Refactored expiration/utilization controller into node controller (#594)

* Refactored expiration controller into node controller * PR comments * PR Comments
aws · Aug 4, 2021 · a53687e · a53687e
1 parent 91bb752
commit a53687e
Show file tree

Hide file tree

Showing 24 changed files with 534 additions and 849 deletions.
diff --git a/cmd/controller/main.go b/cmd/controller/main.go
@@ -24,9 +24,7 @@ import (
 	"github.com/awslabs/karpenter/pkg/cloudprovider/registry"
 	"github.com/awslabs/karpenter/pkg/controllers"
 	"github.com/awslabs/karpenter/pkg/controllers/allocation"
-	"github.com/awslabs/karpenter/pkg/controllers/expiration"
 	"github.com/awslabs/karpenter/pkg/controllers/node"
-	"github.com/awslabs/karpenter/pkg/controllers/reallocation"
 	"github.com/awslabs/karpenter/pkg/controllers/termination"
 	"github.com/go-logr/zapr"
 	"k8s.io/apimachinery/pkg/runtime"
@@ -82,9 +80,7 @@ func main() {
 		HealthProbeBindAddress: fmt.Sprintf(":%d", options.HealthProbePort),
 	})
 	if err := manager.RegisterControllers(ctx,
-		expiration.NewController(manager.GetClient()),
 		allocation.NewController(manager.GetClient(), clientSet.CoreV1(), cloudProvider),
-		reallocation.NewController(manager.GetClient(), cloudProvider),
 		termination.NewController(ctx, manager.GetClient(), clientSet.CoreV1(), cloudProvider),
 		node.NewController(manager.GetClient()),
 	).Start(ctx); err != nil {

diff --git a/pkg/apis/provisioning/v1alpha3/provisioner.go b/pkg/apis/provisioning/v1alpha3/provisioner.go
@@ -113,28 +113,20 @@ var (
 )
 
 var (
-	// Well known, supported labels
+	// Well known labels
 	ArchitectureLabelKey    = "kubernetes.io/arch"
 	OperatingSystemLabelKey = "kubernetes.io/os"
-
+	ZoneLabelKey            = "topology.kubernetes.io/zone"
+	InstanceTypeLabelKey    = "node.kubernetes.io/instance-type"
+	// Reserved labels
+	ProvisionerNameLabelKey = SchemeGroupVersion.Group + "/provisioner-name"
 	// Reserved taints
 	NotReadyTaintKey = SchemeGroupVersion.Group + "/not-ready"
-
-	// Reserved labels
-	ProvisionerNameLabelKey          = SchemeGroupVersion.Group + "/provisioner-name"
-	ProvisionerUnderutilizedLabelKey = SchemeGroupVersion.Group + "/underutilized"
-
 	// Reserved annotations
-	KarpenterDoNotEvictPodAnnotation = SchemeGroupVersion.Group + "/do-not-evict"
-	ProvisionerTTLAfterEmptyKey      = SchemeGroupVersion.Group + "/ttl-after-empty"
-
-	// Use ProvisionerSpec instead
-	ZoneLabelKey         = "topology.kubernetes.io/zone"
-	InstanceTypeLabelKey = "node.kubernetes.io/instance-type"
-
+	DoNotEvictPodAnnotationKey      = SchemeGroupVersion.Group + "/do-not-evict"
+	EmptinessTimestampAnnotationKey = SchemeGroupVersion.Group + "/emptiness-timestamp"
 	// Finalizers
 	TerminationFinalizer = SchemeGroupVersion.Group + "/termination"
-
 	// Default provisioner
 	DefaultProvisioner = types.NamespacedName{Name: "default"}
 )

diff --git a/pkg/apis/provisioning/v1alpha3/provisioner_validation.go b/pkg/apis/provisioning/v1alpha3/provisioner_validation.go
@@ -32,8 +32,7 @@ var (
 		ArchitectureLabelKey,
 		OperatingSystemLabelKey,
 		ProvisionerNameLabelKey,
-		ProvisionerUnderutilizedLabelKey,
-		ProvisionerTTLAfterEmptyKey,
+		EmptinessTimestampAnnotationKey,
 		ZoneLabelKey,
 		InstanceTypeLabelKey,
 	}

diff --git a/pkg/apis/provisioning/v1alpha3/provisioner_validation_test.go b/pkg/apis/provisioning/v1alpha3/provisioner_validation_test.go
@@ -91,7 +91,6 @@ var _ = Describe("Validation", func() {
 				ArchitectureLabelKey,
 				OperatingSystemLabelKey,
 				ProvisionerNameLabelKey,
-				ProvisionerUnderutilizedLabelKey,
 				ZoneLabelKey,
 				InstanceTypeLabelKey,
 			} {

diff --git a/pkg/cloudprovider/aws/instance.go b/pkg/cloudprovider/aws/instance.go
@@ -178,7 +178,7 @@ func (p *InstanceProvider) getInstance(ctx context.Context, id *string, instance
 	}
 	*instance = *describeInstancesOutput.Reservations[0].Instances[0]
 	if len(aws.StringValue(instance.PrivateDnsName)) == 0 {
-		return fmt.Errorf("expected PrivateDnsName to be set")
+		return fmt.Errorf("got instance %s but PrivateDnsName was not set", aws.StringValue(instance.InstanceId))
 	}
 	return nil
 }

diff --git a/pkg/controllers/allocation/controller.go b/pkg/controllers/allocation/controller.go
@@ -85,7 +85,7 @@ func (c *Controller) Reconcile(ctx context.Context, req reconcile.Request) (reco
 			logging.FromContext(ctx).Errorf("Provisioner \"%s\" not found. Create the \"default\" provisioner or specify an alternative using the nodeSelector %s", req.Name, v1alpha3.ProvisionerNameLabelKey)
 			return reconcile.Result{}, nil
 		}
-		return reconcile.Result{}, err
+		return result.RetryIfError(ctx, err)
 	}
 
 	// 2. Wait on a pod batch

diff --git a/pkg/controllers/expiration/controller.go b/pkg/controllers/expiration/controller.go
diff --git a/pkg/controllers/expiration/suite_test.go b/pkg/controllers/expiration/suite_test.go