PR #37959: [Bugfix] Fix Helm chart Deployment using hardcoded labels instead of chart.labels

• Repository: vllm-project/vllm
• Author: simpx
• State: closed | merged: False
• Link: https://github.com/vllm-project/vllm/pull/37959

Description (problem / solution / changelog)

Essential Changes

The Deployment template in examples/online_serving/chart-helm had hardcoded selector.matchLabels and template.metadata.labels:

Meanwhile, the Service template already uses the chart.labels helper (which reads from values.labels). When users customize labels in values.yaml, the Service selector no longer matches the Deployment pod labels, leaving Service endpoints empty.

This PR replaces the hardcoded labels with {{ include "chart.labels" . }} to keep the Deployment selector/labels consistent with the Service selector.

Fixes #37942

Testing

• Verified helm template output before and after: selector and pod labels now match Service selector when values.labels is customized.
• Confirmed default values.yaml labels (environment: test, release: test) still work correctly.

Changed files

• DAILY_LEARNINGS.md (added, +974/-0)
• dgx_spark_issues.md (added, +423/-0)
• examples/online_serving/chart-helm/templates/deployment.yaml (modified, +3/-5)

---

PR #38340: [Doc] fix selector/label mismatch in helm chart

• Repository: vllm-project/vllm
• Author: utsumi-fj
• State: open | merged: False
• Link: https://github.com/vllm-project/vllm/pull/38340

Description (problem / solution / changelog)

Purpose

Fixes #37942

This fixes label/selector consistency in examples/online_serving/chart-helm by:

• introducing stable selector labels (app.kubernetes.io/name, app.kubernetes.io/instance) for Deployment/Service selectors,
• applying common chart labels across resources
• and updating values.schema.json so .Values.labels supports arbitrary string key/value pairs instead of only environment/release.

Breaking change note:

• Deployment.spec.selector.matchLabels changed from hardcoded environment/release to app.kubernetes.io/name/app.kubernetes.io/instance.
• Existing releases may require recreation (or equivalent force-replace flow) when upgrading because Deployment selectors are immutable in Kubernetes.

Test Plan

• helm lint examples/online_serving/chart-helm
• helm template test examples/online_serving/chart-helm
• helm unittest examples/online_serving/chart-helm

Test Result

• helm lint passes
• helm template renders expected selectors/labels
• helm unittest passes

<details> <summary> Test Result Details </summary> <pre><code> $ helm lint examples/online_serving/chart-helm ==> Linting examples/online_serving/chart-helm [INFO] Chart.yaml: icon is recommended 1 chart(s) linted, 0 chart(s) failed </code></pre> <pre><code> $ helm template test examples/online_serving/chart-helm --- # Source: chart-vllm/templates/poddisruptionbudget.yaml apiVersion: policy/v1 kind: PodDisruptionBudget metadata: name: "test-pdb" namespace: default labels: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test helm.sh/chart: chart-vllm-0.0.1 app.kubernetes.io/managed-by: Helm spec: maxUnavailable: 1 --- # Source: chart-vllm/templates/secrets.yaml apiVersion: v1 kind: Secret metadata: name: "test-secrets" namespace: default labels: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test helm.sh/chart: chart-vllm-0.0.1 app.kubernetes.io/managed-by: Helm type: Opaque data: --- # Source: chart-vllm/templates/pvc.yaml apiVersion: v1 kind: PersistentVolumeClaim metadata: name: "test-storage-claim" namespace: default labels: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test helm.sh/chart: chart-vllm-0.0.1 app.kubernetes.io/managed-by: Helm spec: accessModes: - ReadWriteOnce resources: requests: storage: 1Gi --- # Source: chart-vllm/templates/service.yaml apiVersion: v1 kind: Service metadata: name: "test-service" namespace: default labels: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test helm.sh/chart: chart-vllm-0.0.1 app.kubernetes.io/managed-by: Helm spec: type: ClusterIP ports: - name: "service-port" port: 80 targetPort: "container-port" protocol: TCP selector: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test --- # Source: chart-vllm/templates/deployment.yaml apiVersion: apps/v1 kind: Deployment metadata: name: "test-deployment-vllm" namespace: default labels: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test helm.sh/chart: chart-vllm-0.0.1 app.kubernetes.io/managed-by: Helm spec: replicas: 1 strategy: rollingUpdate: maxSurge: 100% maxUnavailable: 0 selector: matchLabels: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test progressDeadlineSeconds: 1200 template: metadata: labels: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test helm.sh/chart: chart-vllm-0.0.1 app.kubernetes.io/managed-by: Helm spec: containers: - name: "vllm" image: "vllm/vllm-openai:latest" command : - vllm - serve - /data/ - --served-model-name - opt-125m - --enforce-eager - --dtype - bfloat16 - --block-size - "16" - --host - 0.0.0.0 - --port - "8000" securityContext: runAsNonRoot: false imagePullPolicy: IfNotPresent env: [] ports: - name: "container-port" containerPort: 8000 readinessProbe: failureThreshold: 3 httpGet: path: /health port: 8000 initialDelaySeconds: 5 periodSeconds: 5 livenessProbe: failureThreshold: 3 httpGet: path: /health port: 8000 initialDelaySeconds: 15 periodSeconds: 10 resources: requests: memory: "16Gi" cpu: "4" nvidia.com/gpu: "1" limits: memory: "16Gi" cpu: "4" nvidia.com/gpu: "1" volumeMounts: - name: test-storage mountPath: /data initContainers: - name: wait-download-model image: amazon/aws-cli:2.6.4 imagePullPolicy: IfNotPresent command: ["/bin/bash"] args: - -eucx - while aws --endpoint-url $S3_ENDPOINT_URL s3 sync --dryrun s3://$S3_BUCKET_NAME/$S3_PATH /data | grep -q download; do sleep 10; done env: - name: S3_ENDPOINT_URL valueFrom: secretKeyRef: name: test-secrets key: s3endpoint - name: S3_BUCKET_NAME valueFrom: secretKeyRef: name: test-secrets key: s3bucketname - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: name: test-secrets key: s3accesskeyid - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: name: test-secrets key: s3accesskey - name: S3_PATH value: "relative_s3_model_path/opt-125m" - name: AWS_EC2_METADATA_DISABLED value: "true" resources: requests: cpu: 200m memory: 1Gi limits: cpu: 500m memory: 2Gi volumeMounts: - name: test-storage mountPath: /data volumes: - name: test-storage persistentVolumeClaim: claimName: test-storage-claim runtimeClassName: nvidia affinity: nodeAffinity: requiredDuringSchedulingIgnoredDuringExecution: nodeSelectorTerms: - matchExpressions: - key: nvidia.com/gpu.product operator: In values: - TYPE_GPU_USED --- # Source: chart-vllm/templates/job.yaml apiVersion: batch/v1 kind: Job metadata: name: "test-init-vllm" namespace: default labels: app.kubernetes.io/name: chart-vllm app.kubernetes.io/instance: test helm.sh/chart: chart-vllm-0.0.1 app.kubernetes.io/managed-by: Helm spec: ttlSecondsAfterFinished: 100 template: metadata: name: init-vllm spec: containers: - name: job-download-model image: amazon/aws-cli:2.6.4 imagePullPolicy: IfNotPresent command: ["/bin/bash"] args: - -eucx - aws --endpoint-url $S3_ENDPOINT_URL s3 sync s3://$S3_BUCKET_NAME/$S3_PATH /data env: - name: S3_ENDPOINT_URL valueFrom: secretKeyRef: name: test-secrets key: s3endpoint - name: S3_BUCKET_NAME valueFrom: secretKeyRef: name: test-secrets key: s3bucketname - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: name: test-secrets key: s3accesskeyid - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: name: test-secrets key: s3accesskey - name: S3_PATH value: "relative_s3_model_path/opt-125m" - name: AWS_EC2_METADATA_DISABLED value: "true" volumeMounts: - name: test-storage mountPath: /data resources: requests: cpu: 200m memory: 1Gi limits: cpu: 500m memory: 2Gi restartPolicy: OnFailure volumes: - name: test-storage persistentVolumeClaim: claimName: "test-storage-claim" </code></pre> <pre><code> $ helm unittest examples/online_serving/chart-helm ### Chart [ chart-vllm ] examples/online_serving/chart-helm PASS test deployment examples/online_serving/chart-helm/tests/deployment_test.yaml PASS test job examples/online_serving/chart-helm/tests/job_test.yaml PASS test pvc examples/online_serving/chart-helm/tests/pvc_test.yaml Charts: 1 passed, 1 total Test Suites: 3 passed, 3 total Tests: 6 passed, 6 total Snapshot: 0 passed, 0 total Time: 44.598564ms </code></pre> </details>

---

<details> <summary> Essential Elements of an Effective PR Description Checklist </summary>

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

</details>

Changed files

• docs/deployment/frameworks/helm.md (modified, +1/-1)
• examples/online_serving/chart-helm/templates/_helpers.tpl (modified, +21/-3)
• examples/online_serving/chart-helm/templates/configmap.yaml (modified, +3/-1)
• examples/online_serving/chart-helm/templates/deployment.yaml (modified, +4/-6)
• examples/online_serving/chart-helm/templates/hpa.yaml (modified, +3/-1)
• examples/online_serving/chart-helm/templates/job.yaml (modified, +3/-1)
• examples/online_serving/chart-helm/templates/poddisruptionbudget.yaml (modified, +3/-1)
• examples/online_serving/chart-helm/templates/pvc.yaml (modified, +3/-1)
• examples/online_serving/chart-helm/templates/secrets.yaml (modified, +3/-1)
• examples/online_serving/chart-helm/templates/service.yaml (modified, +3/-1)
• examples/online_serving/chart-helm/tests/labels_test.yaml (added, +107/-0)
• examples/online_serving/chart-helm/values.schema.json (modified, +4/-13)
• examples/online_serving/chart-helm/values.yaml (modified, +2/-3)