Upgrade to v0.36.0

.gitignore

@@ -4,6 +4,7 @@ coverage.html
 *.test
 *.cpuprofile
 *.heapprofile
+*.swp
 go.work
 go.work.sum
 

go.mod

@@ -115,3 +115,5 @@ require (
 	sigs.k8s.io/json v0.0.0-20221116044647-bc3834ca7abd // indirect
 	sigs.k8s.io/structured-merge-diff/v4 v4.4.1 // indirect
 )
+
+replace sigs.k8s.io/karpenter v0.33.1-0.20240202175636-0e77b7842c28 => github.com/smartnews/karpenter v0.34.0-sn-1

pkg/apis/crds/karpenter.sh_nodepools.yaml

@@ -135,10 +135,15 @@ spec:
                         memory leak protection, and disruption testing.
                       pattern: ^(([0-9]+(s|m|h))+)|(Never)$
                       type: string
+                    utilizationThreshold:
+                      description: |-
+                        UtilizationThreshold is defined as sum of requested resources divided by capacity
+                        below which a node can be considered for disruption.
+                      maximum: 100
+                      minimum: 1
+                      type: integer
                   type: object
                   x-kubernetes-validations:
-                    - message: consolidateAfter cannot be combined with consolidationPolicy=WhenUnderutilized
-                      rule: 'has(self.consolidateAfter) ? self.consolidationPolicy != ''WhenUnderutilized'' || self.consolidateAfter == ''Never'' : true'
                     - message: consolidateAfter must be specified with consolidationPolicy=WhenEmpty
                       rule: 'self.consolidationPolicy == ''WhenEmpty'' ? has(self.consolidateAfter) : true'
                 limits:

pkg/cache/cache.go

@@ -23,7 +23,7 @@ const (
 	// resources. Cache hits enable faster provisioning and reduced API load on
 	// AWS APIs, which can have a serious impact on performance and scalability.
 	// DO NOT CHANGE THIS VALUE WITHOUT DUE CONSIDERATION
-	DefaultTTL = time.Minute
+	DefaultTTL = 5 * time.Minute
 	// UnavailableOfferingsTTL is the time before offerings that were marked as unavailable
 	// are removed from the cache and are available for launch again
 	UnavailableOfferingsTTL = 3 * time.Minute

pkg/controllers/interruption/controller.go

@@ -24,6 +24,7 @@ import (
 	"github.com/samber/lo"
 	"go.uber.org/multierr"
 	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/client-go/util/workqueue"
 	"k8s.io/utils/clock"
 	"knative.dev/pkg/logging"
@@ -197,13 +198,31 @@ func (c *Controller) handleNodeClaim(ctx context.Context, msg messages.Message,
 		if zone != "" && instanceType != "" {
 			c.unavailableOfferingsCache.MarkUnavailable(ctx, string(msg.Kind()), instanceType, zone, v1beta1.CapacityTypeSpot)
 		}
+		// try to create a new nodeclaim immediately but ignore error if it fails
+		if err := c.createNodeClaim(ctx, nodeClaim); err != nil {
+			logging.FromContext(ctx).Errorf("[interruption handling]failed to create a new nodeclaim, %v", err)
+		}
 	}
 	if action != NoAction {
 		return c.deleteNodeClaim(ctx, nodeClaim, node)
 	}
 	return nil
 }
 
+// createNodeClaim creates a new NodeClaim with the same spec of the interrupted one
+func (c *Controller) createNodeClaim(ctx context.Context, oldNodeClaim *v1beta1.NodeClaim) error {
+	newNodeClaim := &v1beta1.NodeClaim{
+		ObjectMeta: metav1.ObjectMeta{
+			GenerateName:    oldNodeClaim.ObjectMeta.GenerateName,
+			Annotations:     oldNodeClaim.ObjectMeta.Annotations,
+			Labels:          oldNodeClaim.ObjectMeta.Labels,
+			OwnerReferences: oldNodeClaim.ObjectMeta.OwnerReferences,
+		},
+		Spec: oldNodeClaim.Spec,
+	}
+	return c.kubeClient.Create(ctx, newNodeClaim)
+}
+
 // deleteNodeClaim removes the NodeClaim from the api-server
 func (c *Controller) deleteNodeClaim(ctx context.Context, nodeClaim *v1beta1.NodeClaim, node *v1.Node) error {
 	if !nodeClaim.DeletionTimestamp.IsZero() {

pkg/providers/instance/instance.go

@@ -257,6 +257,7 @@ func getTags(ctx context.Context, nodeClass *v1beta1.EC2NodeClass, nodeClaim *co
 		corev1beta1.NodePoolLabelKey:       nodeClaim.Labels[corev1beta1.NodePoolLabelKey],
 		corev1beta1.ManagedByAnnotationKey: options.FromContext(ctx).ClusterName,
 		v1beta1.LabelNodeClass:             nodeClass.Name,
+		"Componenet":                       nodeClaim.Labels[corev1beta1.NodePoolLabelKey], // used for aws explore
 	}
 	return lo.Assign(nodeClass.Spec.Tags, staticTags)
 }