Merge pull request #16213 from rifelpet/targetlifecyclestate

aws: Use instance metadata to get warm pool state

pkg/model/iam/iam_builder.go

@@ -1116,6 +1116,9 @@ func addASLifecyclePolicies(p *Policy, enableHookSupport bool) {
 			"autoscaling:DescribeLifecycleHooks",
 		)
 	}
+	// TODO: remove this after k8s 1.29 support is removed
+	// It is no longer needed as of kops 1.29 but to prevent node bootstrap issues
+	// during kops upgrades we keep the permission until it is guaranteed to not be needed.
 	p.unconditionalAction.Insert(
 		"autoscaling:DescribeAutoScalingInstances",
 	)

upup/pkg/fi/nodeup/command.go

@@ -32,6 +32,7 @@ import (
 	"time"
 
 	"github.com/aws/aws-sdk-go/aws"
+	"github.com/aws/aws-sdk-go/aws/awserr"
 	"github.com/aws/aws-sdk-go/aws/ec2metadata"
 	"github.com/aws/aws-sdk-go/aws/session"
 	"github.com/aws/aws-sdk-go/service/autoscaling"
@@ -245,10 +246,13 @@ func (c *NodeUpCommand) Run(out io.Writer) error {
 		}
 		modelContext.InstanceID = string(instanceIDBytes)
 
+		// Check if WarmPool is enabled first, to avoid additional API calls
+		if len(modelContext.NodeupConfig.WarmPoolImages) > 0 {
 			modelContext.ConfigurationMode, err = getAWSConfigurationMode(ctx, modelContext)
 			if err != nil {
 				return err
 			}
+		}
 
 		modelContext.MachineType, err = getMachineType()
 		if err != nil {
@@ -712,6 +716,11 @@ func getNodeConfigFromServers(ctx context.Context, bootConfig *nodeup.BootConfig
 }
 
 func getAWSConfigurationMode(ctx context.Context, c *model.NodeupModelContext) (string, error) {
+	// Check if WarmPool is enabled first, to avoid additional API calls
+	if len(c.NodeupConfig.WarmPoolImages) == 0 {
+		return "", nil
+	}
+
 	// Only worker nodes and apiservers can actually autoscale.
 	// We are not adding describe permissions to the other roles
 	role := c.BootConfig.InstanceGroupRole
@@ -719,20 +728,17 @@ func getAWSConfigurationMode(ctx context.Context, c *model.NodeupModelContext) (
 		return "", nil
 	}
 
-	svc := c.Cloud.(awsup.AWSCloud).Autoscaling()
+	targetLifecycleState, err := vfs.Context.ReadFile("metadata://aws/meta-data/autoscaling/target-lifecycle-state")
-
-	result, err := svc.DescribeAutoScalingInstancesWithContext(ctx, &autoscaling.DescribeAutoScalingInstancesInput{
-		InstanceIds: []*string{&c.InstanceID},
-	})
 	if err != nil {
-		return "", fmt.Errorf("error describing instances: %v", err)
+		var awsErr awserr.RequestFailure
-	}
+		if errors.As(err, &awsErr) && awsErr.StatusCode() == 404 {
-	// If the instance is not a part of an ASG, it won't be in a warm pool either.
+			// The instance isn't in an ASG (karpenter, etc.)
-	if len(result.AutoScalingInstances) < 1 {
 			return "", nil
 		}
-	lifecycle := fi.ValueOf(result.AutoScalingInstances[0].LifecycleState)
+		return "", fmt.Errorf("error reading target-lifecycle-state from instance metadata: %v", err)
-	if strings.HasPrefix(lifecycle, "Warmed:") {
+	}
+
+	if strings.HasPrefix(string(targetLifecycleState), "Warmed:") {
 		klog.Info("instance is entering warm pool")
 		return model.ConfigurationModeWarming, nil
 	} else {