Support running addon resizer on the control plane

addon-resizer/go.mod

@@ -4,6 +4,7 @@ go 1.21
 
 require (
 	github.com/golang/glog v1.2.1
+	github.com/google/go-cmp v0.6.0
 	github.com/prometheus/client_model v0.6.1
 	github.com/prometheus/common v0.52.3
 	github.com/spf13/pflag v1.0.5
@@ -16,6 +17,7 @@ require (
 require (
 	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/emicklei/go-restful/v3 v3.11.0 // indirect
+	github.com/evanphx/json-patch v4.12.0+incompatible // indirect
 	github.com/go-logr/logr v1.3.0 // indirect
 	github.com/go-openapi/jsonpointer v0.19.6 // indirect
 	github.com/go-openapi/jsonreference v0.20.2 // indirect
@@ -33,6 +35,7 @@ require (
 	github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd // indirect
 	github.com/modern-go/reflect2 v1.0.2 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
+	github.com/pkg/errors v0.9.1 // indirect
 	golang.org/x/net v0.23.0 // indirect
 	golang.org/x/oauth2 v0.18.0 // indirect
 	golang.org/x/sys v0.18.0 // indirect

addon-resizer/go.sum

@@ -4,6 +4,8 @@ github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/emicklei/go-restful/v3 v3.11.0 h1:rAQeMHw1c7zTmncogyy8VvRZwtkmkZ4FxERmMY4rD+g=
 github.com/emicklei/go-restful/v3 v3.11.0/go.mod h1:6n3XBCmQQb25CM2LCACGz8ukIrRry+4bhvbpWn3mrbc=
+github.com/evanphx/json-patch v4.12.0+incompatible h1:4onqiflcdA9EOZ4RxV643DvftH5pOlLGNtQ5lPWQu84=
+github.com/evanphx/json-patch v4.12.0+incompatible/go.mod h1:50XU6AFN0ol/bzJsmQLiYLvXMP4fmwYFNcr97nuDLSk=
 github.com/go-logr/logr v1.3.0 h1:2y3SDp0ZXuc6/cjLSZ+Q3ir+QB9T/iG5yYRXqsagWSY=
 github.com/go-logr/logr v1.3.0/go.mod h1:9T104GzyrTigFIr8wt5mBrctHMim0Nb2HLGrmQ40KvY=
 github.com/go-openapi/jsonpointer v0.19.6 h1:eCs3fxoIi3Wh6vtgmLTOjdhSpiqphQ+DaPn38N2ZdrE=
@@ -63,6 +65,8 @@ github.com/onsi/ginkgo/v2 v2.13.0 h1:0jY9lJquiL8fcf3M4LAXN5aMlS/b2BV86HFFPCPMgE4
 github.com/onsi/ginkgo/v2 v2.13.0/go.mod h1:TE309ZR8s5FsKKpuB1YAQYBzCaAfUgatB/xlT/ETL/o=
 github.com/onsi/gomega v1.29.0 h1:KIA/t2t5UBzoirT4H9tsML45GEbo3ouUnBHsCfD2tVg=
 github.com/onsi/gomega v1.29.0/go.mod h1:9sxs+SwGrKI0+PWe4Fxa9tFQQBG5xSsSbMXOI8PPpoQ=
+github.com/pkg/errors v0.9.1 h1:FEBLx1zS214owpjy7qsBeixbURkuhQAwrK5UwLGTwt4=
+github.com/pkg/errors v0.9.1/go.mod h1:bwawxfHBFNV+L2hUp1rHADufV3IMtnDRdf1r5NINEl0=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/prometheus/client_model v0.6.1 h1:ZKSh/rekM+n3CeS952MLRAdFwIKqeY8b62p8ais2e9E=

addon-resizer/nanny/election.go

@@ -18,14 +18,14 @@ package nanny
 
 import (
 	"context"
+	"github.com/golang/glog"
 	"os"
 	"time"
-	"github.com/golang/glog"
 
 	corev1 "k8s.io/api/core/v1"
-	typedcorev1 "k8s.io/client-go/kubernetes/typed/core/v1"
 	"k8s.io/client-go/kubernetes"
 	"k8s.io/client-go/kubernetes/scheme"
+	typedcorev1 "k8s.io/client-go/kubernetes/typed/core/v1"
 	"k8s.io/client-go/tools/leaderelection"
 	"k8s.io/client-go/tools/leaderelection/resourcelock"
 	"k8s.io/client-go/tools/record"
@@ -37,8 +37,8 @@ type Config struct {
 	// wait to force acquire leadership. This is measured against time of
 	// last observed ack.
 	LeaseDuration time.Duration
-	// RenewDeadline is the duration that the acting master will retry
-	// refreshing leadership before giving up.
+	// RenewDeadline is the duration that the acting control plane will
+	// retry refreshing leadership before giving up.
 	RenewDeadline time.Duration
 	// RetryPeriod is the duration to wait between tries of actions.
 	RetryPeriod time.Duration
@@ -59,7 +59,7 @@ func LeadOrDie(cfg Config, clientset kubernetes.Interface, doLead func()) {
 	}
 
 	lock, err := resourcelock.New(
-		resourcelock.EndpointsLeasesResourceLock,
+		resourcelock.LeasesResourceLock,
 		cfg.SystemNamespace,
 		"addon-resizer",
 		clientset.CoreV1(),

addon-resizer/nanny/estimator.go

@@ -43,6 +43,8 @@ func (e LinearEstimator) scale(clusterSize uint64) *corev1.ResourceRequirements
 	return calculateResources(clusterSize, e.Resources)
 }
 
+// updatedResourceEstimator creates a copy of original estimator
+// with updated resources.
 func (e LinearEstimator) updatedResourceEstimator(resources []Resource) ResourceEstimator {
 	newEstimator := LinearEstimator{
 		Resources: resources,
@@ -75,9 +77,11 @@ func (e ExponentialEstimator) scale(clusterSize uint64) *corev1.ResourceRequirem
 	return calculateResources(n, e.Resources)
 }
 
+// updatedResourceEstimator creates a copy of original estimator
+// with updated resources.
 func (e ExponentialEstimator) updatedResourceEstimator(resources []Resource) ResourceEstimator {
 	newEstimator := ExponentialEstimator{
-		Resources:       resources,
+		Resources:      resources,
 		MinClusterSize: e.MinClusterSize,
 		ScaleFactor:    e.ScaleFactor,
 	}
@@ -121,4 +125,4 @@ func calculateResources(clusterSize uint64, resources []Resource) *corev1.Resour
 		Limits:   limits,
 		Requests: requests,
 	}
-}
+}

addon-resizer/nanny/kubernetes_client.go

@@ -48,13 +48,12 @@ const (
 )
 
 type kubernetesClient struct {
-	namespace   string
-	deployment  string
-	pod         string
-	container   string
-	clientset   *kubernetes.Clientset
-	useMetrics  bool
-	runOnMaster bool
+	namespace  string
+	deployment string
+	pod        string
+	container  string
+	clientset  *kubernetes.Clientset
+	useMetrics bool
 }
 
 // CountNodes returns the number of nodes in the cluster:
@@ -185,25 +184,26 @@ func (k *kubernetesClient) countResourcesThroughMetrics(resourceName string) (ui
 }
 
 func (k *kubernetesClient) ContainerResources() (*corev1.ResourceRequirements, error) {
-	if k.runOnMaster {
+	var containers []corev1.Container
+	// When addon resizer runs in a different pod, it cannot get pod name from env,
+	// so the pod name will be empty. In that case, get container information
+	// from the deployment instead.
+	if k.pod == "" {
 		dep, err := k.clientset.AppsV1().Deployments(k.namespace).Get(context.Background(), k.deployment, metav1.GetOptions{})
 		if err != nil {
 			return nil, err
 		}
-		for _, container := range dep.Spec.Template.Spec.Containers {
-			if container.Name == k.container {
-				return &container.Resources, nil
-			}
-		}
+		containers = dep.Spec.Template.Spec.Containers
 	} else {
 		pod, err := k.clientset.CoreV1().Pods(k.namespace).Get(context.Background(), k.pod, metav1.GetOptions{})
 		if err != nil {
 			return nil, err
 		}
-		for _, container := range pod.Spec.Containers {
-			if container.Name == k.container {
-				return &container.Resources, nil
-			}
+		containers = pod.Spec.Containers
+	}
+	for _, container := range containers {
+		if container.Name == k.container {
+			return &container.Resources, nil
 		}
 	}
 	return nil, fmt.Errorf("Container %s was not found in deployment %s in namespace %s.", k.container, k.deployment, k.namespace)
@@ -275,15 +275,14 @@ func getPodSelectorExcludingDonePodsOrDie() string {
 }
 
 // NewKubernetesClient gives a KubernetesClient with the given dependencies.
-func NewKubernetesClient(namespace, deployment, pod, container string, clientset *kubernetes.Clientset, useMetrics, runOnMaster bool) KubernetesClient {
+func NewKubernetesClient(namespace, deployment, pod, container string, clientset *kubernetes.Clientset, useMetrics bool) KubernetesClient {
 	result := &kubernetesClient{
-		namespace:   namespace,
-		deployment:  deployment,
-		pod:         pod,
-		container:   container,
-		clientset:   clientset,
-		useMetrics:  useMetrics,
-		runOnMaster: runOnMaster,
+		namespace:  namespace,
+		deployment: deployment,
+		pod:        pod,
+		container:  container,
+		clientset:  clientset,
+		useMetrics: useMetrics,
 	}
 	return result
-}
+}