Merge pull request #60007 from k82cn/k8s_54313_1

Automatic merge from submit-queue (batch tested with PRs 59367, 60007). If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.

Do not schedule pod to the node under PID pressure.

Signed-off-by: Da K. Ma <klaus1982.cn@gmail.com>

**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
part of #54313 

**Release note**:
```release-note
Added CheckNodePIDPressurePredicate to checks if a pod can be scheduled on
a node reporting pid pressure condition.
```

pkg/scheduler/algorithm/predicates/error.go

@@ -61,6 +61,8 @@ var (
 	ErrNodeUnderMemoryPressure = newPredicateFailureError("NodeUnderMemoryPressure", "node(s) had memory pressure")
 	// ErrNodeUnderDiskPressure is used for NodeUnderDiskPressure predicate error.
 	ErrNodeUnderDiskPressure = newPredicateFailureError("NodeUnderDiskPressure", "node(s) had disk pressure")
+	// ErrNodeUnderPIDPressure is used for NodeUnderPIDPressure predicate error.
+	ErrNodeUnderPIDPressure = newPredicateFailureError("NodeUnderPIDPressure", "node(s) had pid pressure")
 	// ErrNodeOutOfDisk is used for NodeOutOfDisk predicate error.
 	ErrNodeOutOfDisk = newPredicateFailureError("NodeOutOfDisk", "node(s) were out of disk space")
 	// ErrNodeNotReady is used for NodeNotReady predicate error.

pkg/scheduler/algorithm/predicates/predicates.go

@@ -88,6 +88,8 @@ const (
 	CheckNodeMemoryPressurePred = "CheckNodeMemoryPressure"
 	// CheckNodeDiskPressurePred defines the name of predicate CheckNodeDiskPressure.
 	CheckNodeDiskPressurePred = "CheckNodeDiskPressure"
+	// CheckNodePIDPressurePred defines the name of predicate CheckNodePIDPressure.
+	CheckNodePIDPressurePred = "CheckNodePIDPressure"
 
 	// DefaultMaxEBSVolumes is the limit for volumes attached to an instance.
 	// Amazon recommends no more than 40; the system root volume uses at least one.
@@ -132,7 +134,7 @@ var (
 		PodToleratesNodeTaintsPred, PodToleratesNodeNoExecuteTaintsPred, CheckNodeLabelPresencePred,
 		CheckServiceAffinityPred, MaxEBSVolumeCountPred, MaxGCEPDVolumeCountPred,
 		MaxAzureDiskVolumeCountPred, CheckVolumeBindingPred, NoVolumeZoneConflictPred,
-		CheckNodeMemoryPressurePred, CheckNodeDiskPressurePred, MatchInterPodAffinityPred}
+		CheckNodeMemoryPressurePred, CheckNodePIDPressurePred, CheckNodeDiskPressurePred, MatchInterPodAffinityPred}
 )
 
 // NodeInfo interface represents anything that can get node object from node ID.
@@ -1591,6 +1593,16 @@ func CheckNodeDiskPressurePredicate(pod *v1.Pod, meta algorithm.PredicateMetadat
 	return true, nil, nil
 }
 
+// CheckNodePIDPressurePredicate checks if a pod can be scheduled on a node
+// reporting pid pressure condition.
+func CheckNodePIDPressurePredicate(pod *v1.Pod, meta algorithm.PredicateMetadata, nodeInfo *schedulercache.NodeInfo) (bool, []algorithm.PredicateFailureReason, error) {
+	// check if node is under pid pressure
+	if nodeInfo.PIDPressureCondition() == v1.ConditionTrue {
+		return false, []algorithm.PredicateFailureReason{ErrNodeUnderPIDPressure}, nil
+	}
+	return true, nil, nil
+}
+
 // CheckNodeConditionPredicate checks if a pod can be scheduled on a node reporting out of disk,
 // network unavailable and not ready condition. Only node conditions are accounted in this predicate.
 func CheckNodeConditionPredicate(pod *v1.Pod, meta algorithm.PredicateMetadata, nodeInfo *schedulercache.NodeInfo) (bool, []algorithm.PredicateFailureReason, error) {

pkg/scheduler/algorithmprovider/defaults/compatibility_test.go

@@ -480,7 +480,6 @@ func TestCompatibility_v1_Scheduler(t *testing.T) {
 			{"name": "CheckVolumeBinding"},
 			{"name": "TestServiceAffinity", "argument": {"serviceAffinity" : {"labels" : ["region"]}}},
 			{"name": "TestLabelsPresence",  "argument": {"labelsPresence"  : {"labels" : ["foo"], "presence":true}}}
-
 		  ],"priorities": [
 			{"name": "EqualPriority",   "weight": 2},
 			{"name": "ImageLocalityPriority",   "weight": 2},
@@ -529,6 +528,81 @@ func TestCompatibility_v1_Scheduler(t *testing.T) {
 				},
 			},
 		},
+		// Do not change this JSON after the corresponding release has been tagged.
+		// A failure indicates backwards compatibility with the specified release was broken.
+		"1.10": {
+			JSON: `{
+		  "kind": "Policy",
+		  "apiVersion": "v1",
+		  "predicates": [
+			{"name": "MatchNodeSelector"},
+			{"name": "PodFitsResources"},
+			{"name": "PodFitsHostPorts"},
+			{"name": "HostName"},
+			{"name": "NoDiskConflict"},
+			{"name": "NoVolumeZoneConflict"},
+			{"name": "PodToleratesNodeTaints"},
+			{"name": "CheckNodeMemoryPressure"},
+			{"name": "CheckNodeDiskPressure"},
+			{"name": "CheckNodePIDPressure"},
+			{"name": "CheckNodeCondition"},
+			{"name": "MaxEBSVolumeCount"},
+			{"name": "MaxGCEPDVolumeCount"},
+			{"name": "MaxAzureDiskVolumeCount"},
+			{"name": "MatchInterPodAffinity"},
+			{"name": "GeneralPredicates"},
+			{"name": "CheckVolumeBinding"},
+			{"name": "TestServiceAffinity", "argument": {"serviceAffinity" : {"labels" : ["region"]}}},
+			{"name": "TestLabelsPresence",  "argument": {"labelsPresence"  : {"labels" : ["foo"], "presence":true}}}
+		  ],"priorities": [
+			{"name": "EqualPriority",   "weight": 2},
+			{"name": "ImageLocalityPriority",   "weight": 2},
+			{"name": "LeastRequestedPriority",   "weight": 2},
+			{"name": "BalancedResourceAllocation",   "weight": 2},
+			{"name": "SelectorSpreadPriority",   "weight": 2},
+			{"name": "NodePreferAvoidPodsPriority",   "weight": 2},
+			{"name": "NodeAffinityPriority",   "weight": 2},
+			{"name": "TaintTolerationPriority",   "weight": 2},
+			{"name": "InterPodAffinityPriority",   "weight": 2},
+			{"name": "MostRequestedPriority",   "weight": 2}
+		  ]
+		}`,
+			ExpectedPolicy: schedulerapi.Policy{
+				Predicates: []schedulerapi.PredicatePolicy{
+					{Name: "MatchNodeSelector"},
+					{Name: "PodFitsResources"},
+					{Name: "PodFitsHostPorts"},
+					{Name: "HostName"},
+					{Name: "NoDiskConflict"},
+					{Name: "NoVolumeZoneConflict"},
+					{Name: "PodToleratesNodeTaints"},
+					{Name: "CheckNodeMemoryPressure"},
+					{Name: "CheckNodeDiskPressure"},
+					{Name: "CheckNodePIDPressure"},
+					{Name: "CheckNodeCondition"},
+					{Name: "MaxEBSVolumeCount"},
+					{Name: "MaxGCEPDVolumeCount"},
+					{Name: "MaxAzureDiskVolumeCount"},
+					{Name: "MatchInterPodAffinity"},
+					{Name: "GeneralPredicates"},
+					{Name: "CheckVolumeBinding"},
+					{Name: "TestServiceAffinity", Argument: &schedulerapi.PredicateArgument{ServiceAffinity: &schedulerapi.ServiceAffinity{Labels: []string{"region"}}}},
+					{Name: "TestLabelsPresence", Argument: &schedulerapi.PredicateArgument{LabelsPresence: &schedulerapi.LabelsPresence{Labels: []string{"foo"}, Presence: true}}},
+				},
+				Priorities: []schedulerapi.PriorityPolicy{
+					{Name: "EqualPriority", Weight: 2},
+					{Name: "ImageLocalityPriority", Weight: 2},
+					{Name: "LeastRequestedPriority", Weight: 2},
+					{Name: "BalancedResourceAllocation", Weight: 2},
+					{Name: "SelectorSpreadPriority", Weight: 2},
+					{Name: "NodePreferAvoidPodsPriority", Weight: 2},
+					{Name: "NodeAffinityPriority", Weight: 2},
+					{Name: "TaintTolerationPriority", Weight: 2},
+					{Name: "InterPodAffinityPriority", Weight: 2},
+					{Name: "MostRequestedPriority", Weight: 2},
+				},
+			},
+		},
 	}
 
 	registeredPredicates := sets.NewString(factory.ListRegisteredFitPredicates()...)

pkg/scheduler/algorithmprovider/defaults/defaults.go

@@ -17,6 +17,8 @@ limitations under the License.
 package defaults
 
 import (
+	"github.com/golang/glog"
+
 	"k8s.io/apimachinery/pkg/util/sets"
 	utilfeature "k8s.io/apiserver/pkg/util/feature"
 
@@ -26,8 +28,6 @@ import (
 	"k8s.io/kubernetes/pkg/scheduler/algorithm/priorities"
 	"k8s.io/kubernetes/pkg/scheduler/core"
 	"k8s.io/kubernetes/pkg/scheduler/factory"
-
-	"github.com/golang/glog"
 )
 
 const (
@@ -160,6 +160,9 @@ func defaultPredicates() sets.String {
 		// Fit is determined by node disk pressure condition.
 		factory.RegisterFitPredicate(predicates.CheckNodeDiskPressurePred, predicates.CheckNodeDiskPressurePredicate),
 
+		// Fit is determined by node pid pressure condition.
+		factory.RegisterFitPredicate(predicates.CheckNodePIDPressurePred, predicates.CheckNodePIDPressurePredicate),
+
 		// Fit is determined by node conditions: not ready, network unavailable or out of disk.
 		factory.RegisterMandatoryFitPredicate(predicates.CheckNodeConditionPred, predicates.CheckNodeConditionPredicate),
 
@@ -179,10 +182,12 @@ func defaultPredicates() sets.String {
 // ApplyFeatureGates applies algorithm by feature gates.
 func ApplyFeatureGates() {
 	if utilfeature.DefaultFeatureGate.Enabled(features.TaintNodesByCondition) {
-		// Remove "CheckNodeCondition", "CheckNodeMemoryPressure" and "CheckNodeDiskPressure" predicates
+		// Remove "CheckNodeCondition", "CheckNodeMemoryPressure", "CheckNodePIDPressurePred"
+		// and "CheckNodeDiskPressure" predicates
 		factory.RemoveFitPredicate(predicates.CheckNodeConditionPred)
 		factory.RemoveFitPredicate(predicates.CheckNodeMemoryPressurePred)
 		factory.RemoveFitPredicate(predicates.CheckNodeDiskPressurePred)
+		factory.RemoveFitPredicate(predicates.CheckNodePIDPressurePred)
 		// Remove key "CheckNodeCondition", "CheckNodeMemoryPressure" and "CheckNodeDiskPressure"
 		// from ALL algorithm provider
 		// The key will be removed from all providers which in algorithmProviderMap[]
@@ -190,6 +195,7 @@ func ApplyFeatureGates() {
 		factory.RemovePredicateKeyFromAlgorithmProviderMap(predicates.CheckNodeConditionPred)
 		factory.RemovePredicateKeyFromAlgorithmProviderMap(predicates.CheckNodeMemoryPressurePred)
 		factory.RemovePredicateKeyFromAlgorithmProviderMap(predicates.CheckNodeDiskPressurePred)
+		factory.RemovePredicateKeyFromAlgorithmProviderMap(predicates.CheckNodePIDPressurePred)
 
 		// Fit is determined based on whether a pod can tolerate all of the node's taints
 		factory.RegisterMandatoryFitPredicate(predicates.PodToleratesNodeTaintsPred, predicates.PodToleratesNodeTaints)

pkg/scheduler/algorithmprovider/defaults/defaults_test.go

@@ -76,6 +76,7 @@ func TestDefaultPredicates(t *testing.T) {
 		"GeneralPredicates",
 		"CheckNodeMemoryPressure",
 		"CheckNodeDiskPressure",
+		"CheckNodePIDPressure",
 		"CheckNodeCondition",
 		"PodToleratesNodeTaints",
 		predicates.CheckVolumeBindingPred,

pkg/scheduler/core/generic_scheduler.go

@@ -25,6 +25,8 @@ import (
 	"sync/atomic"
 	"time"
 
+	"github.com/golang/glog"
+
 	"k8s.io/api/core/v1"
 	policy "k8s.io/api/policy/v1beta1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@@ -40,8 +42,6 @@ import (
 	"k8s.io/kubernetes/pkg/scheduler/schedulercache"
 	"k8s.io/kubernetes/pkg/scheduler/util"
 	"k8s.io/kubernetes/pkg/scheduler/volumebinder"
-
-	"github.com/golang/glog"
 )
 
 // FailedPredicateMap declares a map[string][]algorithm.PredicateFailureReason type.

pkg/scheduler/core/scheduling_queue.go

@@ -29,8 +29,11 @@ package core
 import (
 	"container/heap"
 	"fmt"
+	"reflect"
 	"sync"
 
+	"github.com/golang/glog"
+
 	"k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/client-go/tools/cache"
@@ -38,10 +41,6 @@ import (
 	"k8s.io/kubernetes/pkg/scheduler/algorithm/predicates"
 	priorityutil "k8s.io/kubernetes/pkg/scheduler/algorithm/priorities/util"
 	"k8s.io/kubernetes/pkg/scheduler/util"
-
-	"reflect"
-
-	"github.com/golang/glog"
 )
 
 // SchedulingQueue is an interface for a queue to store pods waiting to be scheduled.

pkg/scheduler/factory/factory.go

@@ -837,6 +837,9 @@ func (c *configFactory) invalidateCachedPredicatesOnNodeUpdate(newNode *v1.Node,
 			if oldConditions[v1.NodeDiskPressure] != newConditions[v1.NodeDiskPressure] {
 				invalidPredicates.Insert(predicates.CheckNodeDiskPressurePred)
 			}
+			if oldConditions[v1.NodePIDPressure] != newConditions[v1.NodePIDPressure] {
+				invalidPredicates.Insert(predicates.CheckNodePIDPressurePred)
+			}
 			if oldConditions[v1.NodeReady] != newConditions[v1.NodeReady] ||
 				oldConditions[v1.NodeOutOfDisk] != newConditions[v1.NodeOutOfDisk] ||
 				oldConditions[v1.NodeNetworkUnavailable] != newConditions[v1.NodeNetworkUnavailable] {

pkg/scheduler/schedulercache/node_info.go

@@ -62,6 +62,7 @@ type NodeInfo struct {
 	// Cached conditions of node for faster lookup.
 	memoryPressureCondition v1.ConditionStatus
 	diskPressureCondition   v1.ConditionStatus
+	pidPressureCondition    v1.ConditionStatus
 
 	// Whenever NodeInfo changes, generation is bumped.
 	// This is used to avoid cloning it if the object didn't change.
@@ -284,6 +285,14 @@ func (n *NodeInfo) DiskPressureCondition() v1.ConditionStatus {
 	return n.diskPressureCondition
 }
 
+// PIDPressureCondition returns the pid pressure condition status on this node.
+func (n *NodeInfo) PIDPressureCondition() v1.ConditionStatus {
+	if n == nil {
+		return v1.ConditionUnknown
+	}
+	return n.pidPressureCondition
+}
+
 // RequestedResource returns aggregated resource request of pods on this node.
 func (n *NodeInfo) RequestedResource() Resource {
 	if n == nil {
@@ -324,6 +333,7 @@ func (n *NodeInfo) Clone() *NodeInfo {
 		TransientInfo:           n.TransientInfo,
 		memoryPressureCondition: n.memoryPressureCondition,
 		diskPressureCondition:   n.diskPressureCondition,
+		pidPressureCondition:    n.pidPressureCondition,
 		usedPorts:               make(util.HostPortInfo),
 		generation:              n.generation,
 	}
@@ -482,6 +492,8 @@ func (n *NodeInfo) SetNode(node *v1.Node) error {
 			n.memoryPressureCondition = cond.Status
 		case v1.NodeDiskPressure:
 			n.diskPressureCondition = cond.Status
+		case v1.NodePIDPressure:
+			n.pidPressureCondition = cond.Status
 		default:
 			// We ignore other conditions.
 		}
@@ -502,6 +514,7 @@ func (n *NodeInfo) RemoveNode(node *v1.Node) error {
 	n.taints, n.taintsErr = nil, nil
 	n.memoryPressureCondition = v1.ConditionUnknown
 	n.diskPressureCondition = v1.ConditionUnknown
+	n.pidPressureCondition = v1.ConditionUnknown
 	n.generation++
 	return nil
 }

test/integration/scheduler/predicates_test.go

@@ -870,3 +870,53 @@ func TestInterPodAffinity(t *testing.T) {
 		}
 	}
 }
+
+// TestNodePIDPressure verifies that scheduler's CheckNodePIDPressurePredicate predicate
+// functions works correctly.
+func TestNodePIDPressure(t *testing.T) {
+	context := initTest(t, "node-pid-pressure")
+	defer cleanupTest(t, context)
+	// Add a node.
+	node, err := createNode(context.clientSet, "testnode", nil)
+	if err != nil {
+		t.Fatalf("Cannot create node: %v", err)
+	}
+
+	cs := context.clientSet
+
+	// Adds PID pressure condition to the node.
+	node.Status.Conditions = []v1.NodeCondition{
+		{
+			Type:   v1.NodePIDPressure,
+			Status: v1.ConditionTrue,
+		},
+	}
+
+	// Update node condition.
+	err = updateNodeStatus(context.clientSet, node)
+	if err != nil {
+		t.Fatalf("Cannot update node: %v", err)
+	}
+
+	// Creats test pod.
+	testPod := &v1.Pod{
+		ObjectMeta: metav1.ObjectMeta{Name: "pidpressure-fake-name"},
+		Spec: v1.PodSpec{
+			Containers: []v1.Container{
+				{Name: "container", Image: imageutils.GetPauseImageName()},
+			},
+		},
+	}
+
+	testPod, err = cs.CoreV1().Pods(context.ns.Name).Create(testPod)
+	if err != nil {
+		t.Fatalf("Test Failed: error: %v, while creating pod", err)
+	}
+
+	err = waitForPodUnschedulable(cs, testPod)
+	if err != nil {
+		t.Errorf("Test Failed: error, %v, while waiting for scheduled", err)
+	}
+
+	cleanupPods(cs, t, []*v1.Pod{testPod})
+}

test/integration/scheduler/scheduler_test.go

@@ -134,6 +134,7 @@ func TestSchedulerCreationFromConfigMap(t *testing.T) {
 				"CheckNodeCondition", // mandatory predicate
 				"CheckNodeDiskPressure",
 				"CheckNodeMemoryPressure",
+				"CheckNodePIDPressure",
 				"CheckVolumeBinding",
 				"GeneralPredicates",
 				"MatchInterPodAffinity",

test/integration/scheduler/taint_test.go

@@ -45,6 +45,12 @@ import (
 //   2. NodeController taints nodes by node condition
 //   3. Scheduler allows pod to tolerate node condition taints, e.g. network unavailable
 func TestTaintNodeByCondition(t *testing.T) {
+	enabled := utilfeature.DefaultFeatureGate.Enabled("TaintNodesByCondition")
+	defer func() {
+		if !enabled {
+			utilfeature.DefaultFeatureGate.Set("TaintNodesByCondition=False")
+		}
+	}()
 	// Enable TaintNodeByCondition
 	utilfeature.DefaultFeatureGate.Set("TaintNodesByCondition=True")
 

test/integration/scheduler/util.go

@@ -19,6 +19,7 @@ package scheduler
 import (
 	"fmt"
 	"net/http"
+	"net/http/httptest"
 	"testing"
 	"time"
 
@@ -48,8 +49,6 @@ import (
 	"k8s.io/kubernetes/pkg/scheduler/factory"
 	"k8s.io/kubernetes/test/integration/framework"
 	imageutils "k8s.io/kubernetes/test/utils/image"
-
-	"net/http/httptest"
 )
 
 type TestContext struct {
@@ -318,6 +317,12 @@ func createNode(cs clientset.Interface, name string, res *v1.ResourceList) (*v1.
 	return cs.CoreV1().Nodes().Create(n)
 }
 
+// updateNodeStatus updates the status of node.
+func updateNodeStatus(cs clientset.Interface, node *v1.Node) error {
+	_, err := cs.CoreV1().Nodes().UpdateStatus(node)
+	return err
+}
+
 // createNodes creates `numNodes` nodes. The created node names will be in the
 // form of "`prefix`-X" where X is an ordinal.
 func createNodes(cs clientset.Interface, prefix string, res *v1.ResourceList, numNodes int) ([]*v1.Node, error) {