset, but not for the daemonset itself. Karpenter reads the daemonset spec to compute the "daemonset overhead" for a NodePool. We do this by creating fake daemonset pods based on the specs (ref) and then calling RequestsForPods against those pods (ref).

That being said, I think it's reasonable for this interface to expect that the pods provided adhere to the API server's defaulting logic. It should be the responsibility of any component generating "fake" pods to ensure that they meet those requirements. In this case, we would want to update our getDaemonsetPods function to handle that defaulting.

Other than that, this PR looks good to me and I'm happy to take on this shared dependency and drop our custom logic. Apologies that it took a while to get review, our bandwidth right now is really limited and we're trying to get more folks promoted to reviewer / approver. We are actually in a bit of a hurry to get this in for 1.34 - I was about to get a PR up but saw yours - so since the rest of the PR looks good I'm going to go ahead and approve then follow up with the change to getDaemonsetPods.

/lgtm
/approve

@k8s-ci-robot k8s-ci-robot added the lgtm "Looks good to me", indicates that a PR is ready to be merged. label Sep 17, 2025
@k8s-ci-robot
Copy link
Contributor

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: jmdeal, toVersus

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

The pull request process is described here

Needs approval from an approver in each of these files:

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 k8s-ci-robot added the approved Indicates a PR has been approved by an approver from all required OWNERS files. label Sep 17, 2025
@k8s-ci-robot k8s-ci-robot merged commit 06d9ded into kubernetes-sigs:main Sep 17, 2025
15 checks passed
@toVersus toVersus deleted the feat/pod-level-resources branch September 18, 2025 01:59
jigisha620 pushed a commit to jigisha620/karpenter that referenced this pull request Sep 19, 2025
harshad3339 added a commit to acquia/karpenter that referenced this pull request Nov 3, 2025
* chore: bump go version to 1.24.4 (kubernetes-sigs#2298)

* chore: Only log that the command succeeded when it actually did (kubernetes-sigs#2302)

* fix: Fix bug with MarkForDeletion before creating replacements (kubernetes-sigs#2300)

* perf: Refactor the eviction queue to be multithreaded (kubernetes-sigs#2252)

* docs: Add Bizfly Cloud provider (kubernetes-sigs#2303)

* chore: Bump lifecycle cache expiration to one hour (kubernetes-sigs#2307)

* chore: Use cluster state to check replacement NodeClaim existence (kubernetes-sigs#2308)

* chore(deps): bump github.com/samber/lo from 1.50.0 to 1.51.0 in the go-deps group (kubernetes-sigs#2315)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore: bump operatorpkg (kubernetes-sigs#2314)

* chore(deps): bump the k8s-go-deps group across 1 directory with 4 updates (kubernetes-sigs#2317)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore: Refactor Orchestration Queue and Handle Mark/Unmark Deletion in Queue (kubernetes-sigs#2305)

* chore(deps): bump the k8s-go-deps group with 7 updates (kubernetes-sigs#2326)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* perf: multithreaded orchestration queue (kubernetes-sigs#2293)

* test: Add nodeclaim name when you have garbage collection (kubernetes-sigs#2333)

* perf: Reduce multiple patch calls in instance termination (kubernetes-sigs#2324)

* fix: add helm rbac for kwok-provider to update finalizers (kubernetes-sigs#2336)

Signed-off-by: Max Cao <[email protected]>

* feat: configure CRD status operator with larger histogram buckets (kubernetes-sigs#2328)

* chore(deps): bump sigs.k8s.io/yaml from 1.4.0 to 1.5.0 in the k8s-go-deps group (kubernetes-sigs#2339)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore(deps): bump github.com/docker/docker from 28.2.2+incompatible to 28.3.0+incompatible in the go-deps group (kubernetes-sigs#2340)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* fix: Fix re-retrieving object on retry (kubernetes-sigs#2337)

* fix: Fix overriding error with patch call (kubernetes-sigs#2338)

* fix: add missing rlock to disruption queue (kubernetes-sigs#2348)

* test: allow e2e tests to output junit report (kubernetes-sigs#2334)

Signed-off-by: Max Cao <[email protected]>

* docs: Add Oracle Cloud Infrastructure (OCI) provider  (kubernetes-sigs#2342)

* fix: no longer allow the same hostname to take multiple capacity (kubernetes-sigs#2356)

* feat: support auto relaxing min values (kubernetes-sigs#2299)

* fix: update provider ID to ensure that Cloud Provider tests pass (kubernetes-sigs#2363)

* fix: remove unsupported capacity_type label from karpenter_nodeclaims… (kubernetes-sigs#2364)

* fix: update deletionTimestamp on terminating pods when after nodeDeletionTimestamp (kubernetes-sigs#2316)

Co-authored-by: Amanuel Engeda <[email protected]>

* chore: promote ReservedCapacity feature gate to beta (kubernetes-sigs#2365)

* fix: flakiness in expiration tests (kubernetes-sigs#2366)

* test: Bump the termination time for the deletion timestamp (kubernetes-sigs#2367)

* chore(deps): bump github.com/docker/docker from 28.3.0+incompatible to 28.3.1+incompatible in the go-deps group (kubernetes-sigs#2355)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* fix: pod errors when nodepool requirements filter all instance types (kubernetes-sigs#2341)

* refactor: Create a NopValidator for the disruption testing (kubernetes-sigs#2369)

* chore(deps): bump the go-deps group with 2 updates (kubernetes-sigs#2373)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* refactor: Update disruption testing from PR comments (kubernetes-sigs#2372)

* feat: (BREAKING) addition of launch timeout for nodeclaim lifecycle (kubernetes-sigs#2349)

* chore: Consider node.kubernetes.io/not-ready:NoExecute as ephemeral (kubernetes-sigs#2265)

* perf: Optimistically delete from the cache after launch (kubernetes-sigs#2380)

* docs: Node Overlay RFC (kubernetes-sigs#2166)

* fix: handle multiple PDBs for the same pod more gracefully (kubernetes-sigs#2379)

* docs: Add IBM Cloud provider (kubernetes-sigs#2396)

Signed-off-by: Josephine Pfeiffer <[email protected]>

* fix: rate limit eviction when PDBs are blocking (kubernetes-sigs#2399)

* feat: Add the Node Overlay CRD (kubernetes-sigs#2296)

* chore: ignore pods that use unsupported provisioner in the storageClass (kubernetes-sigs#2400)

* feat: Add a feature flag for Node Overlay (kubernetes-sigs#2404)

* feat: Add StaticCapacity feature flag (kubernetes-sigs#2405)

* fix(BREAKING): update naming of karpenter_pods_drained_total (kubernetes-sigs#2421)

* fix: pod metrics when pod is terminal (kubernetes-sigs#2417)

* chore: ignore pods that have unbound pvc with volumeBindingMode immediate (kubernetes-sigs#2415)

* docs: static capacity RFC (kubernetes-sigs#2309)

* chore: bump go version to 1.24.6 (kubernetes-sigs#2432)

* feat: Create optional operator arguments to leverage leader lease functionality (kubernetes-sigs#2433)

* chore(deps): bump the go-deps group with 5 updates (kubernetes-sigs#2442)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore(deps): bump actions/checkout from 4.2.2 to 5.0.0 in /.github/actions/install-pyroscope in the action-deps group (kubernetes-sigs#2428)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore(deps): bump the actions-deps group across 1 directory with 2 updates (kubernetes-sigs#2443)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore(deps): bump actions/cache from 4.2.3 to 4.2.4 in /.github/actions/install-deps in the action-deps group (kubernetes-sigs#2425)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* fix: do not block drifted nodes from being terminated if consolidation is disabled (kubernetes-sigs#2423)

* chore: Pin GH action SHAs for run-bench-test (kubernetes-sigs#2448)

* chore: update operatorpkg (kubernetes-sigs#2455)

* chore: Track NodeClaims in NodePoolState (kubernetes-sigs#2449)

* chore(deps): bump the k8s-go-deps group across 1 directory with 7 updates (kubernetes-sigs#2456)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* perf: Add flag to disable costly metrics controllers (kubernetes-sigs#2354)

* perf: concurrent reconciles CPU-based scaling (kubernetes-sigs#2406)

* perf: Disruption Queue Retry Duration Scaling (kubernetes-sigs#2411)

* perf: Typed Bucket Scaling (kubernetes-sigs#2420)

* ci: Include K8s version 1.33 and 1.34 in testing (kubernetes-sigs#2465)

* chore: increase MaxInstanceTypes to give cloud-providers more control over instance type truncation (kubernetes-sigs#2430)

* chore(deps): bump the go-deps group with 2 updates (kubernetes-sigs#2461)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore(deps): bump amannn/action-semantic-pull-request from 6.0.1 to 6.1.1 in the actions-deps group (kubernetes-sigs#2462)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* ci: revert k8s 1.34 addition (kubernetes-sigs#2475)

* fix: Don't schedule a pod with DRA requirements (kubernetes-sigs#2384)

* fix: support arbitrary reserved capacity labels for drift (kubernetes-sigs#2476)

* chore(deps): bump actions/checkout from 4.2.2 to 5.0.0 in /.github/actions/install-prometheus in the action-deps group (kubernetes-sigs#2426)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* fix: Fix nil pointer exception for multiNodeConsolidation (kubernetes-sigs#2472)

* fix: avoid hash collisions with duplicate match expressions (kubernetes-sigs#2479)

* ci: enable k8s 1.34 tests (kubernetes-sigs#2481)

* fix: Validate unsupported provisioners on bound PVs (kubernetes-sigs#2480)

* refactor: use iterator for iterating state nodes (kubernetes-sigs#2483)

* fix: make toolchain failing due to deletion of asciicheck (kubernetes-sigs#2485)

* fix: Handle PVC edge cases handled by kube-scheduler (kubernetes-sigs#2488)

* chore: Change appName from const to var (kubernetes-sigs#2489)

* fix: Handle unbound volumes with volumeName defined (kubernetes-sigs#2487)

* chore(deps): bump actions/setup-go from 5.5.0 to 6.0.0 in /.github/actions/install-deps in the action-deps group (kubernetes-sigs#2494)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore(deps): bump actions/setup-python from 5.6.0 to 6.0.0 in the actions-deps group (kubernetes-sigs#2493)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore(deps): bump the go-deps group with 6 updates (kubernetes-sigs#2491)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore(deps): bump the k8s-go-deps group with 4 updates (kubernetes-sigs#2492)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore: remove duplicate reconcile logging (kubernetes-sigs#2496)

* chore: bump operatorpkg version (kubernetes-sigs#2500)

* perf: Update the Node Repair Controller for requeue time  (kubernetes-sigs#2286)

* feat: Add NodeOverlay Controller Support (kubernetes-sigs#2306)

* chore(deps): bump the k8s-go-deps group with 3 updates (kubernetes-sigs#2504)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* chore: rolling back to 1.34 (kubernetes-sigs#2512)

* fix: handle nil selector when hashing in topology (kubernetes-sigs#2511)

* feat: Support Pod Level Resources (kubernetes-sigs#2383)

Signed-off-by: Tsubasa Nagasawa <[email protected]>

* fix: merge limits into requests when constructing ds pods (kubernetes-sigs#2514)

* fix: default CPU_REQUESTS when non-positive value is provided (kubernetes-sigs#2516)

* fix(node): prevent empty providerID causing false NodeClaim matches (kubernetes-sigs#2507)

* feat: Support Static Capacity (kubernetes-sigs#2521)

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: Jason Deal <[email protected]>
Co-authored-by: Jonathan Innis <[email protected]>
Co-authored-by: Andrew Mitchell <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Ryan Mistretta <[email protected]>

* fix: over provisioning static nodeclaims during controller crashes (kubernetes-sigs#2534)

* chore: drop consistency error to info log (kubernetes-sigs#2542)

* fix: flaky static provisioning unit test (kubernetes-sigs#2546)

* fix: nodepool crd definition should explicitly say replicas field as alpha (kubernetes-sigs#2554)

* chore: Update NodeRegistrationHealthy SC to use a buffer mechanism (kubernetes-sigs#2520)

---------

Signed-off-by: dependabot[bot] <[email protected]>
Signed-off-by: Max Cao <[email protected]>
Signed-off-by: Josephine Pfeiffer <[email protected]>
Signed-off-by: Tsubasa Nagasawa <[email protected]>
Co-authored-by: Derek Frank <[email protected]>
Co-authored-by: Jonathan Innis <[email protected]>
Co-authored-by: Lê Minh Quân <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Jigisha Patil <[email protected]>
Co-authored-by: Amanuel Engeda <[email protected]>
Co-authored-by: Max Cao <[email protected]>
Co-authored-by: Aidan Rowe <[email protected]>
Co-authored-by: Daniel Lopes <[email protected]>
Co-authored-by: Saurav Agarwalla <[email protected]>
Co-authored-by: cosimomeli <[email protected]>
Co-authored-by: Jason Deal <[email protected]>
Co-authored-by: Reed Schalo <[email protected]>
Co-authored-by: Josephine Pfeiffer <[email protected]>
Co-authored-by: Sumukha Radhakrishna <[email protected]>
Co-authored-by: Andy Townsend <[email protected]>
Co-authored-by: Sumukha Radhakrishna <[email protected]>
Co-authored-by: ryan-mist <[email protected]>
Co-authored-by: Brandon Wagner <[email protected]>
Co-authored-by: Alima Azamat <[email protected]>
Co-authored-by: Andrew Mitchell <[email protected]>
Co-authored-by: Tsubasa Nagasawa <[email protected]>
Co-authored-by: Neil <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

approved Indicates a PR has been approved by an approver from all required OWNERS files. cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. lgtm "Looks good to me", indicates that a PR is ready to be merged. needs-ok-to-test Indicates a PR that requires an org member to verify it is safe to test. size/L Denotes a PR that changes 100-499 lines, ignoring generated files.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Support Pod Level Resources

4 participants