Merge pull request #25531 from ingvagabund/introduce-memory-pressure-to-scheduler

docs/devel/scheduler_algorithm.md

@@ -48,6 +48,7 @@ The purpose of filtering the nodes is to filter out the nodes that do not meet c
 - `MatchNodeSelector`: Check if the labels of the node match the labels specified in the Pod's `nodeSelector` field and, as of Kubernetes v1.2, also match the `scheduler.alpha.kubernetes.io/affinity` pod annotation if present. See [here](../user-guide/node-selection/) for more details on both.
 - `MaxEBSVolumeCount`: Ensure that the number of attached ElasticBlockStore volumes does not exceed a maximum value (by default, 39, since Amazon recommends a maximum of 40 with one of those 40 reserved for the root volume -- see [Amazon's documentation](http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/volume_limits.html#linux-specific-volume-limits)).  The maximum value can be controlled by setting the `KUBE_MAX_PD_VOLS` environment variable.
 - `MaxGCEPDVolumeCount`: Ensure that the number of attached GCE PersistentDisk volumes does not exceed a maximum value (by default, 16, which is the maximum GCE allows -- see [GCE's documentation](https://cloud.google.com/compute/docs/disks/persistent-disks#limits_for_predefined_machine_types)).  The maximum value can be controlled by setting the `KUBE_MAX_PD_VOLS` environment variable.
+- `CheckNodeMemoryPressure`: Check if a pod can be scheduled on a node reporting memory pressure condition. Currently, no ``BestEffort`` should be placed on a node under memory pressure as it gets automatically evicted by kubelet.
 
 The details of the above predicates can be found in [plugin/pkg/scheduler/algorithm/predicates/predicates.go](http://releases.k8s.io/HEAD/plugin/pkg/scheduler/algorithm/predicates/predicates.go). All predicates mentioned above can be used in combination to perform a sophisticated filtering policy. Kubernetes uses some, but not all, of these predicates by default. You can see which ones are used by default in [plugin/pkg/scheduler/algorithmprovider/defaults/defaults.go](http://releases.k8s.io/HEAD/plugin/pkg/scheduler/algorithmprovider/defaults/defaults.go).
 

plugin/pkg/scheduler/algorithm/predicates/error.go

@@ -38,6 +38,7 @@ var (
 	ErrNodeLabelPresenceViolated = newPredicateFailureError("CheckNodeLabelPresence")
 	ErrServiceAffinityViolated   = newPredicateFailureError("CheckServiceAffinity")
 	ErrMaxVolumeCountExceeded    = newPredicateFailureError("MaxVolumeCount")
+	ErrNodeUnderMemoryPressure   = newPredicateFailureError("NodeUnderMemoryPressure")
 	// ErrFakePredicate is used for test only. The fake predicates returning false also returns error
 	// as ErrFakePredicate.
 	ErrFakePredicate = newPredicateFailureError("FakePredicateError")

plugin/pkg/scheduler/algorithm/predicates/predicates.go

@@ -23,6 +23,7 @@ import (
 	"k8s.io/kubernetes/pkg/api"
 	"k8s.io/kubernetes/pkg/api/unversioned"
 	"k8s.io/kubernetes/pkg/client/cache"
+	qosutil "k8s.io/kubernetes/pkg/kubelet/qos/util"
 	"k8s.io/kubernetes/pkg/labels"
 	"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm"
 	priorityutil "k8s.io/kubernetes/plugin/pkg/scheduler/algorithm/priorities/util"
@@ -999,3 +1000,31 @@ func tolerationsToleratesTaints(tolerations []api.Toleration, taints []api.Taint
 
 	return true
 }
+
+// Determine if a pod is scheduled with best-effort QoS
+func isPodBestEffort(pod *api.Pod) bool {
+	return qosutil.GetPodQos(pod) == qosutil.BestEffort
+}
+
+// CheckNodeMemoryPressurePredicate checks if a pod can be scheduled on a node
+// reporting memory pressure condition.
+func CheckNodeMemoryPressurePredicate(pod *api.Pod, nodeInfo *schedulercache.NodeInfo) (bool, error) {
+	node := nodeInfo.Node()
+	if node == nil {
+		return false, fmt.Errorf("node not found")
+	}
+
+	// pod is not BestEffort pod
+	if !isPodBestEffort(pod) {
+		return true, nil
+	}
+
+	// is node under presure?
+	for _, cond := range node.Status.Conditions {
+		if cond.Type == api.NodeMemoryPressure && cond.Status == api.ConditionTrue {
+			return false, ErrNodeUnderMemoryPressure
+		}
+	}
+
+	return true, nil
+}

plugin/pkg/scheduler/algorithm/predicates/predicates_test.go

@@ -2641,3 +2641,111 @@ func TestPodToleratesTaints(t *testing.T) {
 		}
 	}
 }
+
+func makeEmptyNodeInfo(node *api.Node) *schedulercache.NodeInfo {
+	nodeInfo := schedulercache.NewNodeInfo()
+	nodeInfo.SetNode(node)
+	return nodeInfo
+}
+
+func TestPodSchedulesOnNodeWithMemoryPressureCondition(t *testing.T) {
+	// specify best-effort pod
+	bestEffortPod := &api.Pod{
+		Spec: api.PodSpec{
+			Containers: []api.Container{
+				{
+					Name:            "container",
+					Image:           "image",
+					ImagePullPolicy: "Always",
+					// no requirements -> best effort pod
+					Resources: api.ResourceRequirements{},
+				},
+			},
+		},
+	}
+
+	// specify non-best-effort pod
+	nonBestEffortPod := &api.Pod{
+		Spec: api.PodSpec{
+			Containers: []api.Container{
+				{
+					Name:            "container",
+					Image:           "image",
+					ImagePullPolicy: "Always",
+					// at least one requirement -> burstable pod
+					Resources: api.ResourceRequirements{
+						Requests: makeAllocatableResources(100, 100, 100, 100),
+					},
+				},
+			},
+		},
+	}
+
+	// specify a node with no memory pressure condition on
+	noMemoryPressureNode := &api.Node{
+		Status: api.NodeStatus{
+			Conditions: []api.NodeCondition{
+				{
+					Type:   "Ready",
+					Status: "True",
+				},
+			},
+		},
+	}
+
+	// specify a node with memory pressure condition on
+	memoryPressureNode := &api.Node{
+		Status: api.NodeStatus{
+			Conditions: []api.NodeCondition{
+				{
+					Type:   "MemoryPressure",
+					Status: "True",
+				},
+			},
+		},
+	}
+
+	tests := []struct {
+		pod      *api.Pod
+		nodeInfo *schedulercache.NodeInfo
+		fits     bool
+		name     string
+	}{
+		{
+			pod:      bestEffortPod,
+			nodeInfo: makeEmptyNodeInfo(noMemoryPressureNode),
+			fits:     true,
+			name:     "best-effort pod schedulable on node without memory pressure condition on",
+		},
+		{
+			pod:      bestEffortPod,
+			nodeInfo: makeEmptyNodeInfo(memoryPressureNode),
+			fits:     false,
+			name:     "best-effort pod not schedulable on node with memory pressure condition on",
+		},
+		{
+			pod:      nonBestEffortPod,
+			nodeInfo: makeEmptyNodeInfo(memoryPressureNode),
+			fits:     true,
+			name:     "non best-effort pod schedulable on node with memory pressure condition on",
+		},
+		{
+			pod:      nonBestEffortPod,
+			nodeInfo: makeEmptyNodeInfo(noMemoryPressureNode),
+			fits:     true,
+			name:     "non best-effort pod schedulable on node without memory pressure condition on",
+		},
+	}
+
+	for _, test := range tests {
+		fits, err := CheckNodeMemoryPressurePredicate(test.pod, test.nodeInfo)
+		if fits != test.fits {
+			t.Errorf("%s: expected %v got %v", test.name, test.fits, fits)
+		}
+
+		if err != nil && err != ErrNodeUnderMemoryPressure {
+			t.Errorf("%s: unexpected error: %v", test.name, err)
+			continue
+		}
+	}
+}

plugin/pkg/scheduler/algorithmprovider/defaults/defaults.go

@@ -153,6 +153,9 @@ func defaultPredicates() sets.String {
 				return predicates.NewTolerationMatchPredicate(args.NodeInfo)
 			},
 		),
+
+		// Fit is determined by node memory pressure condition.
+		factory.RegisterFitPredicate("CheckNodeMemoryPressure", predicates.CheckNodeMemoryPressurePredicate),
 	)
 }