Merge pull request #33112 from vishh/fix-qos

Automatic merge from submit-queue

Ignore opaque or counted resources from Pod QoS

Fixes #33108

pkg/kubelet/qos/qos.go

@@ -19,6 +19,7 @@ package qos
 import (
 	"k8s.io/kubernetes/pkg/api"
 	"k8s.io/kubernetes/pkg/api/resource"
+	"k8s.io/kubernetes/pkg/util/sets"
 )
 
 // isResourceGuaranteed returns true if the container's resource requirements are Guaranteed.
@@ -53,6 +54,9 @@ func GetPodQOS(pod *api.Pod) QOSClass {
 	for _, container := range pod.Spec.Containers {
 		// process requests
 		for name, quantity := range container.Resources.Requests {
+			if !supportedQoSComputeResources.Has(string(name)) {
+				continue
+			}
 			if quantity.Cmp(zeroQuantity) == 1 {
 				delta := quantity.Copy()
 				if _, exists := requests[name]; !exists {
@@ -65,6 +69,9 @@ func GetPodQOS(pod *api.Pod) QOSClass {
 		}
 		// process limits
 		for name, quantity := range container.Resources.Limits {
+			if !supportedQoSComputeResources.Has(string(name)) {
+				continue
+			}
 			if quantity.Cmp(zeroQuantity) == 1 {
 				delta := quantity.Copy()
 				if _, exists := limits[name]; !exists {
@@ -75,7 +82,7 @@ func GetPodQOS(pod *api.Pod) QOSClass {
 				}
 			}
 		}
-		if len(container.Resources.Limits) != len(supportedComputeResources) {
+		if len(limits) != len(supportedQoSComputeResources) {
 			isGuaranteed = false
 		}
 	}
@@ -92,8 +99,7 @@ func GetPodQOS(pod *api.Pod) QOSClass {
 		}
 	}
 	if isGuaranteed &&
-		len(requests) == len(limits) &&
+		len(requests) == len(limits) {
-		len(limits) == len(supportedComputeResources) {
 		return Guaranteed
 	}
 	return Burstable
@@ -118,17 +124,14 @@ func GetQOS(container *api.Container) QOSList {
 	return resourceToQOS
 }
 
-// supportedComputeResources is the list of supported compute resources
+// supportedComputeResources is the list of compute resources for with QoS is supported.
-var supportedComputeResources = []api.ResourceName{
+var supportedQoSComputeResources = sets.NewString(string(api.ResourceCPU), string(api.ResourceMemory))
-	api.ResourceCPU,
-	api.ResourceMemory,
-}
 
 // allResources returns a set of all possible resources whose mapped key value is true if present on the container
 func allResources(container *api.Container) map[api.ResourceName]bool {
 	resources := map[api.ResourceName]bool{}
-	for _, resource := range supportedComputeResources {
+	for _, resource := range supportedQoSComputeResources.List() {
-		resources[resource] = false
+		resources[api.ResourceName(resource)] = false
 	}
 	for resource := range container.Resources.Requests {
 		resources[resource] = true

pkg/kubelet/qos/qos_test.go

@@ -34,6 +34,11 @@ func getResourceList(cpu, memory string) api.ResourceList {
 	return res
 }
 
+func addResource(rName, value string, rl api.ResourceList) api.ResourceList {
+	rl[api.ResourceName(rName)] = resource.MustParse(value)
+	return rl
+}
+
 func getResourceRequirements(requests, limits api.ResourceList) api.ResourceRequirements {
 	res := api.ResourceRequirements{}
 	res.Requests = requests
@@ -70,6 +75,12 @@ func TestGetPodQOS(t *testing.T) {
 			}),
 			expected: Guaranteed,
 		},
+		{
+			pod: newPod("guaranteed-with-gpu", []api.Container{
+				newContainer("guaranteed", getResourceList("100m", "100Mi"), addResource("nvidia-gpu", "2", getResourceList("100m", "100Mi"))),
+			}),
+			expected: Guaranteed,
+		},
 		{
 			pod: newPod("guaranteed-guaranteed", []api.Container{
 				newContainer("guaranteed", getResourceList("100m", "100Mi"), getResourceList("100m", "100Mi")),
@@ -77,6 +88,13 @@ func TestGetPodQOS(t *testing.T) {
 			}),
 			expected: Guaranteed,
 		},
+		{
+			pod: newPod("guaranteed-guaranteed-with-gpu", []api.Container{
+				newContainer("guaranteed", getResourceList("100m", "100Mi"), addResource("nvidia-gpu", "2", getResourceList("100m", "100Mi"))),
+				newContainer("guaranteed", getResourceList("100m", "100Mi"), getResourceList("100m", "100Mi")),
+			}),
+			expected: Guaranteed,
+		},
 		{
 			pod: newPod("best-effort-best-effort", []api.Container{
 				newContainer("best-effort", getResourceList("", ""), getResourceList("", "")),
@@ -85,21 +103,28 @@ func TestGetPodQOS(t *testing.T) {
 			expected: BestEffort,
 		},
 		{
-			pod: newPod("best-effort", []api.Container{
+			pod: newPod("best-effort-best-effort-with-gpu", []api.Container{
+				newContainer("best-effort", getResourceList("", ""), addResource("nvidia-gpu", "2", getResourceList("", ""))),
 				newContainer("best-effort", getResourceList("", ""), getResourceList("", "")),
 			}),
 			expected: BestEffort,
 		},
+		{
+			pod: newPod("best-effort-with-gpu", []api.Container{
+				newContainer("best-effort", getResourceList("", ""), addResource("nvidia-gpu", "2", getResourceList("", ""))),
+			}),
+			expected: BestEffort,
+		},
 		{
 			pod: newPod("best-effort-burstable", []api.Container{
-				newContainer("best-effort", getResourceList("", ""), getResourceList("", "")),
+				newContainer("best-effort", getResourceList("", ""), addResource("nvidia-gpu", "2", getResourceList("", ""))),
 				newContainer("burstable", getResourceList("1", ""), getResourceList("2", "")),
 			}),
 			expected: Burstable,
 		},
 		{
 			pod: newPod("best-effort-guaranteed", []api.Container{
-				newContainer("best-effort", getResourceList("", ""), getResourceList("", "")),
+				newContainer("best-effort", getResourceList("", ""), addResource("nvidia-gpu", "2", getResourceList("", ""))),
 				newContainer("guaranteed", getResourceList("10m", "100Mi"), getResourceList("10m", "100Mi")),
 			}),
 			expected: Burstable,
@@ -118,21 +143,21 @@ func TestGetPodQOS(t *testing.T) {
 			expected: Burstable,
 		},
 		{
-			pod: newPod("burstable", []api.Container{
+			pod: newPod("burstable-1", []api.Container{
 				newContainer("burstable", getResourceList("10m", "100Mi"), getResourceList("100m", "200Mi")),
 			}),
 			expected: Burstable,
 		},
 		{
-			pod: newPod("burstable", []api.Container{
+			pod: newPod("burstable-2", []api.Container{
-				newContainer("burstable", getResourceList("0", "0"), getResourceList("100m", "200Mi")),
+				newContainer("burstable", getResourceList("0", "0"), addResource("nvidia-gpu", "2", getResourceList("100m", "200Mi"))),
 			}),
 			expected: Burstable,
 		},
 	}
-	for _, testCase := range testCases {
+	for id, testCase := range testCases {
 		if actual := GetPodQOS(testCase.pod); testCase.expected != actual {
-			t.Errorf("invalid qos pod %s, expected: %s, actual: %s", testCase.pod.Name, testCase.expected, actual)
+			t.Errorf("[%d]: invalid qos pod %s, expected: %s, actual: %s", id, testCase.pod.Name, testCase.expected, actual)
 		}
 	}
 }