[kubectl-plugin] remove CPU limits by default (#3243)

Signed-off-by: Andrew Sy Kim <[email protected]>

Author: andrewsykim

kubectl-plugin/pkg/util/generation/generation.go

@@ -69,8 +69,8 @@ func (rayJobObject *RayJobYamlObject) GenerateRayJobApplyConfig() *rayv1ac.RayJo
 	return rayJobApplyConfig
 }
 
-// generateResources returns a corev1.ResourceList with the given CPU, memory, ephemeral storage, and GPU values for both requests and limits
-func generateResources(cpu, memory, ephemeralStorage, gpu string) corev1.ResourceList {
+// generateRequestResources returns a corev1.ResourceList with the given CPU, memory, ephemeral storage, and GPU values for only resource requests
+func generateRequestResources(cpu, memory, ephemeralStorage, gpu string) corev1.ResourceList {
 	resources := corev1.ResourceList{
 		corev1.ResourceCPU:    resource.MustParse(cpu),
 		corev1.ResourceMemory: resource.MustParse(memory),
@@ -87,6 +87,23 @@ func generateResources(cpu, memory, ephemeralStorage, gpu string) corev1.Resourc
 	return resources
 }
 
+// generateLimitResources returns a corev1.ResourceList with the given memory, ephemeral storage, and GPU values for only resource limits
+func generateLimitResources(memory, ephemeralStorage, gpu string) corev1.ResourceList {
+	resources := corev1.ResourceList{
+		corev1.ResourceMemory: resource.MustParse(memory),
+	}
+	if ephemeralStorage != "" {
+		resources[corev1.ResourceEphemeralStorage] = resource.MustParse(ephemeralStorage)
+	}
+
+	gpuResource := resource.MustParse(gpu)
+	if !gpuResource.IsZero() {
+		resources[corev1.ResourceName(util.ResourceNvidiaGPU)] = gpuResource
+	}
+
+	return resources
+}
+
 func (rayClusterSpecObject *RayClusterSpecObject) generateRayClusterSpec() *rayv1ac.RayClusterSpecApplyConfiguration {
 	// TODO: Look for better workaround/fixes for RayStartParams. Currently using `WithRayStartParams()` requires
 	// a non-empty map with valid key value pairs and will not populate the field with empty/nil values. This
@@ -100,8 +117,10 @@ func (rayClusterSpecObject *RayClusterSpecObject) generateRayClusterSpec() *rayv
 	maps.Copy(headRayStartParams, rayClusterSpecObject.HeadRayStartParams)
 	maps.Copy(workerRayStartParams, rayClusterSpecObject.WorkerRayStartParams)
 
-	headResources := generateResources(rayClusterSpecObject.HeadCPU, rayClusterSpecObject.HeadMemory, rayClusterSpecObject.HeadEphemeralStorage, rayClusterSpecObject.HeadGPU)
-	workerResources := generateResources(rayClusterSpecObject.WorkerCPU, rayClusterSpecObject.WorkerMemory, rayClusterSpecObject.WorkerEphemeralStorage, rayClusterSpecObject.WorkerGPU)
+	headRequestResources := generateRequestResources(rayClusterSpecObject.HeadCPU, rayClusterSpecObject.HeadMemory, rayClusterSpecObject.HeadEphemeralStorage, rayClusterSpecObject.HeadGPU)
+	headLimitResources := generateLimitResources(rayClusterSpecObject.HeadMemory, rayClusterSpecObject.HeadEphemeralStorage, rayClusterSpecObject.HeadGPU)
+	workerRequestResources := generateRequestResources(rayClusterSpecObject.WorkerCPU, rayClusterSpecObject.WorkerMemory, rayClusterSpecObject.WorkerEphemeralStorage, rayClusterSpecObject.WorkerGPU)
+	workerLimitResources := generateLimitResources(rayClusterSpecObject.WorkerMemory, rayClusterSpecObject.WorkerEphemeralStorage, rayClusterSpecObject.WorkerGPU)
 
 	rayClusterSpec := rayv1ac.RayClusterSpec().
 		WithRayVersion(rayClusterSpecObject.RayVersion).
@@ -114,8 +133,8 @@ func (rayClusterSpecObject *RayClusterSpecObject) generateRayClusterSpec() *rayv
 						WithName("ray-head").
 						WithImage(rayClusterSpecObject.Image).
 						WithResources(corev1ac.ResourceRequirements().
-							WithRequests(headResources).
-							WithLimits(headResources)).
+							WithRequests(headRequestResources).
+							WithLimits(headLimitResources)).
 						WithPorts(corev1ac.ContainerPort().WithContainerPort(6379).WithName("gcs-server"),
 							corev1ac.ContainerPort().WithContainerPort(8265).WithName("dashboard"),
 							corev1ac.ContainerPort().WithContainerPort(10001).WithName("client")))))).
@@ -130,8 +149,8 @@ func (rayClusterSpecObject *RayClusterSpecObject) generateRayClusterSpec() *rayv
 						WithName("ray-worker").
 						WithImage(rayClusterSpecObject.Image).
 						WithResources(corev1ac.ResourceRequirements().
-							WithRequests(workerResources).
-							WithLimits(workerResources))))))
+							WithRequests(workerRequestResources).
+							WithLimits(workerLimitResources))))))
 
 	return rayClusterSpec
 }

kubectl-plugin/pkg/util/generation/generation_test.go

@@ -171,7 +171,6 @@ spec:
             name: client
           resources:
             limits:
-              cpu: "1"
               memory: 5Gi
               nvidia.com/gpu: "1"
             requests:
@@ -191,7 +190,6 @@ spec:
           name: ray-worker
           resources:
             limits:
-              cpu: "2"
               memory: 10Gi
             requests:
               cpu: "2"
@@ -202,42 +200,52 @@ spec:
 
 func TestGenerateResources(t *testing.T) {
 	tests := []struct {
-		expectedResources corev1.ResourceList
-		name              string
-		cpu               string
-		memory            string
-		ephemeralStorage  string
-		gpu               string
+		expectedRequestResources corev1.ResourceList
+		expectedLimitResources   corev1.ResourceList
+		name                     string
+		cpu                      string
+		memory                   string
+		ephemeralStorage         string
+		gpu                      string
 	}{
 		{
 			name:             "should generate resources with CPU, memory, ephemeral storage, and GPU",
 			cpu:              "1",
 			memory:           "5Gi",
 			ephemeralStorage: "10Gi",
 			gpu:              "1",
-			expectedResources: corev1.ResourceList{
+			expectedRequestResources: corev1.ResourceList{
 				corev1.ResourceCPU:                          resource.MustParse("1"),
 				corev1.ResourceMemory:                       resource.MustParse("5Gi"),
 				corev1.ResourceEphemeralStorage:             resource.MustParse("10Gi"),
 				corev1.ResourceName(util.ResourceNvidiaGPU): resource.MustParse("1"),
 			},
+			expectedLimitResources: corev1.ResourceList{
+				corev1.ResourceMemory:                       resource.MustParse("5Gi"),
+				corev1.ResourceEphemeralStorage:             resource.MustParse("10Gi"),
+				corev1.ResourceName(util.ResourceNvidiaGPU): resource.MustParse("1"),
+			},
 		},
 		{
 			name:             "should only generate resources with CPU and memory if ephemeral storage isn't set and GPUs are 0",
 			cpu:              "1",
 			memory:           "5Gi",
 			ephemeralStorage: "",
 			gpu:              "0",
-			expectedResources: corev1.ResourceList{
+			expectedRequestResources: corev1.ResourceList{
 				corev1.ResourceCPU:    resource.MustParse("1"),
 				corev1.ResourceMemory: resource.MustParse("5Gi"),
 			},
+			expectedLimitResources: corev1.ResourceList{
+				corev1.ResourceMemory: resource.MustParse("5Gi"),
+			},
 		},
 	}
 
 	for _, test := range tests {
 		t.Run(test.name, func(t *testing.T) {
-			assert.Equal(t, test.expectedResources, generateResources(test.cpu, test.memory, test.ephemeralStorage, test.gpu))
+			assert.Equal(t, test.expectedRequestResources, generateRequestResources(test.cpu, test.memory, test.ephemeralStorage, test.gpu))
+			assert.Equal(t, test.expectedLimitResources, generateLimitResources(test.memory, test.ephemeralStorage, test.gpu))
 		})
 	}
 }
@@ -285,7 +293,6 @@ func TestGenerateRayClusterSpec(t *testing.T) {
 									corev1.ResourceName(util.ResourceNvidiaGPU): resource.MustParse("1"),
 								},
 								Limits: &corev1.ResourceList{
-									corev1.ResourceCPU:                          resource.MustParse("1"),
 									corev1.ResourceMemory:                       resource.MustParse("5Gi"),
 									corev1.ResourceEphemeralStorage:             resource.MustParse("10Gi"),
 									corev1.ResourceName(util.ResourceNvidiaGPU): resource.MustParse("1"),
@@ -331,7 +338,6 @@ func TestGenerateRayClusterSpec(t *testing.T) {
 										corev1.ResourceMemory: resource.MustParse("10Gi"),
 									},
 									Limits: &corev1.ResourceList{
-										corev1.ResourceCPU:    resource.MustParse("2"),
 										corev1.ResourceMemory: resource.MustParse("10Gi"),
 									},
 								},