Merge branch 'kubernetes:master' into master

.github/workflows/ci.yaml

@@ -17,7 +17,7 @@ jobs:
       - name: Set up Go
         uses: actions/setup-go@v2
         with:
-          go-version: '1.22.1'
+          go-version: '1.22.2'
 
       - uses: actions/checkout@v2
         with:

cluster-autoscaler/OWNERS

@@ -1,14 +1,14 @@
 approvers:
+- aleksandra-malinowska
 - BigDarkClown
 - feiskyer
 - towca
 - x13n
 reviewers:
+- aleksandra-malinowska
 - BigDarkClown
 - feiskyer
 - vadasambar
 - x13n
-emeritus_approvers:
-- aleksandra-malinowska # 2022-09-30
 labels:
 - area/cluster-autoscaler

cluster-autoscaler/apis/go.mod

@@ -1,32 +1,32 @@
 module k8s.io/autoscaler/cluster-autoscaler/apis
 
-go 1.21
+go 1.22.0
 
 require (
-	github.com/onsi/ginkgo/v2 v2.16.0
-	github.com/onsi/gomega v1.31.1
-	k8s.io/apimachinery v0.29.2
-	k8s.io/client-go v0.29.2
-	k8s.io/code-generator v0.29.2
+	github.com/onsi/ginkgo/v2 v2.19.0
+	github.com/onsi/gomega v1.33.1
+	k8s.io/apimachinery v0.31.0-alpha.2
+	k8s.io/client-go v0.31.0-alpha.2
+	k8s.io/code-generator v0.31.0-alpha.2
 	sigs.k8s.io/structured-merge-diff/v4 v4.4.1
 )
 
 require (
-	github.com/davecgh/go-spew v1.1.1 // indirect
+	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
 	github.com/emicklei/go-restful/v3 v3.11.0 // indirect
-	github.com/evanphx/json-patch v4.12.0+incompatible // indirect
+	github.com/fxamacker/cbor/v2 v2.7.0-beta // indirect
 	github.com/go-logr/logr v1.4.1 // indirect
 	github.com/go-openapi/jsonpointer v0.19.6 // indirect
 	github.com/go-openapi/jsonreference v0.20.2 // indirect
-	github.com/go-openapi/swag v0.22.3 // indirect
-	github.com/go-task/slim-sprig v0.0.0-20230315185526-52ccab3ef572 // indirect
+	github.com/go-openapi/swag v0.22.4 // indirect
+	github.com/go-task/slim-sprig/v3 v3.0.0 // indirect
 	github.com/gogo/protobuf v1.3.2 // indirect
-	github.com/golang/protobuf v1.5.3 // indirect
+	github.com/golang/protobuf v1.5.4 // indirect
 	github.com/google/gnostic-models v0.6.8 // indirect
 	github.com/google/go-cmp v0.6.0 // indirect
 	github.com/google/gofuzz v1.2.0 // indirect
-	github.com/google/pprof v0.0.0-20210720184732-4bb14d4b1be1 // indirect
-	github.com/google/uuid v1.3.0 // indirect
+	github.com/google/pprof v0.0.0-20240525223248-4bfdf5a9a2af // indirect
+	github.com/google/uuid v1.3.1 // indirect
 	github.com/josharian/intern v1.0.0 // indirect
 	github.com/json-iterator/go v1.1.12 // indirect
 	github.com/mailru/easyjson v0.7.7 // indirect
@@ -35,24 +35,26 @@ require (
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/pkg/errors v0.9.1 // indirect
 	github.com/spf13/pflag v1.0.5 // indirect
-	golang.org/x/mod v0.14.0 // indirect
-	golang.org/x/net v0.20.0 // indirect
-	golang.org/x/oauth2 v0.10.0 // indirect
-	golang.org/x/sys v0.16.0 // indirect
-	golang.org/x/term v0.16.0 // indirect
-	golang.org/x/text v0.14.0 // indirect
+	github.com/x448/float16 v0.8.4 // indirect
+	golang.org/x/mod v0.17.0 // indirect
+	golang.org/x/net v0.25.0 // indirect
+	golang.org/x/oauth2 v0.20.0 // indirect
+	golang.org/x/sync v0.7.0 // indirect
+	golang.org/x/sys v0.20.0 // indirect
+	golang.org/x/term v0.20.0 // indirect
+	golang.org/x/text v0.15.0 // indirect
 	golang.org/x/time v0.3.0 // indirect
-	golang.org/x/tools v0.17.0 // indirect
-	google.golang.org/appengine v1.6.7 // indirect
-	google.golang.org/protobuf v1.31.0 // indirect
+	golang.org/x/tools v0.21.0 // indirect
+	google.golang.org/protobuf v1.33.0 // indirect
+	gopkg.in/evanphx/json-patch.v4 v4.12.0 // indirect
 	gopkg.in/inf.v0 v0.9.1 // indirect
 	gopkg.in/yaml.v2 v2.4.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
-	k8s.io/api v0.29.2 // indirect
-	k8s.io/gengo v0.0.0-20230829151522-9cce18d56c01 // indirect
-	k8s.io/klog/v2 v2.110.1 // indirect
-	k8s.io/kube-openapi v0.0.0-20231010175941-2dd684a91f00 // indirect
+	k8s.io/api v0.31.0-alpha.2 // indirect
+	k8s.io/gengo/v2 v2.0.0-20240228010128-51d4e06bde70 // indirect
+	k8s.io/klog/v2 v2.120.1 // indirect
+	k8s.io/kube-openapi v0.0.0-20240228011516-70dd3763d340 // indirect
 	k8s.io/utils v0.0.0-20230726121419-3b25d923346b // indirect
 	sigs.k8s.io/json v0.0.0-20221116044647-bc3834ca7abd // indirect
-	sigs.k8s.io/yaml v1.3.0 // indirect
+	sigs.k8s.io/yaml v1.4.0 // indirect
 )

cluster-autoscaler/apis/provisioningrequest/autoscaling.x-k8s.io/v1beta1/types.go

@@ -197,7 +197,11 @@ const (
 	// ProvisioningClassCheckCapacity denotes that CA will check if current cluster state can fulfill this request,
 	// and reserve the capacity for a specified time.
 	ProvisioningClassCheckCapacity string = "check-capacity.autoscaling.x-k8s.io"
-	// ProvisioningClassAtomicScaleUp denotes that CA try to provision the capacity
+	// ProvisioningClassBestEffortAtomicScaleUp denotes that CA try to provision the capacity
 	// in an atomic manner.
-	ProvisioningClassAtomicScaleUp string = "atomic-scale-up.autoscaling.x-k8s.io"
+	ProvisioningClassBestEffortAtomicScaleUp string = "best-effort-atomic-scale-up.autoscaling.x-k8s.io"
+	// ProvisioningRequestPodAnnotationKey is a key used to annotate pods consuming provisioning request.
+	ProvisioningRequestPodAnnotationKey = "autoscaling.x-k8s.io/consume-provisioning-request"
+	// ProvisioningClassPodAnnotationKey is a key used to add annotation about Provisioning Class
+	ProvisioningClassPodAnnotationKey = "autoscaling.x-k8s.io/provisioning-class-name"
 )

cluster-autoscaler/cloudprovider/POLICY.md

@@ -80,7 +80,7 @@ in-tree cloudprovider follows the following rules:
       handles. There is an expectation that at least some of the owners will
       join Kubernetes organization (by following the
       [process](https://github.com/kubernetes/community/blob/master/community-membership.md))
-      within one release cycly, so that they can approve PRs to their
+      within one release cycle, so that they can approve PRs to their
       cloudprovider.
   * Cloudprovider shouldn't introduce new dependencies (such as clients/SDKs)
     to top-level go.mod vendor, unless those dependencies are already imported
@@ -109,8 +109,8 @@ maintenance request_ (CMR) mechanism.
    (ex. 1.26).
  * CMR will need to be discussed on sig-autoscaling meeting and approved by
    sig leads before being issued. It will also be announced on sig-autoscaling
-   slack channel and highlited in sig-autoscaling meeting notes.
- * A CMR may be issued no later then [enhancements
+   slack channel and highlighted in sig-autoscaling meeting notes.
+ * A CMR may be issued no later than [enhancements
    freeze](https://github.com/kubernetes/sig-release/blob/master/releases/release_phases.md#enhancements-freeze)
    of a given Kubernetes minor version.
  * If a given cloud provider was added more than one release cycle ago and there

cluster-autoscaler/cloudprovider/aws/auto_scaling_groups.go

@@ -308,46 +308,78 @@ func (m *asgCache) DeleteInstances(instances []*AwsInstanceRef) error {
 		}
 	}
 
+	placeHolderInstancesCount := m.GetPlaceHolderInstancesCount(instances)
+	// Check if there are any placeholder instances in the list.
+	if placeHolderInstancesCount > 0 {
+		// Log the check for placeholders in the ASG.
+		klog.V(4).Infof("Detected %d placeholder instance(s) in ASG %s",
+			placeHolderInstancesCount, commonAsg.Name)
+
+		asgNames := []string{commonAsg.Name}
+		asgDetail, err := m.awsService.getAutoscalingGroupsByNames(asgNames)
+
+		if err != nil {
+			klog.Errorf("Error retrieving ASG details %s: %v", commonAsg.Name, err)
+			return err
+		}
+
+		activeInstancesInAsg := len(asgDetail[0].Instances)
+		desiredCapacityInAsg := int(*asgDetail[0].DesiredCapacity)
+		klog.V(4).Infof("asg %s has placeholders instances with desired capacity = %d and active instances = %d. updating ASG to match active instances count",
+			commonAsg.Name, desiredCapacityInAsg, activeInstancesInAsg)
+
+		// If the difference between the active instances and the desired capacity is greater than 1,
+		// it means that the ASG is under-provisioned and the desired capacity is not being reached.
+		// In this case, we would reduce the size of ASG by the count of unprovisioned instances
+		// which is equal to the total count of active instances in ASG
+
+		err = m.setAsgSizeNoLock(commonAsg, activeInstancesInAsg)
+
+		if err != nil {
+			klog.Errorf("Error reducing ASG %s size to %d: %v", commonAsg.Name, activeInstancesInAsg, err)
+			return err
+		}
+	}
+
 	for _, instance := range instances {
-		// check if the instance is a placeholder - a requested instance that was never created by the node group
-		// if it is, just decrease the size of the node group, as there's no specific instance we can remove
-		if m.isPlaceholderInstance(instance) {
-			klog.V(4).Infof("instance %s is detected as a placeholder, decreasing ASG requested size instead "+
-				"of deleting instance", instance.Name)
-			m.decreaseAsgSizeByOneNoLock(commonAsg)
-		} else {
-			// check if the instance is already terminating - if it is, don't bother terminating again
-			// as doing so causes unnecessary API calls and can cause the curSize cached value to decrement
-			// unnecessarily.
-			lifecycle, err := m.findInstanceLifecycle(*instance)
-			if err != nil {
-				return err
-			}
 
-			if lifecycle != nil &&
-				*lifecycle == autoscaling.LifecycleStateTerminated ||
-				*lifecycle == autoscaling.LifecycleStateTerminating ||
-				*lifecycle == autoscaling.LifecycleStateTerminatingWait ||
-				*lifecycle == autoscaling.LifecycleStateTerminatingProceed {
-				klog.V(2).Infof("instance %s is already terminating in state %s, will skip instead", instance.Name, *lifecycle)
-				continue
-			}
+		if m.isPlaceholderInstance(instance) {
+			// skipping placeholder as placeholder instances don't exist
+			// and we have already reduced ASG size during placeholder check.
+			continue
+		}
+		// check if the instance is already terminating - if it is, don't bother terminating again
+		// as doing so causes unnecessary API calls and can cause the curSize cached value to decrement
+		// unnecessarily.
+		lifecycle, err := m.findInstanceLifecycle(*instance)
+		if err != nil {
+			return err
+		}
 
-			params := &autoscaling.TerminateInstanceInAutoScalingGroupInput{
-				InstanceId:                     aws.String(instance.Name),
-				ShouldDecrementDesiredCapacity: aws.Bool(true),
-			}
-			start := time.Now()
-			resp, err := m.awsService.TerminateInstanceInAutoScalingGroup(params)
-			observeAWSRequest("TerminateInstanceInAutoScalingGroup", err, start)
-			if err != nil {
-				return err
-			}
-			klog.V(4).Infof(*resp.Activity.Description)
+		if lifecycle != nil &&
+			*lifecycle == autoscaling.LifecycleStateTerminated ||
+			*lifecycle == autoscaling.LifecycleStateTerminating ||
+			*lifecycle == autoscaling.LifecycleStateTerminatingWait ||
+			*lifecycle == autoscaling.LifecycleStateTerminatingProceed {
+			klog.V(2).Infof("instance %s is already terminating in state %s, will skip instead", instance.Name, *lifecycle)
+			continue
+		}
 
-			// Proactively decrement the size so autoscaler makes better decisions
-			commonAsg.curSize--
+		params := &autoscaling.TerminateInstanceInAutoScalingGroupInput{
+			InstanceId:                     aws.String(instance.Name),
+			ShouldDecrementDesiredCapacity: aws.Bool(true),
+		}
+		start := time.Now()
+		resp, err := m.awsService.TerminateInstanceInAutoScalingGroup(params)
+		observeAWSRequest("TerminateInstanceInAutoScalingGroup", err, start)
+		if err != nil {
+			return err
 		}
+		klog.V(4).Infof(*resp.Activity.Description)
+
+		// Proactively decrement the size so autoscaler makes better decisions
+		commonAsg.curSize--
+
 	}
 	return nil
 }
@@ -624,3 +656,16 @@ func (m *asgCache) buildInstanceRefFromAWS(instance *autoscaling.Instance) AwsIn
 func (m *asgCache) Cleanup() {
 	close(m.interrupt)
 }
+
+// GetPlaceHolderInstancesCount returns count of placeholder instances in the cache
+func (m *asgCache) GetPlaceHolderInstancesCount(instances []*AwsInstanceRef) int {
+
+	placeholderInstancesCount := 0
+	for _, instance := range instances {
+		if strings.HasPrefix(instance.Name, placeholderInstanceNamePrefix) {
+			placeholderInstancesCount++
+
+		}
+	}
+	return placeholderInstancesCount
+}