[fix][kubectl-plugin] set worker group CPU limit #2958
Merged
+77
−12
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
davidxia
commented
Feb 6, 2025
| return nil | ||
| } | ||
|
|
||
| func createWorkerGroupSpec(options *CreateWorkerGroupOptions) rayv1.WorkerGroupSpec { |
There was a problem hiding this comment.
create a function for unit testing
| @@ -144,6 +161,7 @@ func (options *CreateWorkerGroupOptions) Run(ctx context.Context, factory cmduti | |||
| corev1.ResourceMemory: resource.MustParse(options.workerMemory), | |||
| }, | |||
| Limits: corev1.ResourceList{ | |||
| corev1.ResourceCPU: resource.MustParse(options.workerCPU), | |||
There was a problem hiding this comment.
the important addition
|
Let me know when this is ready for review. |
|
Ready for review. I’ll fix conflicts later. |
when creating a new worker group with `kubectl ray create workergroup`. Write a unit test. I noticed we are setting resource limits equal to resource requests everywhere else but in this command. I have a K8s [LimitRange] that prevented the creation of these worker Pods that had CPU limit defaulting to less than their CPU requests. Describing the RayCluster showed this warning event. `Failed to create worker Pod hyperkube/, Pod "dxia-test-other-group-worker-pm2sh" is invalid: spec.containers[0].resources.requests: Invalid value: "2": must be less than or equal to cpu limit of 250m` Signed-off-by: David Xia <[email protected]> [LimitRange]: https://kubernetes.io/docs/concepts/policy/limit-range/
andrewsykim
added a commit
that referenced
this pull request
Feb 10, 2025
* [RayService] More envtests that follow the most common scenario in the RayService code path (#2880) Signed-off-by: Rueian <[email protected]> * [RayService] Remove outdated env tests (#2886) Signed-off-by: kaihsun <[email protected]> * [RayService] Refactor envtests (#2888) * [docs][ray-operator] fix typo in Golang version (#2893) Project uses Golang 1.22 not 1.20. Signed-off-by: David Xia <[email protected]> * [Fix][kubectl-plugin] Fix no context nil error SIGSEGV in tests (#2892) Signed-off-by: Chi-Sheng Liu <[email protected]> * [CI] Enable testifylint rule (#2896) Signed-off-by: Chi-Sheng Liu <[email protected]> * [CI] Enable `testifylint` `error-nil` rule (#2907) * [kubectl-plugin][feat] support specifying number of head GPUs (#2895) when creating a RayCluster with `kubectl ray create cluster NAME --head-gpu N`. Similar to the `--worker-gpu` switch. Signed-off-by: David Xia <[email protected]> * Use webhook.CustomValidator instead of deprecated webhook.Validator. (#2803) * Use webhook.CustomValidator instead of deprecated webhook.Validator. * Move RayClusterWebhook to pkg/webhook/v1. * [kubectl-plugin] update context error messages (#2891) to tell user they can use `--context` to set the K8s context. Add tests Signed-off-by: David Xia <[email protected]> Co-authored-by: Chi-Sheng Liu <[email protected]> * [Fix][kubectl-plugin] make tests use a temporary kube config (#2894) file that's not at the default path. Right now tests fail if there's a K8s current context set with a command like `kubectl config use-context my-context`. This change allows tests to pass regardless of current context. Signed-off-by: David Xia <[email protected]> * [CI] Enable testifylint formatter rule (#2915) Signed-off-by: 400Ping <[email protected]> * [CI] Enable `testifylint` `empty` rule (#2908) * [CI] Enable testifylint bool-compare rule (#2911) Signed-off-by: 400Ping <[email protected]> * [CI] Auto download golang tools in pre-commit (#2917) Signed-off-by: Chi-Sheng Liu <[email protected]> * [RayService] a safeguard for preventing overriding the pending cluster during a upgrade (#2887) Signed-off-by: Rueian <[email protected]> * [RayService] Refactor unit tests for ShouldPrepareNewCluster (#2928) Signed-off-by: kaihsun <[email protected]> * [chore][kubectl-plugin] use consistent capitalization (#2922) in comments and console messages. * capitalize "Ray" when used as a proper noun * capitalize Ray K8s CRDs like "RayJob" and "RayCluster" * capitalize acronyms like "YAML" * fix some minor typos No functional changes. Signed-off-by: David Xia <[email protected]> * [CI] Enable `testifylint` `require-error` rule (#2909) * [kubectl-plugin] support general `kubectl` switches like `--context` (#2883) * [CI] Fix lint error (require-error) (#2931) Signed-off-by: Chi-Sheng Liu <[email protected]> * [CI] Enable `testifylint` `float-compare` rule (#2910) * [docs] move pre-commit instructions to main dev docs (#2921) * [CI] Enable `testifylint` `expected-actual` rule (#2914) * [CI] Generate CRD json schema separately in pre-commit (#2930) * [release][1/N] Update YAMLs from Ray 2.9 to Ray 2.41 (#2934) * Delete `[raycluster|rayjob|rayservice]_types_test.go` unnecessary tests (#2935) * [release][2/N] Update RayCluster Helm chart from Ray 2.9 to Ray 2.41 (#2936) Signed-off-by: kaihsun <[email protected]> * [release][3/N] Update RayService e2e tests YAML files from Ray 2.9 to Ray 2.41 (#2937) * [release][4/N] Update Ray images / versions in kubectl plugin (#2938) Signed-off-by: kaihsun <[email protected]> * [release][5/N] Update some RayJob YAMLs from Ray 2.9 to Ray 2.41 (#2941) Signed-off-by: kaihsun <[email protected]> * [release][6/N] Remove unnecessary YAMLs (#2946) Signed-off-by: kaihsun <[email protected]> * [docs][kubectl-plugin] add dev docs (#2912) Signed-off-by: David Xia <[email protected]> Co-authored-by: Chi-Sheng Liu <[email protected]> * [CI] Add shellcheck and fix error of it (#2933) * [chore][kubectl-plugin] use better test assertions (#2955) Fix two places. * use `assert.EqualError()` instead of `assert.Error()` to check the error string is what we expect * use testify assertion instead of string checking for better error output Signed-off-by: David Xia <[email protected]> * [chore] add Markdown linting pre-commit hook (#2953) Ignore most rules we violate for now. Fix these two violations. ``` pre-commit run markdownlint --all-files --show-diff-on-failure ... clients/python-client/README.md:106 MD001/heading-increment Heading levels should only increment by one level at a time [Expected: h3; Actual: h4] CHANGELOG.md:420:1 MD030/list-marker-space Spaces after list markers [Expected: 1; Actual: 2] ``` Signed-off-by: David Xia <[email protected]> * [chore][kubectl-plugin] use consistent capitalization (#2950) in comments and console messages. capitalize "Ray" when used as a proper noun. Similar to #2922. Signed-off-by: David Xia <[email protected]> * [Fix][RayJob] Invalid quote for RayJob submitter (#2949) Closes: #2943 Signed-off-by: Chi-Sheng Liu <[email protected]> * [release][7/N] Update RayService YAMLs (#2956) * [CI] Enable testifylint len rule (#2945) Signed-off-by: LeoLiao123 <[email protected]> * [docs][kubectl-plugin] improve help messages (#2952) Signed-off-by: David Xia <[email protected]> * [kubectl-plugin] Fix panic when GPU resource is not set (#2954) Signed-off-by: win5923 <[email protected]> * [release][8/N] Upgrade Stable Diffusion RayService to Ray 2.41 (#2960) * [docs][kubectl-plugin] fix incorrect example commands (#2951) Signed-off-by: David Xia <[email protected]> Co-authored-by: Chi-Sheng Liu <[email protected]> * [fix][kubectl-plugin] set worker group CPU limit (#2958) when creating a new worker group with `kubectl ray create workergroup`. Write a unit test. I noticed we are setting resource limits equal to resource requests everywhere else but in this command. I have a K8s [LimitRange] that prevented the creation of these worker Pods that had CPU limit defaulting to less than their CPU requests. Describing the RayCluster showed this warning event. `Failed to create worker Pod hyperkube/, Pod "dxia-test-other-group-worker-pm2sh" is invalid: spec.containers[0].resources.requests: Invalid value: "2": must be less than or equal to cpu limit of 250m` Signed-off-by: David Xia <[email protected]> [LimitRange]: https://kubernetes.io/docs/concepts/policy/limit-range/ * [RayJob] Deflaky RayJob e2e tests (#2963) Signed-off-by: kaihsun <[email protected]> * [RayService] Deflaky RayService envtest (#2962) Signed-off-by: kaihsun <[email protected]> * [release][9/N] Update text summarizer RayService to Ray 2.41 (#2961) * [RayService] adapter vllm 0.6.1.post2 (#2823) * adapter vllm 0.6.1.post2 * fix var define * Unify the cpu Settings in serve.py in service.yaml, all set to 1 * Maintain the configuration with vllm0.5x * [Release] Upgrade ray-job.batch-inference.yaml image to 2.41 (#2971) Signed-off-by: Chi-Sheng Liu <[email protected]> * [chore][docs] enable Markdownlint rule MD010 (#2975) * [CI] Change Pre-commit-shellcheck-to-shellcheck-py (#2974) Signed-off-by: owenowenisme <[email protected]> * [release] Update Yunikorn YAML file to Ray 2.41 (#2969) Signed-off-by: Cheng-Yeh Chung <[email protected]> * [release] Update YuniKorn YAML files to Ray 2.41 (#2976) Signed-off-by: win5923 <[email protected]> * [chore][docs] enable Markdownlint rule MD004 (#2973) [Unordered list style][1] [1]: https://github.com/DavidAnson/markdownlint/blob/main/doc/Rules.md#md004---unordered-list-style * [Test] Use GcsFaultToleranceOptions in test and backward compatibility (#2972) * Update samples to use Ray 2.41.0 images (#2964) Signed-off-by: Andrew Sy Kim <[email protected]> * Update TPU Ray CR manifests to use Ray 2.41.0 (#2965) * [Refactor] Use constants for image tag, image repo, and versions in golang to avoid hard-coded strings (#2978) --------- Signed-off-by: Rueian <[email protected]> Signed-off-by: kaihsun <[email protected]> Signed-off-by: David Xia <[email protected]> Signed-off-by: Chi-Sheng Liu <[email protected]> Signed-off-by: 400Ping <[email protected]> Signed-off-by: LeoLiao123 <[email protected]> Signed-off-by: win5923 <[email protected]> Signed-off-by: owenowenisme <[email protected]> Signed-off-by: Cheng-Yeh Chung <[email protected]> Signed-off-by: Andrew Sy Kim <[email protected]> Co-authored-by: Rueian <[email protected]> Co-authored-by: Kai-Hsun Chen <[email protected]> Co-authored-by: David Xia <[email protected]> Co-authored-by: Chi-Sheng Liu <[email protected]> Co-authored-by: David Xia <[email protected]> Co-authored-by: Mykhailo Bobrovskyi <[email protected]> Co-authored-by: Ping <[email protected]> Co-authored-by: Owen Lin <[email protected]> Co-authored-by: Leo Liao <[email protected]> Co-authored-by: Blocka <[email protected]> Co-authored-by: zrant <[email protected]> Co-authored-by: kenchung285 <[email protected]> Co-authored-by: fscnick <[email protected]> Co-authored-by: ryanaoleary <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
when creating a new worker group with
kubectl ray create workergroup. Write a unit test.I noticed we are setting resource limits equal to resource requests everywhere else but in this command. I have a K8s LimitRange that prevented the creation of these worker Pods that had CPU limit defaulting to less than their CPU requests. Describing the RayCluster showed this warning event.
Failed to create worker Pod hyperkube/, Pod "dxia-test-other-group-worker-pm2sh" is invalid: spec.containers[0].resources.requests: Invalid value: "2": must be less than or equal to cpu limit of 250mSigned-off-by: David Xia [email protected]
Checks