Merge pull request #68683 from chrisohaver/trackdns

cluster/addons/dns/coredns/coredns.yaml.base

@@ -112,7 +112,7 @@ spec:
         imagePullPolicy: IfNotPresent
         resources:
           limits:
-            memory: 170Mi
+            memory: 1000Mi
           requests:
             cpu: 100m
             memory: 70Mi

cluster/addons/dns/coredns/coredns.yaml.in

@@ -112,7 +112,7 @@ spec:
         imagePullPolicy: IfNotPresent
         resources:
           limits:
-            memory: 170Mi
+            memory: 1000Mi
           requests:
             cpu: 100m
             memory: 70Mi

cluster/addons/dns/coredns/coredns.yaml.sed

@@ -112,7 +112,7 @@ spec:
         imagePullPolicy: IfNotPresent
         resources:
           limits:
-            memory: 170Mi
+            memory: 1000Mi
           requests:
             cpu: 100m
             memory: 70Mi

cluster/gce/config-default.sh

@@ -259,7 +259,7 @@ fi
 
 # Optional: Install cluster DNS.
 # Set CLUSTER_DNS_CORE_DNS to 'true' to install CoreDNS instead of kube-dns.
-CLUSTER_DNS_CORE_DNS="${CLUSTER_DNS_CORE_DNS:-false}"
+CLUSTER_DNS_CORE_DNS="${CLUSTER_DNS_CORE_DNS:-true}"
 ENABLE_CLUSTER_DNS="${KUBE_ENABLE_CLUSTER_DNS:-true}"
 DNS_SERVER_IP="${KUBE_DNS_SERVER_IP:-10.0.0.10}"
 DNS_DOMAIN="${KUBE_DNS_DOMAIN:-cluster.local}"

cluster/gce/config-test.sh

@@ -266,7 +266,7 @@ fi
 
 # Optional: Install cluster DNS.
 # Set CLUSTER_DNS_CORE_DNS to 'true' to install CoreDNS instead of kube-dns.
-CLUSTER_DNS_CORE_DNS="${CLUSTER_DNS_CORE_DNS:-false}"
+CLUSTER_DNS_CORE_DNS="${CLUSTER_DNS_CORE_DNS:-true}"
 ENABLE_CLUSTER_DNS="${KUBE_ENABLE_CLUSTER_DNS:-true}"
 DNS_SERVER_IP="10.0.0.10"
 DNS_DOMAIN="cluster.local"

test/e2e/framework/framework.go

@@ -251,9 +251,18 @@ func (f *Framework) BeforeEach() {
 
 	if TestContext.GatherKubeSystemResourceUsageData != "false" && TestContext.GatherKubeSystemResourceUsageData != "none" {
 		var err error
+		var nodeMode NodesSet
+		switch TestContext.GatherKubeSystemResourceUsageData {
+		case "master":
+			nodeMode = MasterAndDNSNodes
+		case "masteranddns":
+			nodeMode = MasterAndDNSNodes
+		default:
+			nodeMode = AllNodes
+		}
 		f.gatherer, err = NewResourceUsageGatherer(f.ClientSet, ResourceGathererOptions{
-			InKubemark:                  ProviderIs("kubemark"),
-			MasterOnly:                  TestContext.GatherKubeSystemResourceUsageData == "master",
+			InKubemark: ProviderIs("kubemark"),
+			Nodes:      nodeMode,
 			ResourceDataGatheringPeriod: 60 * time.Second,
 			ProbeDuration:               15 * time.Second,
 			PrintVerboseLogs:            false,

test/e2e/framework/resource_usage_gatherer.go

@@ -202,12 +202,20 @@ type ContainerResourceGatherer struct {
 
 type ResourceGathererOptions struct {
 	InKubemark                  bool
-	MasterOnly                  bool
+	Nodes                       NodesSet
 	ResourceDataGatheringPeriod time.Duration
 	ProbeDuration               time.Duration
 	PrintVerboseLogs            bool
 }
 
+type NodesSet int
+
+const (
+	AllNodes          NodesSet = 0 // All containers on all nodes
+	MasterNodes       NodesSet = 1 // All containers on Master nodes only
+	MasterAndDNSNodes NodesSet = 2 // All containers on Master nodes and DNS containers on other nodes
+)
+
 func NewResourceUsageGatherer(c clientset.Interface, options ResourceGathererOptions, pods *v1.PodList) (*ContainerResourceGatherer, error) {
 	g := ContainerResourceGatherer{
 		client:       c,
@@ -237,13 +245,23 @@ func NewResourceUsageGatherer(c clientset.Interface, options ResourceGathererOpt
 				return nil, err
 			}
 		}
+		dnsNodes := make(map[string]bool)
 		for _, pod := range pods.Items {
+			if (options.Nodes == MasterNodes) && !system.IsMasterNode(pod.Spec.NodeName) {
+				continue
+			}
+			if (options.Nodes == MasterAndDNSNodes) && !system.IsMasterNode(pod.Spec.NodeName) && pod.Labels["k8s-app"] != "kube-dns" {
+				continue
+			}
 			for _, container := range pod.Status.InitContainerStatuses {
 				g.containerIDs = append(g.containerIDs, container.Name)
 			}
 			for _, container := range pod.Status.ContainerStatuses {
 				g.containerIDs = append(g.containerIDs, container.Name)
 			}
+			if options.Nodes == MasterAndDNSNodes {
+				dnsNodes[pod.Spec.NodeName] = true
+			}
 		}
 		nodeList, err := c.CoreV1().Nodes().List(metav1.ListOptions{})
 		if err != nil {
@@ -252,7 +270,7 @@ func NewResourceUsageGatherer(c clientset.Interface, options ResourceGathererOpt
 		}
 
 		for _, node := range nodeList.Items {
-			if !options.MasterOnly || system.IsMasterNode(node.Name) {
+			if options.Nodes == AllNodes || system.IsMasterNode(node.Name) || dnsNodes[node.Name] {
 				g.workerWg.Add(1)
 				g.workers = append(g.workers, resourceGatherWorker{
 					c:                           c,
@@ -266,7 +284,7 @@ func NewResourceUsageGatherer(c clientset.Interface, options ResourceGathererOpt
 					probeDuration:               options.ProbeDuration,
 					printVerboseLogs:            options.PrintVerboseLogs,
 				})
-				if options.MasterOnly {
+				if options.Nodes == MasterNodes {
 					break
 				}
 			}

test/e2e/scheduling/nvidia-gpus.go

@@ -137,7 +137,7 @@ func SetupNVIDIAGPUNode(f *framework.Framework, setupResourceGatherer bool) *fra
 	var rsgather *framework.ContainerResourceGatherer
 	if setupResourceGatherer {
 		framework.Logf("Starting ResourceUsageGather for the created DaemonSet pods.")
-		rsgather, err = framework.NewResourceUsageGatherer(f.ClientSet, framework.ResourceGathererOptions{InKubemark: false, MasterOnly: false, ResourceDataGatheringPeriod: 2 * time.Second, ProbeDuration: 2 * time.Second, PrintVerboseLogs: true}, pods)
+		rsgather, err = framework.NewResourceUsageGatherer(f.ClientSet, framework.ResourceGathererOptions{InKubemark: false, Nodes: framework.AllNodes, ResourceDataGatheringPeriod: 2 * time.Second, ProbeDuration: 2 * time.Second, PrintVerboseLogs: true}, pods)
 		framework.ExpectNoError(err, "creating ResourceUsageGather for the daemonset pods")
 		go rsgather.StartGatheringData()
 	}