Extend scale test & improve controller-runtime dashboard

Signed-off-by: Stefan Büringer [email protected]
kubernetes-sigs · Jan 21, 2025 · 3f02ca7 · 3f02ca7
1 parent 1250b13
commit 3f02ca7
Show file tree

Hide file tree

Showing 7 changed files with 317 additions and 183 deletions.
diff --git a/test/e2e/cluster_upgrade_runtimesdk.go b/test/e2e/cluster_upgrade_runtimesdk.go
@@ -21,6 +21,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"strconv"
 	"strings"
 	"time"
 
@@ -90,9 +91,13 @@ type ClusterUpgradeWithRuntimeSDKSpecInput struct {
 	// If not specified, this is a no-op.
 	PostUpgrade func(managementClusterProxy framework.ClusterProxy, workloadClusterNamespace, workloadClusterName string)
 
+	// ExtensionConfigName is the name of the ExtensionConfig. Defaults to "k8s-upgrade-with-runtimesdk".
+	ExtensionConfigName string
+
 	// ExtensionServiceNamespace is the namespace where the service for the Runtime SDK is located
 	// and is used to configure in the test-namespace scoped ExtensionConfig.
 	ExtensionServiceNamespace string
+
 	// ExtensionServiceName is the name of the service to configure in the test-namespace scoped ExtensionConfig.
 	ExtensionServiceName string
 }
@@ -133,6 +138,9 @@ func ClusterUpgradeWithRuntimeSDKSpec(ctx context.Context, inputGetter func() Cl
 
 		Expect(input.ExtensionServiceNamespace).ToNot(BeEmpty())
 		Expect(input.ExtensionServiceName).ToNot(BeEmpty())
+		if input.ExtensionConfigName == "" {
+			input.ExtensionConfigName = specName
+		}
 
 		if input.ControlPlaneMachineCount == nil {
 			controlPlaneMachineCount = 1
@@ -161,8 +169,11 @@ func ClusterUpgradeWithRuntimeSDKSpec(ctx context.Context, inputGetter func() Cl
 
 		By("Deploy Test Extension ExtensionConfig")
 
+		// In this test we are defaulting all handlers to blocking because we expect the handlers to block the
+		// cluster lifecycle by default. Setting defaultAllHandlersToBlocking to true enforces that the test-extension
+		// automatically creates the ConfigMap with blocking preloaded responses.
 		Expect(input.BootstrapClusterProxy.GetClient().Create(ctx,
-			extensionConfig(specName, namespace.Name, input.ExtensionServiceNamespace, input.ExtensionServiceName))).
+			extensionConfig(input.ExtensionConfigName, namespace.Name, input.ExtensionServiceNamespace, input.ExtensionServiceName, true, true))).
 			To(Succeed(), "Failed to create the extension config")
 
 		By("Creating a workload cluster; creation waits for BeforeClusterCreateHook to gate the operation")
@@ -304,8 +315,8 @@ func ClusterUpgradeWithRuntimeSDKSpec(ctx context.Context, inputGetter func() Cl
 		if !input.SkipCleanup {
 			// Delete the extensionConfig first to ensure the BeforeDeleteCluster hook doesn't block deletion.
 			Eventually(func() error {
-				return input.BootstrapClusterProxy.GetClient().Delete(ctx, extensionConfig(specName, namespace.Name, input.ExtensionServiceNamespace, input.ExtensionServiceName))
-			}, 10*time.Second, 1*time.Second).Should(Succeed(), "delete extensionConfig failed")
+				return input.BootstrapClusterProxy.GetClient().Delete(ctx, extensionConfig(input.ExtensionConfigName, namespace.Name, input.ExtensionServiceNamespace, input.ExtensionServiceName, true, true))
+			}, 10*time.Second, 1*time.Second).Should(Succeed(), "Deleting ExtensionConfig failed")
 
 			Byf("Deleting cluster %s", klog.KObj(clusterResources.Cluster))
 			// While https://github.com/kubernetes-sigs/cluster-api/issues/2955 is addressed in future iterations, there is a chance
@@ -429,8 +440,8 @@ func machineSetPreflightChecksTestHandler(ctx context.Context, c client.Client,
 // We make sure this cluster-wide object does not conflict with others by using a random generated
 // name and a NamespaceSelector selecting on the namespace of the current test.
 // Thus, this object is "namespaced" to the current test even though it's a cluster-wide object.
-func extensionConfig(name, namespace, extensionServiceNamespace, extensionServiceName string) *runtimev1.ExtensionConfig {
-	return &runtimev1.ExtensionConfig{
+func extensionConfig(name, namespace, extensionServiceNamespace, extensionServiceName string, selectNamespace, defaultAllHandlersToBlocking bool) *runtimev1.ExtensionConfig {
+	cfg := &runtimev1.ExtensionConfig{
 		ObjectMeta: metav1.ObjectMeta{
 			// Note: We have to use a constant name here as we have to be able to reference it in the ClusterClass
 			// when configuring external patches.
@@ -448,25 +459,24 @@ func extensionConfig(name, namespace, extensionServiceNamespace, extensionServic
 					Namespace: extensionServiceNamespace,
 				},
 			},
-			NamespaceSelector: &metav1.LabelSelector{
-				// Note: we are limiting the test extension to be used by the namespace where the test is run.
-				MatchExpressions: []metav1.LabelSelectorRequirement{
-					{
-						Key:      "kubernetes.io/metadata.name",
-						Operator: metav1.LabelSelectorOpIn,
-						Values:   []string{namespace},
-					},
-				},
-			},
 			Settings: map[string]string{
-				// In the E2E test we are defaulting all handlers to blocking because cluster_upgrade_runtimesdk_test
-				// expects the handlers to block the cluster lifecycle by default.
-				// Setting this value to true enforces that the test-extension automatically creates the ConfigMap with
-				// blocking preloaded responses.
-				"defaultAllHandlersToBlocking": "true",
+				"defaultAllHandlersToBlocking": strconv.FormatBool(defaultAllHandlersToBlocking),
 			},
 		},
 	}
+	if selectNamespace {
+		cfg.Spec.NamespaceSelector = &metav1.LabelSelector{
+			// Note: we are limiting the test extension to be used by the namespace where the test is run.
+			MatchExpressions: []metav1.LabelSelectorRequirement{
+				{
+					Key:      "kubernetes.io/metadata.name",
+					Operator: metav1.LabelSelectorOpIn,
+					Values:   []string{namespace},
+				},
+			},
+		}
+	}
+	return cfg
 }
 
 // Check that each hook in hooks has been called at least once by checking if its actualResponseStatus is in the hook response configmap.

diff --git a/test/e2e/data/infrastructure-inmemory/main/bases/cluster-with-topology.yaml b/test/e2e/data/infrastructure-inmemory/main/bases/cluster-with-topology.yaml
@@ -12,6 +12,7 @@ spec:
     serviceDomain: ${SERVICE_DOMAIN:="cluster.local"}
   topology:
     class: in-memory
+    classNamespace: ${NAMESPACE}
     version: ${KUBERNETES_VERSION}
     controlPlane:
       replicas: ${CONTROL_PLANE_MACHINE_COUNT}
@@ -20,3 +21,8 @@ spec:
         - class: default-worker
           name: md-0
           replicas: ${WORKER_MACHINE_COUNT}
+    variables:
+      - name: kubeadmControlPlaneMaxSurge
+        value: "1"
+      - name: imageRepository
+        value: "kindest"
diff --git a/test/e2e/data/infrastructure-inmemory/main/clusterclass-in-memory.yaml b/test/e2e/data/infrastructure-inmemory/main/clusterclass-in-memory.yaml
@@ -1,56 +1,3 @@
-apiVersion: cluster.x-k8s.io/v1beta1
-kind: ClusterClass
-metadata:
-  name: in-memory
-spec:
-  controlPlane:
-    metadata:
-      annotations:
-    machineInfrastructure:
-      ref:
-        apiVersion: infrastructure.cluster.x-k8s.io/v1alpha1
-        kind: InMemoryMachineTemplate
-        name: in-memory-control-plane
-    ref:
-      apiVersion: controlplane.cluster.x-k8s.io/v1beta1
-      kind: KubeadmControlPlaneTemplate
-      name: in-memory-control-plane
-    machineHealthCheck:
-      unhealthyConditions:
-        - type: Ready
-          status: Unknown
-          timeout: 300s
-        - type: Ready
-          status: "False"
-          timeout: 300s
-  infrastructure:
-    ref:
-      apiVersion: infrastructure.cluster.x-k8s.io/v1alpha1
-      kind: InMemoryClusterTemplate
-      name: in-memory-cluster
-  workers:
-    machineDeployments:
-      - class: default-worker
-        template:
-          bootstrap:
-            ref:
-              apiVersion: bootstrap.cluster.x-k8s.io/v1beta1
-              kind: KubeadmConfigTemplate
-              name: in-memory-default-worker-bootstraptemplate
-          infrastructure:
-            ref:
-              apiVersion: infrastructure.cluster.x-k8s.io/v1alpha1
-              kind: InMemoryMachineTemplate
-              name: in-memory-default-worker-machinetemplate
-        machineHealthCheck:
-          unhealthyConditions:
-            - type: Ready
-              status: Unknown
-              timeout: 300s
-            - type: Ready
-              status: "False"
-              timeout: 300s
----
 apiVersion: infrastructure.cluster.x-k8s.io/v1alpha1
 kind: InMemoryClusterTemplate
 metadata:
@@ -95,19 +42,19 @@ spec:
       behaviour:
         vm:
           provisioning:
-            startupDuration: "30s"
+            startupDuration: "10s"
             startupJitter: "0.2"
         node:
           provisioning:
-            startupDuration: "10s"
+            startupDuration: "2s"
             startupJitter: "0.2"
         apiServer:
           provisioning:
-            startupDuration: "10s"
+            startupDuration: "2s"
             startupJitter: "0.2"
         etcd:
           provisioning:
-            startupDuration: "10s"
+            startupDuration: "2s"
             startupJitter: "0.2"
 ---
 apiVersion: infrastructure.cluster.x-k8s.io/v1alpha1
@@ -120,19 +67,19 @@ spec:
       behaviour:
         vm:
           provisioning:
-            startupDuration: "30s"
+            startupDuration: "10s"
             startupJitter: "0.2"
         node:
           provisioning:
-            startupDuration: "10s"
+            startupDuration: "2s"
             startupJitter: "0.2"
         apiServer:
           provisioning:
-            startupDuration: "10s"
+            startupDuration: "2s"
             startupJitter: "0.2"
         etcd:
           provisioning:
-            startupDuration: "10s"
+            startupDuration: "2s"
             startupJitter: "0.2"
 ---
 apiVersion: bootstrap.cluster.x-k8s.io/v1beta1
@@ -146,4 +93,62 @@ spec:
         nodeRegistration:
           criSocket: unix:///var/run/containerd/containerd.sock
           kubeletExtraArgs:
-            eviction-hard: nodefs.available<0%,nodefs.inodesFree<0%,imagefs.available<0%
+            eviction-hard: nodefs.available<0%,nodefs.inodesFree<0%,imagefs.available<0%
+---
+apiVersion: cluster.x-k8s.io/v1beta1
+kind: ClusterClass
+metadata:
+  name: in-memory
+spec:
+  controlPlane:
+    metadata:
+      annotations:
+    machineInfrastructure:
+      ref:
+        apiVersion: infrastructure.cluster.x-k8s.io/v1alpha1
+        kind: InMemoryMachineTemplate
+        name: in-memory-control-plane
+    ref:
+      apiVersion: controlplane.cluster.x-k8s.io/v1beta1
+      kind: KubeadmControlPlaneTemplate
+      name: in-memory-control-plane
+    machineHealthCheck:
+      unhealthyConditions:
+      - type: Ready
+        status: Unknown
+        timeout: 300s
+      - type: Ready
+        status: "False"
+        timeout: 300s
+  infrastructure:
+    ref:
+      apiVersion: infrastructure.cluster.x-k8s.io/v1alpha1
+      kind: InMemoryClusterTemplate
+      name: in-memory-cluster
+  workers:
+    machineDeployments:
+    - class: default-worker
+      template:
+        bootstrap:
+          ref:
+            apiVersion: bootstrap.cluster.x-k8s.io/v1beta1
+            kind: KubeadmConfigTemplate
+            name: in-memory-default-worker-bootstraptemplate
+        infrastructure:
+          ref:
+            apiVersion: infrastructure.cluster.x-k8s.io/v1alpha1
+            kind: InMemoryMachineTemplate
+            name: in-memory-default-worker-machinetemplate
+      machineHealthCheck:
+        unhealthyConditions:
+        - type: Ready
+          status: Unknown
+          timeout: 300s
+        - type: Ready
+          status: "False"
+          timeout: 300s
+  patches:
+  - name: test-patch
+    external:
+      generateExtension: generate-patches.scale
+      discoverVariablesExtension: discover-variables.scale