deploy/cloud/operator/internal/dynamo/component_worker.go (1)

45-45: Nit: Use a typed ResourceName for the GPU key

Minor readability/consistency improvement and avoids any ambiguity around typed keys in ResourceList.

Apply this diff:

-           "nvidia.com/gpu": resource.MustParse("1"),
+           corev1.ResourceName("nvidia.com/gpu"): resource.MustParse("1"),

deploy/cloud/operator/internal/controller/dynamocomponentdeployment_controller_test.go (1)

708-711: Add a test for the default case (no CPU/mem limits in spec)

To guard the new default, add a case where spec.limits omits CPU/memory and assert the generated Pod’s container.Resources.Limits includes only GPU (and no CPU/memory). Also verify the /dev/shm EmptyDir size handling when no memory limit is present (should either be unset or follow the intended fallback).

Here’s a minimal example you can adapt:

func Test_generateLeaderWorkerSet_defaults_NoCpuMemLimits(t *testing.T) {
  g := gomega.NewWithT(t)
  s := scheme.Scheme
  _ = v1alpha1.AddToScheme(s)
  _ = corev1.AddToScheme(s)
  _ = leaderworkersetv1.AddToScheme(s)

  dcd := &v1alpha1.DynamoComponentDeployment{
    ObjectMeta: metav1.ObjectMeta{Name: "default-lws", Namespace: "default"},
    Spec: v1alpha1.DynamoComponentDeploymentSpec{
      DynamoComponent: "comp", DynamoTag: "tag",
      DynamoComponentDeploymentSharedSpec: v1alpha1.DynamoComponentDeploymentSharedSpec{
        ComponentType:   string(commonconsts.ComponentTypeWorker),
        ServiceName:     "svc",
        DynamoNamespace: ptr.To("default"),
        Multinode:       &v1alpha1.MultinodeSpec{NodeCount: 2},
        Resources: &common.Resources{
          Requests: &common.ResourceItem{CPU: "300m", Memory: "500Mi"},
          Limits:   &common.ResourceItem{GPU: "1"}, // no CPU/memory limits
        },
        ExtraPodSpec: &dynamoCommon.ExtraPodSpec{
          MainContainer: &corev1.Container{Image: "test:latest"},
        },
      },
    },
  }

  fakeClient := fake.NewClientBuilder().WithScheme(s).WithObjects(dcd).Build()
  r := &DynamoComponentDeploymentReconciler{
    Client:   fakeClient,
    Recorder: record.NewFakeRecorder(10),
  }

  got, _, err := r.generateLeaderWorkerSet(context.Background(), generateResourceOption{
    dynamoComponentDeployment: dcd,
    instanceID:                ptr.To(0),
  })
  g.Expect(err).ToNot(gomega.HaveOccurred())

  // Ensure only GPU is limited; CPU/memory limits are absent
  limits := got.Spec.LeaderWorkerTemplate.LeaderTemplate.Spec.Containers[0].Resources.Limits
  g.Expect(limits).To(gomega.HaveKeyWithValue(corev1.ResourceName("nvidia.com/gpu"), resource.MustParse("1")))
  g.Expect(limits).ToNot(gomega.HaveKey(corev1.ResourceCPU))
  g.Expect(limits).ToNot(gomega.HaveKey(corev1.ResourceMemory))
}

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro

💡 Knowledge Base configuration:

• MCP integration is disabled by default for public repositories
• Jira integration is disabled by default for public repositories
• Linear integration is disabled by default for public repositories

You can enable these sources in your CodeRabbit configuration.

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1474
File: deploy/cloud/operator/internal/controller/dynamocomponent_controller.go:1308-1312
Timestamp: 2025-06-11T21:29:28.650Z
Learning: User julienmancuso expects replies in English; avoid switching languages unless explicitly requested.