Enable PodTolerateNodeTaints predicate in DaemonSet controller

pkg/controller/daemon/BUILD

@@ -27,6 +27,7 @@ go_library(
         "//pkg/client/listers/extensions/v1beta1:go_default_library",
         "//pkg/controller:go_default_library",
         "//pkg/util/metrics:go_default_library",
+        "//plugin/pkg/scheduler/algorithm:go_default_library",
         "//plugin/pkg/scheduler/algorithm/predicates:go_default_library",
         "//plugin/pkg/scheduler/schedulercache:go_default_library",
         "//vendor:github.com/golang/glog",

pkg/controller/daemon/daemoncontroller.go

@@ -45,6 +45,7 @@ import (
 	extensionslisters "k8s.io/kubernetes/pkg/client/listers/extensions/v1beta1"
 	"k8s.io/kubernetes/pkg/controller"
 	"k8s.io/kubernetes/pkg/util/metrics"
+	"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm"
 	"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm/predicates"
 	"k8s.io/kubernetes/plugin/pkg/scheduler/schedulercache"
 
@@ -779,13 +780,14 @@ func (dsc *DaemonSetsController) nodeShouldRunDaemonPod(node *v1.Node, ds *exten
 
 	nodeInfo := schedulercache.NewNodeInfo(pods...)
 	nodeInfo.SetNode(node)
-	_, reasons, err := predicates.GeneralPredicates(newPod, nil, nodeInfo)
+	_, reasons, err := daemonSetPredicates(newPod, nodeInfo)
 	if err != nil {
-		glog.Warningf("GeneralPredicates failed on ds '%s/%s' due to unexpected error: %v", ds.ObjectMeta.Namespace, ds.ObjectMeta.Name, err)
+		glog.Warningf("daemonSetPredicates failed on ds '%s/%s' due to unexpected error: %v", ds.ObjectMeta.Namespace, ds.ObjectMeta.Name, err)
 		return false, false, false, err
 	}
+
 	for _, r := range reasons {
-		glog.V(4).Infof("GeneralPredicates failed on ds '%s/%s' for reason: %v", ds.ObjectMeta.Namespace, ds.ObjectMeta.Name, r.GetReason())
+		glog.V(4).Infof("daemonSetPredicates failed on ds '%s/%s' for reason: %v", ds.ObjectMeta.Namespace, ds.ObjectMeta.Name, r.GetReason())
 		switch reason := r.(type) {
 		case *predicates.InsufficientResourceError:
 			dsc.eventRecorder.Eventf(ds, v1.EventTypeNormal, FailedPlacementReason, "failed to place pod on %q: %s", node.ObjectMeta.Name, reason.Error())
@@ -801,7 +803,9 @@ func (dsc *DaemonSetsController) nodeShouldRunDaemonPod(node *v1.Node, ds *exten
 				predicates.ErrNodeLabelPresenceViolated,
 				// this one is probably intentional since it's a workaround for not having
 				// pod hard anti affinity.
-				predicates.ErrPodNotFitsHostPorts:
+				predicates.ErrPodNotFitsHostPorts,
+				// DaemonSet is expected to respect taints and tolerations
+				predicates.ErrTaintsTolerationsNotMatch:
 				wantToRun, shouldSchedule, shouldContinueRunning = false, false, false
 			// unintentional
 			case
@@ -818,9 +822,9 @@ func (dsc *DaemonSetsController) nodeShouldRunDaemonPod(node *v1.Node, ds *exten
 			// unexpected
 			case
 				predicates.ErrPodAffinityNotMatch,
-				predicates.ErrServiceAffinityViolated,
+				predicates.ErrServiceAffinityViolated:
-				predicates.ErrTaintsTolerationsNotMatch:
+				glog.Warningf("unexpected predicate failure reason: %s", reason.GetReason())
-				return false, false, false, fmt.Errorf("unexpected reason: GeneralPredicates should not return reason %s", reason.GetReason())
+				return false, false, false, fmt.Errorf("unexpected reason: daemonSetPredicates should not return reason %s", reason.GetReason())
 			default:
 				glog.V(4).Infof("unknown predicate failure reason: %s", reason.GetReason())
 				wantToRun, shouldSchedule, shouldContinueRunning = false, false, false
@@ -834,6 +838,30 @@ func (dsc *DaemonSetsController) nodeShouldRunDaemonPod(node *v1.Node, ds *exten
 	return
 }
 
+// daemonSetPredicates checks if a DaemonSet's pod can be scheduled on a node using GeneralPredicates
+// and PodToleratesNodeTaints predicate
+func daemonSetPredicates(pod *v1.Pod, nodeInfo *schedulercache.NodeInfo) (bool, []algorithm.PredicateFailureReason, error) {
+	var predicateFails []algorithm.PredicateFailureReason
+
+	fit, reasons, err := predicates.GeneralPredicates(pod, nil, nodeInfo)
+	if err != nil {
+		return false, predicateFails, err
+	}
+	if !fit {
+		predicateFails = append(predicateFails, reasons...)
+	}
+
+	fit, reasons, err = predicates.PodToleratesNodeTaints(pod, nil, nodeInfo)
+	if err != nil {
+		return false, predicateFails, err
+	}
+	if !fit {
+		predicateFails = append(predicateFails, reasons...)
+	}
+
+	return len(predicateFails) == 0, predicateFails, nil
+}
+
 // byCreationTimestamp sorts a list by creation timestamp, using their names as a tie breaker.
 type byCreationTimestamp []*extensions.DaemonSet
 

pkg/controller/daemon/daemoncontroller_test.go

@@ -45,6 +45,15 @@ var (
 	alwaysReady           = func() bool { return true }
 )
 
+const (
+	noSchedule = `
+	[{
+		"key": "dedicated",
+		"value": "user1",
+		"effect": "NoSchedule"
+	}]`
+)
+
 func getKey(ds *extensions.DaemonSet, t *testing.T) string {
 	if key, err := controller.KeyFunc(ds); err != nil {
 		t.Errorf("Unexpected error getting key for ds %v: %v", ds.Name, err)
@@ -708,6 +717,63 @@ func TestDaemonKillFailedPods(t *testing.T) {
 	}
 }
 
+// DaemonSet should not launch a pod on a tainted node when the pod doesn't tolerate that taint.
+func TestTaintedNodeDaemonDoesNotLaunchUntoleratePod(t *testing.T) {
+	manager, podControl, _ := newTestController()
+
+	node := newNode("tainted", nil)
+	setNodeTaint(node, noSchedule)
+	manager.nodeStore.Add(node)
+
+	ds := newDaemonSet("untolerate")
+	manager.dsStore.Add(ds)
+
+	syncAndValidateDaemonSets(t, manager, ds, podControl, 0, 0)
+}
+
+// DaemonSet should launch a pod on a tainted node when the pod can tolerate that taint.
+func TestTaintedNodeDaemonLaunchesToleratePod(t *testing.T) {
+	manager, podControl, _ := newTestController()
+
+	node := newNode("tainted", nil)
+	setNodeTaint(node, noSchedule)
+	manager.nodeStore.Add(node)
+
+	ds := newDaemonSet("tolerate")
+	setDaemonSetToleration(ds, noSchedule)
+	manager.dsStore.Add(ds)
+
+	syncAndValidateDaemonSets(t, manager, ds, podControl, 1, 0)
+}
+
+// DaemonSet should launch a pod on an untainted node when the pod has tolerations.
+func TestNodeDaemonLaunchesToleratePod(t *testing.T) {
+	manager, podControl, _ := newTestController()
+
+	node := newNode("untainted", nil)
+	manager.nodeStore.Add(node)
+
+	ds := newDaemonSet("tolerate")
+	setDaemonSetToleration(ds, noSchedule)
+	manager.dsStore.Add(ds)
+
+	syncAndValidateDaemonSets(t, manager, ds, podControl, 1, 0)
+}
+
+func setNodeTaint(node *v1.Node, taint string) {
+	if node.ObjectMeta.Annotations == nil {
+		node.ObjectMeta.Annotations = make(map[string]string)
+	}
+	node.ObjectMeta.Annotations[v1.TaintsAnnotationKey] = taint
+}
+
+func setDaemonSetToleration(ds *extensions.DaemonSet, toleration string) {
+	if ds.Spec.Template.ObjectMeta.Annotations == nil {
+		ds.Spec.Template.ObjectMeta.Annotations = make(map[string]string)
+	}
+	ds.Spec.Template.ObjectMeta.Annotations[v1.TolerationsAnnotationKey] = toleration
+}
+
 func TestNodeShouldRunDaemonPod(t *testing.T) {
 	cases := []struct {
 		podsOnNode                                       []*v1.Pod