merge into default handler and add uts/e2e

Signed-off-by: Abner-1 <yuanyuxing.yyx@alibaba-inc.com>

.github/workflows/e2e-1.18.yaml

@@ -34,7 +34,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Install-CSI
         run: |

.github/workflows/e2e-1.24.yaml

@@ -35,7 +35,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Install-CSI
         run: |

.github/workflows/e2e-1.28.yaml

@@ -34,7 +34,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf-with-vpa.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Install-CSI
         run: |
@@ -117,7 +117,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf-with-vpa.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Build image
         run: |
@@ -196,7 +196,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf-with-vpa.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Build image
         run: |
@@ -288,7 +288,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf-with-vpa.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Build image
         run: |
@@ -380,7 +380,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf-with-vpa.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Build image
         run: |
@@ -472,7 +472,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf-with-vpa.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Build image
         run: |
@@ -542,7 +542,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf-with-vpa.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Build image
         run: |
@@ -592,6 +592,89 @@ jobs:
               done < <(kubectl get pods -n kruise-system -l control-plane=controller-manager --no-headers | awk '{print $1}')             
           fi
           exit $retVal
+  clonesetAndInplace:
+    runs-on: ubuntu-20.04
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: true
+      - name: Setup Go
+        uses: actions/setup-go@v5
+        with:
+          go-version: ${{ env.GO_VERSION }}
+      - name: Setup Kind Cluster
+        uses: helm/kind-action@v1.10.0
+        with:
+          node_image: ${{ env.KIND_IMAGE }}
+          cluster_name: ${{ env.KIND_CLUSTER_NAME }}
+          config: ./test/kind-conf-with-vpa.yaml
+          version: ${{ env.KIND_VERSION }}
+      - name: Install-CSI
+        run: |
+          make install-csi
+
+      - name: Build image
+        run: |
+          export IMAGE="openkruise/kruise-manager:e2e-${GITHUB_RUN_ID}"
+          docker build --pull --no-cache . -t $IMAGE
+          kind load docker-image --name=${KIND_CLUSTER_NAME} $IMAGE || { echo >&2 "kind not installed or error loading image: $IMAGE"; exit 1; }
+      - name: Install Kruise
+        run: |
+          set -ex
+          kubectl cluster-info
+          IMG=openkruise/kruise-manager:e2e-${GITHUB_RUN_ID} ./scripts/deploy_kind.sh
+          NODES=$(kubectl get node | wc -l)
+          for ((i=1;i<10;i++));
+          do
+            set +e
+            PODS=$(kubectl get pod -n kruise-system | grep '1/1' | wc -l)
+            set -e
+            if [ "$PODS" -eq "$NODES" ]; then
+              break
+            fi
+            sleep 3
+          done
+          set +e
+          PODS=$(kubectl get pod -n kruise-system | grep '1/1' | wc -l)
+          kubectl get node -o yaml
+          kubectl get all -n kruise-system -o yaml
+          kubectl get pod -n kruise-system --no-headers | grep daemon | awk '{print $1}' | xargs kubectl logs -n kruise-system
+          kubectl get pod -n kruise-system --no-headers | grep daemon | awk '{print $1}' | xargs kubectl logs -n kruise-system --previous=true
+          set -e
+          if [ "$PODS" -eq "$NODES" ]; then
+            echo "Wait for kruise-manager and kruise-daemon ready successfully"
+          else
+            echo "Timeout to wait for kruise-manager and kruise-daemon ready"
+            exit 1
+          fi
+      - name: Run E2E Tests
+        run: |
+          export KUBECONFIG=/home/runner/.kube/config
+          make ginkgo
+          set +e
+          ./bin/ginkgo -p -timeout 120m -v --focus='\[apps\] (InplaceVPA)' test/e2e
+          retVal=$?
+          restartCount=$(kubectl get pod -n kruise-system -l control-plane=controller-manager --no-headers | awk '{print $4}')
+          if [ "${restartCount}" -eq "0" ];then
+              echo "Kruise-manager has not restarted"
+          else
+              kubectl get pod -n kruise-system -l control-plane=controller-manager --no-headers
+              echo "Kruise-manager has restarted, abort!!!"
+              kubectl get pod -n kruise-system --no-headers -l control-plane=controller-manager | awk '{print $1}' | xargs kubectl logs -p -n kruise-system
+              exit 1
+          fi
+          if [ "$retVal" -ne 0 ];then
+              echo "test fail, dump kruise-manager logs"
+              while read pod; do 
+                   kubectl logs -n kruise-system $pod
+              done < <(kubectl get pods -n kruise-system -l control-plane=controller-manager --no-headers | awk '{print $1}') 
+              echo "test fail, dump kruise-daemon logs"
+              while read pod; do 
+                   kubectl logs -n kruise-system $pod
+              done < <(kubectl get pods -n kruise-system -l control-plane=daemon --no-headers | awk '{print $1}')         
+          fi
+          exit $retVal
+
   other:
     runs-on: ubuntu-20.04
     steps:
@@ -607,7 +690,7 @@ jobs:
         with:
           node_image: ${{ env.KIND_IMAGE }}
           cluster_name: ${{ env.KIND_CLUSTER_NAME }}
-          config: ./test/kind-conf-none-fg.yaml
+          config: ./test/kind-conf-with-vpa.yaml
           version: ${{ env.KIND_VERSION }}
       - name: Build image
         run: |
@@ -648,7 +731,7 @@ jobs:
           export KUBECONFIG=/home/runner/.kube/config
           make ginkgo
           set +e
-          ./bin/ginkgo -timeout 90m -v --skip='\[apps\] (AppStatefulSetStorage|StatefulSet|PullImage|PullImages|ContainerRecreateRequest|DaemonSet|SidecarSet|EphemeralJob)' --skip='\[policy\] PodUnavailableBudget' test/e2e
+          ./bin/ginkgo -timeout 90m -v --skip='\[apps\] (InplaceVPA|AppStatefulSetStorage|StatefulSet|PullImage|PullImages|ContainerRecreateRequest|DaemonSet|SidecarSet|EphemeralJob)' --skip='\[policy\] PodUnavailableBudget' test/e2e
           retVal=$?
           restartCount=$(kubectl get pod -n kruise-system -l control-plane=controller-manager --no-headers | awk '{print $4}')
           if [ "${restartCount}" -eq "0" ];then

Makefile

@@ -152,9 +152,15 @@ endif
 create-cluster: $(tools/kind)
 	tools/hack/create-cluster.sh
 
+DISABLE_CSI ?= false
+
 .PHONY: install-csi
 install-csi:
+ifeq ($(DISABLE_CSI), true)
+	@echo "CSI is disabled, skip"
+else
 	cd tools/hack/csi-driver-host-path; ./install-snapshot.sh
+endif
 
 # delete-cluster deletes a kube cluster.
 .PHONY: delete-cluster

apis/apps/pub/inplace_update.go

@@ -62,6 +62,12 @@ type InPlaceUpdateState struct {
 	// UpdateEnvFromMetadata indicates there are envs from annotations/labels that should be in-place update.
 	UpdateEnvFromMetadata bool `json:"updateEnvFromMetadata,omitempty"`
 
+	// UpdateResources indicates there are resources that should be in-place update.
+	UpdateResources bool `json:"updateResources,omitempty"`
+
+	// VerticalUpdateOnly indicates there are only vertical update in this revision.
+	VerticalUpdateOnly bool `json:"verticalUpdateOnly"`
+
 	// NextContainerImages is the containers with lower priority that waiting for in-place update images in next batch.
 	NextContainerImages map[string]string `json:"nextContainerImages,omitempty"`
 
@@ -94,8 +100,8 @@ type InPlaceUpdateContainerBatch struct {
 // InPlaceUpdateContainerStatus records the statuses of the container that are mainly used
 // to determine whether the InPlaceUpdate is completed.
 type InPlaceUpdateContainerStatus struct {
-	ImageID  string                  `json:"imageID,omitempty"`
-	Resource v1.ResourceRequirements `json:"resource,omitempty"`
+	ImageID   string                  `json:"imageID,omitempty"`
+	Resources v1.ResourceRequirements `json:"resources,omitempty"`
 }
 
 // InPlaceUpdateStrategy defines the strategies for in-place update.
@@ -144,6 +150,10 @@ type RuntimeContainerHashes struct {
 	// PlainHash is the hash that directly calculated from pod.spec.container[x].
 	// Usually it is calculated by Kubelet and will be in annotation of each runtime container.
 	PlainHash uint64 `json:"plainHash"`
+	// PlainHashWithoutResources is the hash that directly calculated from pod.spec.container[x]
+	// over fields with Resources field zero'd out.
+	// Usually it is calculated by Kubelet and will be in annotation of each runtime container.
+	PlainHashWithoutResources uint64 `json:"plainHashWithoutResources"`
 	// ExtractedEnvFromMetadataHash is the hash that calculated from pod.spec.container[x],
 	// whose envs from annotations/labels have already been extracted to the real values.
 	ExtractedEnvFromMetadataHash uint64 `json:"extractedEnvFromMetadataHash,omitempty"`

apis/apps/v1alpha1/cloneset_types.go

@@ -17,10 +17,11 @@ limitations under the License.
 package v1alpha1
 
 import (
-	appspub "github.com/openkruise/kruise/apis/apps/pub"
 	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/util/intstr"
+
+	appspub "github.com/openkruise/kruise/apis/apps/pub"
 )
 
 const (
@@ -178,6 +179,8 @@ type CloneSetStatus struct {
 
 	// UpdatedAvailableReplicas is the number of Pods created by the CloneSet controller from the CloneSet version
 	// indicated by updateRevision and have a Ready Condition for at least minReadySeconds.
+	// Notice: when enable InPlaceWorkloadVerticalScaling, only resource resize updating pod will also be unavailable.
+	// This means these pod will be counted in maxUnavailable.
 	UpdatedAvailableReplicas int32 `json:"updatedAvailableReplicas,omitempty"`
 
 	// ExpectedUpdatedReplicas is the number of Pods that should be updated by CloneSet controller.
@@ -237,6 +240,7 @@ type CloneSetCondition struct {
 // +kubebuilder:printcolumn:name="DESIRED",type="integer",JSONPath=".spec.replicas",description="The desired number of pods."
 // +kubebuilder:printcolumn:name="UPDATED",type="integer",JSONPath=".status.updatedReplicas",description="The number of pods updated."
 // +kubebuilder:printcolumn:name="UPDATED_READY",type="integer",JSONPath=".status.updatedReadyReplicas",description="The number of pods updated and ready."
+// +kubebuilder:printcolumn:name="UPDATED_AVAILABLE",type="integer",JSONPath=".status.updatedAvailableReplicas",description="The number of pods updated and available."
 // +kubebuilder:printcolumn:name="READY",type="integer",JSONPath=".status.readyReplicas",description="The number of pods ready."
 // +kubebuilder:printcolumn:name="TOTAL",type="integer",JSONPath=".status.replicas",description="The number of currently all pods."
 // +kubebuilder:printcolumn:name="AGE",type="date",JSONPath=".metadata.creationTimestamp",description="CreationTimestamp is a timestamp representing the server time when this object was created. It is not guaranteed to be set in happens-before order across separate operations. Clients may not set this value. It is represented in RFC3339 form and is in UTC."

config/crd/bases/apps.kruise.io_clonesets.yaml

@@ -29,6 +29,10 @@ spec:
       jsonPath: .status.updatedReadyReplicas
       name: UPDATED_READY
       type: integer
+    - description: The number of pods updated and available.
+      jsonPath: .status.updatedAvailableReplicas
+      name: UPDATED_AVAILABLE
+      type: integer
     - description: The number of pods ready.
       jsonPath: .status.readyReplicas
       name: READY
@@ -512,6 +516,8 @@ spec:
                 description: |-
                   UpdatedAvailableReplicas is the number of Pods created by the CloneSet controller from the CloneSet version
                   indicated by updateRevision and have a Ready Condition for at least minReadySeconds.
+                  Notice: when enable InPlaceWorkloadVerticalScaling, only resource resize updating pod will also be unavailable.
+                  This means these pod will be counted in maxUnavailable.
                 format: int32
                 type: integer
               updatedReadyReplicas:

config/manager/manager.yaml

@@ -60,8 +60,8 @@ spec:
             port: 8000
         resources:
           limits:
-            cpu: 100m
-            memory: 200Mi
+            cpu: 2
+            memory: 2Gi
           requests:
             cpu: 100m
             memory: 200Mi

pkg/controller/cloneset/cloneset_event_handler.go

@@ -138,6 +138,13 @@ func (e *podEventHandler) Update(ctx context.Context, evt event.UpdateEvent, q w
 
 	// If it has a ControllerRef, that's all that matters.
 	if curControllerRef != nil {
+		// TODO(Abner-1): delete it when fixes only resize resource
+		//old, _ := json.Marshal(oldPod)
+		//cur, _ := json.Marshal(curPod)
+		//patches, _ := jsonpatch.CreatePatch(old, cur)
+		//pjson, _ := json.Marshal(patches)
+		//klog.V(4).InfoS("Pod updated json", "pod", klog.KObj(curPod), "patch", pjson)
+
 		req := resolveControllerRef(curPod.Namespace, curControllerRef)
 		if req == nil {
 			return

pkg/controller/cloneset/core/cloneset_core.go

@@ -32,6 +32,8 @@ import (
 	appspub "github.com/openkruise/kruise/apis/apps/pub"
 	appsv1alpha1 "github.com/openkruise/kruise/apis/apps/v1alpha1"
 	clonesetutils "github.com/openkruise/kruise/pkg/controller/cloneset/utils"
+	"github.com/openkruise/kruise/pkg/features"
+	utilfeature "github.com/openkruise/kruise/pkg/util/feature"
 	"github.com/openkruise/kruise/pkg/util/inplaceupdate"
 )
 
@@ -211,6 +213,15 @@ func (c *commonControl) IgnorePodUpdateEvent(oldPod, curPod *v1.Pod) bool {
 			}
 		}
 	}
+	// only inplace resource resize
+	if utilfeature.DefaultFeatureGate.Enabled(features.InPlaceWorkloadVerticalScaling) &&
+		len(curPod.Labels) > 0 && appspub.LifecycleStateType(curPod.Labels[appspub.LifecycleStateKey]) != appspub.LifecycleStateNormal {
+		opts := c.GetUpdateOptions()
+		opts = inplaceupdate.SetOptionsDefaults(opts)
+		if err := containersUpdateCompleted(curPod, opts.CheckContainersUpdateCompleted); err == nil {
+			return false
+		}
+	}
 
 	return true
 }

pkg/controller/cloneset/sync/cloneset_sync_utils.go

@@ -17,6 +17,7 @@ limitations under the License.
 package sync
 
 import (
+	"encoding/json"
 	"flag"
 	"math"
 	"reflect"
@@ -87,6 +88,12 @@ func (e expectationDiffs) isEmpty() bool {
 	return reflect.DeepEqual(e, expectationDiffs{})
 }
 
+// String implement this to print information in klog
+func (e expectationDiffs) String() string {
+	b, _ := json.Marshal(e)
+	return string(b)
+}
+
 type IsPodUpdateFunc func(pod *v1.Pod, updateRevision string) bool
 
 // This is the most important algorithm in cloneset-controller.

pkg/daemon/containermeta/container_meta_controller.go

@@ -344,7 +344,10 @@ func (c *Controller) manageContainerMetaSet(pod *v1.Pod, kubePodStatus *kubeletc
 				Name:         status.Name,
 				ContainerID:  status.ID.String(),
 				RestartCount: int32(status.RestartCount),
-				Hashes:       appspub.RuntimeContainerHashes{PlainHash: status.Hash},
+				Hashes: appspub.RuntimeContainerHashes{
+					PlainHash:                 status.Hash,
+					PlainHashWithoutResources: status.HashWithoutResources,
+				},
 			}
 		}
 		if utilfeature.DefaultFeatureGate.Enabled(features.InPlaceUpdateEnvFromMetadata) {