Merge branch 'main' into bump-core

Makefile

@@ -49,6 +49,7 @@ run: ## Run Karpenter controller binary against your local cluster
 		CLUSTER_NAME=${CLUSTER_NAME} \
 		INTERRUPTION_QUEUE=${CLUSTER_NAME} \
 		FEATURE_GATES="SpotToSpotConsolidation=true" \
+		LOG_LEVEL="debug" \
 		go run ./cmd/controller/main.go
 
 test: ## Run tests

charts/karpenter/templates/service.yaml

@@ -5,10 +5,15 @@ metadata:
   namespace: {{ .Release.Namespace }}
   labels:
     {{- include "karpenter.labels" . | nindent 4 }}
-  {{- with .Values.additionalAnnotations }}
+  {{- if or .Values.additionalAnnotations .Values.service.annotations }}
   annotations:
+  {{- with .Values.additionalAnnotations }}
     {{- toYaml . | nindent 4 }}
   {{- end }}
+  {{- with .Values.service.annotations }}
+    {{- toYaml . | nindent 4 }}
+  {{- end }}
+  {{- end }}
 spec:
   type: ClusterIP
   ports:

charts/karpenter/values.yaml

@@ -12,6 +12,9 @@ additionalAnnotations: {}
 imagePullPolicy: IfNotPresent
 # -- Image pull secrets for Docker images.
 imagePullSecrets: []
+service:
+  # -- Additional annotations for the Service.
+  annotations: {}
 serviceAccount:
   # -- Specifies if a ServiceAccount should be created.
   create: true

examples/v1/max-node-lifetime.yaml

@@ -32,7 +32,7 @@ spec:
         kind: EC2NodeClass
         name: default
       expireAfter: 168h # expire nodes after 7 days = 7 * 24h
-      terminationGracePeirod: 24h # grace period after 1 day = 7 * 24h, for a max node lifetime of 8 days
+      terminationGracePeriod: 24h # grace period after 1 day = 7 * 24h, for a max node lifetime of 8 days
   disruption:
     consolidationPolicy: WhenEmpty
     consolidateAfter: 60s # scale down nodes after 60 seconds without workloads (excluding daemons)

go.mod

@@ -8,9 +8,9 @@ require (
 	github.com/avast/retry-go v3.0.0+incompatible
 	github.com/aws/aws-sdk-go v1.55.5
 	github.com/aws/aws-sdk-go-v2 v1.32.2
-	github.com/aws/aws-sdk-go-v2/config v1.27.43
+	github.com/aws/aws-sdk-go-v2/config v1.28.0
 	github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.16.17
-	github.com/aws/aws-sdk-go-v2/service/ec2 v1.182.0
+	github.com/aws/aws-sdk-go-v2/service/ec2 v1.183.0
 	github.com/aws/aws-sdk-go-v2/service/iam v1.37.2
 	github.com/aws/aws-sdk-go-v2/service/sts v1.32.2
 	github.com/aws/karpenter-provider-aws/tools/kompat v0.0.0-20240410220356-6b868db24881

go.sum

@@ -12,8 +12,8 @@ github.com/aws/aws-sdk-go v1.55.5 h1:KKUZBfBoyqy5d3swXyiC7Q76ic40rYcbqH7qjh59kzU
 github.com/aws/aws-sdk-go v1.55.5/go.mod h1:eRwEWoyTWFMVYVQzKMNHWP5/RV4xIUGMQfXQHfHkpNU=
 github.com/aws/aws-sdk-go-v2 v1.32.2 h1:AkNLZEyYMLnx/Q/mSKkcMqwNFXMAvFto9bNsHqcTduI=
 github.com/aws/aws-sdk-go-v2 v1.32.2/go.mod h1:2SK5n0a2karNTv5tbP1SjsX0uhttou00v/HpXKM1ZUo=
-github.com/aws/aws-sdk-go-v2/config v1.27.43 h1:p33fDDihFC390dhhuv8nOmX419wjOSDQRb+USt20RrU=
-github.com/aws/aws-sdk-go-v2/config v1.27.43/go.mod h1:pYhbtvg1siOOg8h5an77rXle9tVG8T+BWLWAo7cOukc=
+github.com/aws/aws-sdk-go-v2/config v1.28.0 h1:FosVYWcqEtWNxHn8gB/Vs6jOlNwSoyOCA/g/sxyySOQ=
+github.com/aws/aws-sdk-go-v2/config v1.28.0/go.mod h1:pYhbtvg1siOOg8h5an77rXle9tVG8T+BWLWAo7cOukc=
 github.com/aws/aws-sdk-go-v2/credentials v1.17.41 h1:7gXo+Axmp+R4Z+AK8YFQO0ZV3L0gizGINCOWxSLY9W8=
 github.com/aws/aws-sdk-go-v2/credentials v1.17.41/go.mod h1:u4Eb8d3394YLubphT4jLEwN1rLNq2wFOlT6OuxFwPzU=
 github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.16.17 h1:TMH3f/SCAWdNtXXVPPu5D6wrr4G5hI1rAxbcocKfC7Q=
@@ -24,8 +24,8 @@ github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.6.21 h1:6jZVETqmYCadGFvrYE
 github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.6.21/go.mod h1:1SR0GbLlnN3QUmYaflZNiH1ql+1qrSiB2vwcJ+4UM60=
 github.com/aws/aws-sdk-go-v2/internal/ini v1.8.1 h1:VaRN3TlFdd6KxX1x3ILT5ynH6HvKgqdiXoTxAF4HQcQ=
 github.com/aws/aws-sdk-go-v2/internal/ini v1.8.1/go.mod h1:FbtygfRFze9usAadmnGJNc8KsP346kEe+y2/oyhGAGc=
-github.com/aws/aws-sdk-go-v2/service/ec2 v1.182.0 h1:LaeziEhHZ/SJZYBK223QVzl3ucHvA9IP4tQMcxGrc9I=
-github.com/aws/aws-sdk-go-v2/service/ec2 v1.182.0/go.mod h1:kYXaB4FzyhEJjvrJ84oPnMElLiEAjGxxUunVW2tBSng=
+github.com/aws/aws-sdk-go-v2/service/ec2 v1.183.0 h1:LgwYvo4kycfT/UD7vjQhSVZSatxHAI41/54q9O6jljI=
+github.com/aws/aws-sdk-go-v2/service/ec2 v1.183.0/go.mod h1:kYXaB4FzyhEJjvrJ84oPnMElLiEAjGxxUunVW2tBSng=
 github.com/aws/aws-sdk-go-v2/service/fis v1.30.2 h1:qw7ZkSCy0akQJbJdIgRQaqXEHe7PrA3DHvE4VvemFJw=
 github.com/aws/aws-sdk-go-v2/service/fis v1.30.2/go.mod h1:CArS66NFuL1fBiSLVfWZV6oQjicsdViLm7Ic9Lte7x4=
 github.com/aws/aws-sdk-go-v2/service/iam v1.37.2 h1:E7vCDUFeDN8uOk8Nb2d4E1howWS1TR4HrKABXsvttIs=

pkg/providers/instance/instance.go

@@ -97,6 +97,7 @@ func (p *DefaultProvider) Create(ctx context.Context, nodeClass *v1.EC2NodeClass
 	}
 	instanceTypes, err := cloudprovider.InstanceTypes(instanceTypes).Truncate(schedulingRequirements, maxInstanceTypes)
 	if err != nil {
+		log.FromContext(ctx).Error(err, "truncating instance types")
 		return nil, fmt.Errorf("truncating instance types, %w", err)
 	}
 	tags := getTags(ctx, nodeClass, nodeClaim)
@@ -107,6 +108,7 @@ func (p *DefaultProvider) Create(ctx context.Context, nodeClass *v1.EC2NodeClass
 		fleetInstance, err = p.launchInstance(ctx, nodeClass, nodeClaim, instanceTypes, tags)
 	}
 	if err != nil {
+		log.FromContext(ctx).Error(err, "launching instance")
 		return nil, err
 	}
 	efaEnabled := lo.Contains(lo.Keys(nodeClaim.Spec.Resources.Requests), v1.ResourceEFA)

test/suites/integration/extended_resources_test.go

@@ -44,8 +44,6 @@ var _ = Describe("Extended Resources", func() {
 	})
 	It("should provision nodes for a deployment that requests nvidia.com/gpu", func() {
 		ExpectNvidiaDevicePluginCreated()
-		// TODO: jmdeal@ remove AL2 pin once AL2023 accelerated AMIs are available
-		nodeClass.Spec.AMISelectorTerms = []v1.AMISelectorTerm{{Alias: "al2@latest"}}
 		numPods := 1
 		dep := test.Deployment(test.DeploymentOptions{
 			Replicas: int32(numPods),
@@ -224,8 +222,6 @@ var _ = Describe("Extended Resources", func() {
 		}
 		// Only select private subnets since instances with multiple network instances at launch won't get a public IP.
 		nodeClass.Spec.SubnetSelectorTerms[0].Tags["Name"] = "*Private*"
-		// TODO: jmdeal@ remove AL2 pin once AL2023 accelerated AMIs are available
-		nodeClass.Spec.AMISelectorTerms = []v1.AMISelectorTerm{{Alias: "al2@latest"}}
 
 		numPods := 1
 		dep := test.Deployment(test.DeploymentOptions{

website/content/en/docs/troubleshooting.md

@@ -202,7 +202,7 @@ Karpenter does not support [in-tree storage plugins](https://kubernetes.io/blog/
 
 #### Pods were scheduled due to a race condition in Kubernetes
 
-Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `toplogySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
+Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `topologySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
 
 The following is a list of known CSI drivers which support a startupTaint to eliminate this issue:
 - [aws-ebs-csi-driver](https://github.com/kubernetes-sigs/aws-ebs-csi-driver/blob/master/docs/install.md#configure-node-startup-taint)

website/content/en/preview/troubleshooting.md

@@ -202,7 +202,7 @@ Karpenter does not support [in-tree storage plugins](https://kubernetes.io/blog/
 
 #### Pods were scheduled due to a race condition in Kubernetes
 
-Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `toplogySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
+Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `topologySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
 
 The following is a list of known CSI drivers which support a startupTaint to eliminate this issue:
 - [aws-ebs-csi-driver](https://github.com/kubernetes-sigs/aws-ebs-csi-driver/blob/master/docs/install.md#configure-node-startup-taint)

website/content/en/v0.32/troubleshooting.md

@@ -285,7 +285,7 @@ Karpenter does not support [in-tree storage plugins](https://kubernetes.io/blog/
 
 #### Pods were scheduled due to a race condition in Kubernetes
 
-Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `toplogySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
+Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `topologySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
 
 The following is a list of known CSI drivers which support a startupTaint to eliminate this issue:
 - [aws-ebs-csi-driver](https://github.com/kubernetes-sigs/aws-ebs-csi-driver/blob/master/docs/install.md#configure-node-startup-taint)

website/content/en/v0.36/troubleshooting.md

@@ -297,7 +297,7 @@ Karpenter does not support [in-tree storage plugins](https://kubernetes.io/blog/
 
 #### Pods were scheduled due to a race condition in Kubernetes
 
-Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `toplogySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
+Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `topologySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
 
 The following is a list of known CSI drivers which support a startupTaint to eliminate this issue:
 - [aws-ebs-csi-driver](https://github.com/kubernetes-sigs/aws-ebs-csi-driver/blob/master/docs/install.md#configure-node-startup-taint)

website/content/en/v0.37/troubleshooting.md

@@ -297,7 +297,7 @@ Karpenter does not support [in-tree storage plugins](https://kubernetes.io/blog/
 
 #### Pods were scheduled due to a race condition in Kubernetes
 
-Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `toplogySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
+Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `topologySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
 
 The following is a list of known CSI drivers which support a startupTaint to eliminate this issue:
 - [aws-ebs-csi-driver](https://github.com/kubernetes-sigs/aws-ebs-csi-driver/blob/master/docs/install.md#configure-node-startup-taint)

website/content/en/v1.0/troubleshooting.md

@@ -202,7 +202,7 @@ Karpenter does not support [in-tree storage plugins](https://kubernetes.io/blog/
 
 #### Pods were scheduled due to a race condition in Kubernetes
 
-Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `toplogySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
+Due to [this race condition in Kubernetes](https://github.com/kubernetes/kubernetes/issues/95911), it's possible that the scheduler and the CSINode can race during node registration such that the scheduler assumes that a node can mount more volumes than the node attachments support. There is currently no universal solve for this problem other than enforcing `topologySpreadConstraints` and `podAntiAffinity` on your workloads that use PVCs such that you attempt to reduce the number of PVCs that schedule to a given node.
 
 The following is a list of known CSI drivers which support a startupTaint to eliminate this issue:
 - [aws-ebs-csi-driver](https://github.com/kubernetes-sigs/aws-ebs-csi-driver/blob/master/docs/install.md#configure-node-startup-taint)