karmada-metrics-adapter: reduce memory usage

When there is a large amount of pod usage in the member cluster, metrics-adapter will consume a lot of memory. The reason is that it caches all the information of all pods in the cluster. However, we don't need all this information, so we trim some of the information to reduce memory usage.

Signed-off-by: chaunceyjiang <[email protected]>

pkg/metricsadapter/controller.go

@@ -37,6 +37,7 @@ import (
 	"github.com/karmada-io/karmada/pkg/metricsadapter/provider"
 	"github.com/karmada-io/karmada/pkg/util"
 	"github.com/karmada-io/karmada/pkg/util/fedinformer/genericmanager"
+	"github.com/karmada-io/karmada/pkg/util/fedinformer/typedmanager"
 	"github.com/karmada-io/karmada/pkg/util/gclient"
 )
 
@@ -49,7 +50,7 @@ var (
 type MetricsController struct {
 	InformerFactory       informerfactory.SharedInformerFactory
 	ClusterLister         clusterlister.ClusterLister
-	InformerManager       genericmanager.MultiClusterInformerManager
+	InformerManager       typedmanager.MultiClusterInformerManager
 	MultiClusterDiscovery multiclient.MultiClusterDiscoveryInterface
 	queue                 workqueue.RateLimitingInterface
 	restConfig            *rest.Config
@@ -62,7 +63,7 @@ func NewMetricsController(restConfig *rest.Config, factory informerfactory.Share
 		InformerFactory:       factory,
 		ClusterLister:         clusterLister,
 		MultiClusterDiscovery: multiclient.NewMultiClusterDiscoveryClient(clusterLister, kubeFactory),
-		InformerManager:       genericmanager.GetInstance(),
+		InformerManager:       typedmanager.GetInstance(),
 		restConfig:            restConfig,
 		queue: workqueue.NewRateLimitingQueueWithConfig(workqueue.DefaultControllerRateLimiter(), workqueue.RateLimitingQueueConfig{
 			Name: "metrics-adapter",
@@ -171,6 +172,7 @@ func (m *MetricsController) handleClusters() bool {
 	if !m.InformerManager.IsManagerExist(clusterName) {
 		klog.Info("Try to build informer manager for cluster ", clusterName)
 		controlPlaneClient := gclient.NewForConfigOrDie(m.restConfig)
+		clusterClient, err := util.NewClusterClientSet(clusterName, controlPlaneClient, nil)
 		clusterDynamicClient, err := util.NewClusterDynamicClientSet(clusterName, controlPlaneClient)
 		if err != nil {
 			return false
@@ -181,7 +183,7 @@ func (m *MetricsController) handleClusters() bool {
 			klog.Warningf("unable to access cluster %s, Error: %+v", clusterName, err)
 			return true
 		}
-		_ = m.InformerManager.ForCluster(clusterName, clusterDynamicClient.DynamicClientSet, 0)
+		_ = m.InformerManager.ForCluster(clusterName, clusterClient.KubeClient, 0)
 	}
 	err = m.MultiClusterDiscovery.Set(clusterName)
 	if err != nil {
@@ -190,8 +192,8 @@ func (m *MetricsController) handleClusters() bool {
 	}
 	sci := m.InformerManager.GetSingleClusterManager(clusterName)
 	// Just trigger the informer to work
-	_ = sci.Lister(provider.PodsGVR)
-	_ = sci.Lister(provider.NodesGVR)
+	_, _ = sci.Lister(provider.PodsGVR)
+	_, _ = sci.Lister(provider.NodesGVR)
 
 	sci.Start()
 	_ = sci.WaitForCacheSync()