Add hooks for cluster health detection

pkg/controller/node/deletion_utils.go → pkg/controller/node/controller_utils.go

@@ -22,7 +22,6 @@ import (
 	"time"
 
 	"k8s.io/kubernetes/pkg/api"
-	"k8s.io/kubernetes/pkg/api/unversioned"
 	"k8s.io/kubernetes/pkg/client/cache"
 	clientset "k8s.io/kubernetes/pkg/client/clientset_generated/internalclientset"
 	"k8s.io/kubernetes/pkg/client/record"
@@ -36,6 +35,20 @@ import (
 	"github.com/golang/glog"
 )
 
+// This function is expected to get a slice of NodeReadyConditions for all Nodes in a given zone.
+func ComputeZoneState(nodeReadyConditions []*api.NodeCondition) zoneState {
+	seenReady := false
+	for i := range nodeReadyConditions {
+		if nodeReadyConditions[i] != nil && nodeReadyConditions[i].Status == api.ConditionTrue {
+			seenReady = true
+		}
+	}
+	if seenReady {
+		return stateNormal
+	}
+	return stateFullSegmentation
+}
+
 // cleanupOrphanedPods deletes pods that are bound to nodes that don't
 // exist.
 func cleanupOrphanedPods(pods []*api.Pod, nodeStore cache.Store, forcefulDeletePodFunc func(*api.Pod) error) {
@@ -124,16 +137,6 @@ func forcefullyDeleteNode(kubeClient clientset.Interface, nodeName string, force
 	return nil
 }
 
-// forceUpdateAllProbeTimes bumps all observed timestamps in saved nodeStatuses to now. This makes
-// all eviction timer to reset.
-func forceUpdateAllProbeTimes(now unversioned.Time, statusData map[string]nodeStatusData) {
-	for k, v := range statusData {
-		v.probeTimestamp = now
-		v.readyTransitionTimestamp = now
-		statusData[k] = v
-	}
-}
-
 // maybeDeleteTerminatingPod non-gracefully deletes pods that are terminating
 // that should not be gracefully terminated.
 func (nc *NodeController) maybeDeleteTerminatingPod(obj interface{}, nodeStore cache.Store, forcefulDeletePodFunc func(*api.Pod) error) {

pkg/util/node/node.go

@@ -24,6 +24,7 @@ import (
 
 	"github.com/golang/glog"
 	"k8s.io/kubernetes/pkg/api"
+	"k8s.io/kubernetes/pkg/api/unversioned"
 )
 
 func GetHostname(hostnameOverride string) string {
@@ -59,3 +60,24 @@ func GetNodeHostIP(node *api.Node) (net.IP, error) {
 	}
 	return nil, fmt.Errorf("host IP unknown; known addresses: %v", addresses)
 }
+
+// Helper function that builds a string identifier that is unique per failure-zone
+// Returns empty-string for no zone
+func GetZoneKey(node *api.Node) string {
+	labels := node.Labels
+	if labels == nil {
+		return ""
+	}
+
+	region, _ := labels[unversioned.LabelZoneRegion]
+	failureDomain, _ := labels[unversioned.LabelZoneFailureDomain]
+
+	if region == "" && failureDomain == "" {
+		return ""
+	}
+
+	// We include the null character just in case region or failureDomain has a colon
+	// (We do assume there's no null characters in a region or failureDomain)
+	// As a nice side-benefit, the null character is not printed by fmt.Print or glog
+	return region + ":\x00:" + failureDomain
+}

plugin/pkg/scheduler/algorithm/priorities/selector_spreading.go

@@ -23,6 +23,7 @@ import (
 	"k8s.io/kubernetes/pkg/api"
 	"k8s.io/kubernetes/pkg/api/unversioned"
 	"k8s.io/kubernetes/pkg/labels"
+	utilnode "k8s.io/kubernetes/pkg/util/node"
 	utilruntime "k8s.io/kubernetes/pkg/util/runtime"
 	"k8s.io/kubernetes/plugin/pkg/scheduler/algorithm"
 	schedulerapi "k8s.io/kubernetes/plugin/pkg/scheduler/api"
@@ -54,27 +55,6 @@ func NewSelectorSpreadPriority(podLister algorithm.PodLister, serviceLister algo
 	return selectorSpread.CalculateSpreadPriority
 }
 
-// Helper function that builds a string identifier that is unique per failure-zone
-// Returns empty-string for no zone
-func getZoneKey(node *api.Node) string {
-	labels := node.Labels
-	if labels == nil {
-		return ""
-	}
-
-	region, _ := labels[unversioned.LabelZoneRegion]
-	failureDomain, _ := labels[unversioned.LabelZoneFailureDomain]
-
-	if region == "" && failureDomain == "" {
-		return ""
-	}
-
-	// We include the null character just in case region or failureDomain has a colon
-	// (We do assume there's no null characters in a region or failureDomain)
-	// As a nice side-benefit, the null character is not printed by fmt.Print or glog
-	return region + ":\x00:" + failureDomain
-}
-
 // CalculateSpreadPriority spreads pods across hosts and zones, considering pods belonging to the same service or replication controller.
 // When a pod is scheduled, it looks for services or RCs that match the pod, then finds existing pods that match those selectors.
 // It favors nodes that have fewer existing matching pods.
@@ -189,7 +169,7 @@ func (s *SelectorSpread) CalculateSpreadPriority(pod *api.Pod, nodeNameToInfo ma
 			continue
 		}
 
-		zoneId := getZoneKey(node)
+		zoneId := utilnode.GetZoneKey(node)
 		if zoneId == "" {
 			continue
 		}
@@ -220,7 +200,7 @@ func (s *SelectorSpread) CalculateSpreadPriority(pod *api.Pod, nodeNameToInfo ma
 
 		// If there is zone information present, incorporate it
 		if haveZones {
-			zoneId := getZoneKey(node)
+			zoneId := utilnode.GetZoneKey(node)
 			if zoneId != "" {
 				zoneScore := maxPriority * (float32(maxCountByZone-countsByZone[zoneId]) / float32(maxCountByZone))
 				fScore = (fScore * (1.0 - zoneWeighting)) + (zoneWeighting * zoneScore)