Merge pull request #709 from damemi/update-helm-23

Update helm chart version to v0.23

Dockerfile

@@ -11,7 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-FROM golang:1.17.7
+FROM golang:1.17.3
 
 WORKDIR /go/src/sigs.k8s.io/descheduler
 COPY . .

charts/descheduler/Chart.yaml

@@ -1,7 +1,7 @@
 apiVersion: v1
 name: descheduler
-version: 0.23.2
-appVersion: 0.23.1
+version: 0.23.0
+appVersion: 0.23.0
 description: Descheduler for Kubernetes is used to rebalance clusters by evicting pods that can potentially be scheduled on better nodes. In the current implementation, descheduler does not schedule replacement of evicted pods but relies on the default scheduler for that.
 keywords:
 - kubernetes

charts/descheduler/templates/clusterrole.yaml

@@ -14,7 +14,7 @@ rules:
   verbs: ["get", "watch", "list"]
 - apiGroups: [""]
   resources: ["namespaces"]
-  verbs: ["get", "watch", "list"]
+  verbs: ["get", "list"]
 - apiGroups: [""]
   resources: ["pods"]
   verbs: ["get", "watch", "list", "delete"]

cmd/descheduler/app/server.go

@@ -73,7 +73,7 @@ func NewDeschedulerCommand(out io.Writer) *cobra.Command {
 			}
 
 			ctx, done := signal.NotifyContext(context.Background(), syscall.SIGINT, syscall.SIGTERM)
-
+			defer done()
 			pathRecorderMux := mux.NewPathRecorderMux("descheduler")
 			if !s.DisableMetrics {
 				pathRecorderMux.Handle("/metrics", legacyregistry.HandlerWithReset())
@@ -81,20 +81,15 @@ func NewDeschedulerCommand(out io.Writer) *cobra.Command {
 
 			healthz.InstallHandler(pathRecorderMux, healthz.NamedCheck("Descheduler", healthz.PingHealthz.Check))
 
-			stoppedCh, err := SecureServing.Serve(pathRecorderMux, 0, ctx.Done())
-			if err != nil {
+			if _, err := SecureServing.Serve(pathRecorderMux, 0, ctx.Done()); err != nil {
 				klog.Fatalf("failed to start secure server: %v", err)
 				return
 			}
 
-			err = Run(ctx, s)
+			err := Run(ctx, s)
 			if err != nil {
 				klog.ErrorS(err, "descheduler server")
 			}
-
-			done()
-			// wait for metrics server to close
-			<-stoppedCh
 		},
 	}
 	cmd.SetOut(out)

docs/user-guide.md

@@ -2,15 +2,14 @@
 
 Starting with descheduler release v0.10.0 container images are available in the official k8s container registry.
 
-Descheduler Version | Container Image                            | Architectures           |
-------------------- |--------------------------------------------|-------------------------|
-v0.23.1             | k8s.gcr.io/descheduler/descheduler:v0.23.1 | AMD64<br>ARM64<br>ARMv7 |
-v0.22.0             | k8s.gcr.io/descheduler/descheduler:v0.22.0 | AMD64<br>ARM64<br>ARMv7 |
-v0.21.0             | k8s.gcr.io/descheduler/descheduler:v0.21.0 | AMD64<br>ARM64<br>ARMv7 |
-v0.20.0             | k8s.gcr.io/descheduler/descheduler:v0.20.0 | AMD64<br>ARM64          |
-v0.19.0             | k8s.gcr.io/descheduler/descheduler:v0.19.0 | AMD64                   |
-v0.18.0             | k8s.gcr.io/descheduler/descheduler:v0.18.0 | AMD64                   |
-v0.10.0             | k8s.gcr.io/descheduler/descheduler:v0.10.0 | AMD64                   |
+Descheduler Version | Container Image                                     | Architectures           |
+------------------- |-----------------------------------------------------|-------------------------|
+v0.22.0             | k8s.gcr.io/descheduler/descheduler:v0.22.0          | AMD64<br>ARM64<br>ARMv7 |
+v0.21.0             | k8s.gcr.io/descheduler/descheduler:v0.21.0          | AMD64<br>ARM64<br>ARMv7 |
+v0.20.0             | k8s.gcr.io/descheduler/descheduler:v0.20.0          | AMD64<br>ARM64          |
+v0.19.0             | k8s.gcr.io/descheduler/descheduler:v0.19.0          | AMD64                   |
+v0.18.0             | k8s.gcr.io/descheduler/descheduler:v0.18.0          | AMD64                   |
+v0.10.0             | k8s.gcr.io/descheduler/descheduler:v0.10.0          | AMD64                   |
 
 Note that multi-arch container images cannot be pulled by [kind](https://kind.sigs.k8s.io) from a registry. Therefore
 starting with descheduler release v0.20.0 use the below process to download the official descheduler

kubernetes/base/rbac.yaml

@@ -12,7 +12,7 @@ rules:
   verbs: ["get", "watch", "list"]
 - apiGroups: [""]
   resources: ["namespaces"]
-  verbs: ["get", "watch", "list"]
+  verbs: ["get", "list"]
 - apiGroups: [""]
   resources: ["pods"]
   verbs: ["get", "watch", "list", "delete"]

kubernetes/cronjob/cronjob.yaml

@@ -16,7 +16,7 @@ spec:
           priorityClassName: system-cluster-critical
           containers:
           - name: descheduler
-            image: k8s.gcr.io/descheduler/descheduler:v0.23.1
+            image: k8s.gcr.io/descheduler/descheduler:v0.22.0
             volumeMounts:
             - mountPath: /policy-dir
               name: policy-volume

kubernetes/deployment/deployment.yaml

@@ -19,7 +19,7 @@ spec:
       serviceAccountName: descheduler-sa
       containers:
         - name: descheduler
-          image: k8s.gcr.io/descheduler/descheduler:v0.23.1
+          image: k8s.gcr.io/descheduler/descheduler:v0.22.0
           imagePullPolicy: IfNotPresent
           command:
             - "/bin/descheduler"

kubernetes/job/job.yaml

@@ -14,7 +14,7 @@ spec:
       priorityClassName: system-cluster-critical
       containers:
         - name: descheduler
-          image: k8s.gcr.io/descheduler/descheduler:v0.23.1
+          image: k8s.gcr.io/descheduler/descheduler:v0.22.0
           volumeMounts:
           - mountPath: /policy-dir
             name: policy-volume

pkg/descheduler/descheduler.go

@@ -69,7 +69,13 @@ func Run(ctx context.Context, rs *options.DeschedulerServer) error {
 		return err
 	}
 
-	return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion)
+	// tie in root ctx with our wait stopChannel
+	stopChannel := make(chan struct{})
+	go func() {
+		<-ctx.Done()
+		close(stopChannel)
+	}()
+	return RunDeschedulerStrategies(ctx, rs, deschedulerPolicy, evictionPolicyGroupVersion, stopChannel)
 }
 
 type strategyFunction func(ctx context.Context, client clientset.Interface, strategy api.DeschedulerStrategy, nodes []*v1.Node, podEvictor *evictions.PodEvictor, getPodsAssignedToNode podutil.GetPodsAssignedToNodeFunc)
@@ -150,16 +156,13 @@ func cachedClient(
 	return fakeClient, nil
 }
 
-func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string) error {
+func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer, deschedulerPolicy *api.DeschedulerPolicy, evictionPolicyGroupVersion string, stopChannel chan struct{}) error {
 	sharedInformerFactory := informers.NewSharedInformerFactory(rs.Client, 0)
 	nodeInformer := sharedInformerFactory.Core().V1().Nodes()
 	podInformer := sharedInformerFactory.Core().V1().Pods()
 	namespaceInformer := sharedInformerFactory.Core().V1().Namespaces()
 	priorityClassInformer := sharedInformerFactory.Scheduling().V1().PriorityClasses()
 
-	ctx, cancel := context.WithCancel(ctx)
-	defer cancel()
-
 	// create the informers
 	namespaceInformer.Informer()
 	priorityClassInformer.Informer()
@@ -169,8 +172,8 @@ func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer
 		return fmt.Errorf("build get pods assigned to node function error: %v", err)
 	}
 
-	sharedInformerFactory.Start(ctx.Done())
-	sharedInformerFactory.WaitForCacheSync(ctx.Done())
+	sharedInformerFactory.Start(stopChannel)
+	sharedInformerFactory.WaitForCacheSync(stopChannel)
 
 	strategyFuncs := map[api.StrategyName]strategyFunction{
 		"RemoveDuplicates":                            strategies.RemoveDuplicatePods,
@@ -220,13 +223,13 @@ func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer
 		nodes, err := nodeutil.ReadyNodes(ctx, rs.Client, nodeInformer, nodeSelector)
 		if err != nil {
 			klog.V(1).InfoS("Unable to get ready nodes", "err", err)
-			cancel()
+			close(stopChannel)
 			return
 		}
 
 		if len(nodes) <= 1 {
 			klog.V(1).InfoS("The cluster size is 0 or 1 meaning eviction causes service disruption or degradation. So aborting..")
-			cancel()
+			close(stopChannel)
 			return
 		}
 
@@ -289,9 +292,9 @@ func RunDeschedulerStrategies(ctx context.Context, rs *options.DeschedulerServer
 
 		// If there was no interval specified, send a signal to the stopChannel to end the wait.Until loop after 1 iteration
 		if rs.DeschedulingInterval.Seconds() == 0 {
-			cancel()
+			close(stopChannel)
 		}
-	}, rs.DeschedulingInterval, ctx.Done())
+	}, rs.DeschedulingInterval, stopChannel)
 
 	return nil
 }

pkg/descheduler/descheduler_test.go

@@ -35,6 +35,9 @@ func TestTaintsUpdated(t *testing.T) {
 		},
 	}
 
+	stopChannel := make(chan struct{})
+	defer close(stopChannel)
+
 	rs, err := options.NewDeschedulerServer()
 	if err != nil {
 		t.Fatalf("Unable to initialize server: %v", err)
@@ -44,7 +47,7 @@ func TestTaintsUpdated(t *testing.T) {
 	errChan := make(chan error, 1)
 	defer close(errChan)
 	go func() {
-		err := RunDeschedulerStrategies(ctx, rs, dp, "v1beta1")
+		err := RunDeschedulerStrategies(ctx, rs, dp, "v1beta1", stopChannel)
 		errChan <- err
 	}()
 	select {
@@ -98,69 +101,3 @@ func TestTaintsUpdated(t *testing.T) {
 		t.Fatalf("Unable to evict pod, node taint did not get propagated to descheduler strategies")
 	}
 }
-
-func TestRootCancel(t *testing.T) {
-	ctx, cancel := context.WithCancel(context.Background())
-	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
-	n2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
-	client := fakeclientset.NewSimpleClientset(n1, n2)
-	dp := &api.DeschedulerPolicy{
-		Strategies: api.StrategyList{}, // no strategies needed for this test
-	}
-
-	rs, err := options.NewDeschedulerServer()
-	if err != nil {
-		t.Fatalf("Unable to initialize server: %v", err)
-	}
-	rs.Client = client
-	rs.DeschedulingInterval = 100 * time.Millisecond
-	errChan := make(chan error, 1)
-	defer close(errChan)
-
-	go func() {
-		err := RunDeschedulerStrategies(ctx, rs, dp, "v1beta1")
-		errChan <- err
-	}()
-	cancel()
-	select {
-	case err := <-errChan:
-		if err != nil {
-			t.Fatalf("Unable to run descheduler strategies: %v", err)
-		}
-	case <-time.After(1 * time.Second):
-		t.Fatal("Root ctx should have canceled immediately")
-	}
-}
-
-func TestRootCancelWithNoInterval(t *testing.T) {
-	ctx, cancel := context.WithCancel(context.Background())
-	n1 := test.BuildTestNode("n1", 2000, 3000, 10, nil)
-	n2 := test.BuildTestNode("n2", 2000, 3000, 10, nil)
-	client := fakeclientset.NewSimpleClientset(n1, n2)
-	dp := &api.DeschedulerPolicy{
-		Strategies: api.StrategyList{}, // no strategies needed for this test
-	}
-
-	rs, err := options.NewDeschedulerServer()
-	if err != nil {
-		t.Fatalf("Unable to initialize server: %v", err)
-	}
-	rs.Client = client
-	rs.DeschedulingInterval = 0
-	errChan := make(chan error, 1)
-	defer close(errChan)
-
-	go func() {
-		err := RunDeschedulerStrategies(ctx, rs, dp, "v1beta1")
-		errChan <- err
-	}()
-	cancel()
-	select {
-	case err := <-errChan:
-		if err != nil {
-			t.Fatalf("Unable to run descheduler strategies: %v", err)
-		}
-	case <-time.After(1 * time.Second):
-		t.Fatal("Root ctx should have canceled immediately")
-	}
-}

test/e2e/e2e_test.go

@@ -984,7 +984,9 @@ func TestDeschedulingInterval(t *testing.T) {
 		if err != nil || len(evictionPolicyGroupVersion) == 0 {
 			t.Errorf("Error when checking support for eviction: %v", err)
 		}
-		if err := descheduler.RunDeschedulerStrategies(ctx, s, deschedulerPolicy, evictionPolicyGroupVersion); err != nil {
+
+		stopChannel := make(chan struct{})
+		if err := descheduler.RunDeschedulerStrategies(ctx, s, deschedulerPolicy, evictionPolicyGroupVersion, stopChannel); err != nil {
 			t.Errorf("Error running descheduler strategies: %+v", err)
 		}
 		c <- true