fix descheduler object limiter with multiple profiles

Signed-off-by: songtao98 <songtao2603060@gmail.com>
koordinator-sh · Sep 5, 2024 · 8a0c4bb · 8a0c4bb
1 parent 36cb609
commit 8a0c4bb
Show file tree

Hide file tree

Showing 2 changed files with 21 additions and 2 deletions.
diff --git a/pkg/descheduler/controllers/migration/controller.go b/pkg/descheduler/controllers/migration/controller.go
@@ -87,6 +87,8 @@ type Reconciler struct {
 	limiterMap      map[deschedulerconfig.MigrationLimitObjectType]map[string]*rate.Limiter
 	limiterCacheMap map[deschedulerconfig.MigrationLimitObjectType]*gocache.Cache
 	limiterLock     sync.Mutex
+
+	reconcilerUID types.UID
 }
 
 func New(args runtime.Object, handle framework.Handle) (framework.Plugin, error) {
@@ -136,6 +138,7 @@ func New(args runtime.Object, handle framework.Handle) (framework.Plugin, error)
 	if err = c.Watch(source.Kind(options.Manager.GetCache(), r.reservationInterpreter.GetReservationType()), &handler.Funcs{}); err != nil {
 		return nil, err
 	}
+	r.reconcilerUID = UUIDGenerateFn()
 	return r, nil
 }
 
@@ -242,6 +245,16 @@ func (r *Reconciler) Reconcile(ctx context.Context, request reconcile.Request) (
 		return reconcile.Result{}, nil
 	}
 
+	// handle the case that the job is created by other reconciler
+	// NOTE:
+	// 1. only the PMJs created by MigrationController have annotation AnnotationJobCreatedBy, if a PMJ was created
+	//    manually by users, we should handle it anyway.
+	// 2. if a PMJ was indeed created by MigrationController but koord-descheduler has restarted(new UUID generated),
+	//    we just ignore it.
+	if jobUID, ok := job.Annotations[AnnotationJobCreatedBy]; ok && jobUID != string(r.reconcilerUID) {
+		return reconcile.Result{}, nil
+	}
+
 	result, err := r.doMigrate(ctx, job)
 	if err != nil {
 		klog.Errorf("Failed to reconcile MigrationJob %v, err: %v", request.NamespacedName, err)

diff --git a/pkg/descheduler/controllers/migration/evict.go b/pkg/descheduler/controllers/migration/evict.go
@@ -21,6 +21,7 @@ import (
 
 	corev1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/types"
 	"k8s.io/klog/v2"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
@@ -30,6 +31,10 @@ import (
 	"github.com/koordinator-sh/koordinator/pkg/descheduler/framework"
 )
 
+const (
+	AnnotationJobCreatedBy = "koordinator.sh/job-created-by"
+)
+
 // Evict evicts a pod
 func (r *Reconciler) Evict(ctx context.Context, pod *corev1.Pod, evictOptions framework.EvictOptions) bool {
 	framework.FillEvictOptionsFromContext(ctx, &evictOptions)
@@ -49,11 +54,11 @@ func (r *Reconciler) Evict(ctx context.Context, pod *corev1.Pod, evictOptions fr
 		return false
 	}
 
-	err := CreatePodMigrationJob(ctx, pod, evictOptions, r.Client, r.args)
+	err := CreatePodMigrationJob(ctx, pod, evictOptions, r.Client, r.args, r.reconcilerUID)
 	return err == nil
 }
 
-func CreatePodMigrationJob(ctx context.Context, pod *corev1.Pod, evictOptions framework.EvictOptions, client client.Client, args *deschedulerconfig.MigrationControllerArgs) error {
+func CreatePodMigrationJob(ctx context.Context, pod *corev1.Pod, evictOptions framework.EvictOptions, client client.Client, args *deschedulerconfig.MigrationControllerArgs, reconcilerUID types.UID) error {
 	if evictOptions.DeleteOptions == nil {
 		evictOptions.DeleteOptions = args.DefaultDeleteOptions
 	}
@@ -63,6 +68,7 @@ func CreatePodMigrationJob(ctx context.Context, pod *corev1.Pod, evictOptions fr
 			Annotations: map[string]string{
 				evictor.AnnotationEvictReason:  evictOptions.Reason,
 				evictor.AnnotationEvictTrigger: evictOptions.PluginName,
+				AnnotationJobCreatedBy:         string(reconcilerUID),
 			},
 		},
 		Spec: sev1alpha1.PodMigrationJobSpec{