Files
Jian QiuandClaude Sonnet 4.5 78daf0d2ae fix: skip GC for ManifestWorks managed by ManifestWorkReplicaSet (#1299)
Skip garbage collection for ManifestWorks that have the
ManifestWorkReplicaSet controller label, as these should be
managed exclusively by the ManifestWorkReplicaSet controller.

Changes:
- Fix logic bug in controller to properly check for ReplicaSet label
- Add unit tests for label-based GC skip behavior
- Add integration test to verify GC skip for ReplicaSet-managed works

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Signed-off-by: Jian Qiu <jqiu@redhat.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
2025-12-16 08:56:07 +00:00

119 lines
4.1 KiB
Go

package manifestworkgarbagecollection
import (
"context"
"fmt"
"time"
apierrors "k8s.io/apimachinery/pkg/api/errors"
"k8s.io/apimachinery/pkg/api/meta"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
utilruntime "k8s.io/apimachinery/pkg/util/runtime"
"k8s.io/client-go/tools/cache"
"k8s.io/klog/v2"
workclientset "open-cluster-management.io/api/client/work/clientset/versioned"
workinformers "open-cluster-management.io/api/client/work/informers/externalversions/work/v1"
worklisters "open-cluster-management.io/api/client/work/listers/work/v1"
workapiv1 "open-cluster-management.io/api/work/v1"
workapiv1alpha1 "open-cluster-management.io/api/work/v1alpha1"
"open-cluster-management.io/sdk-go/pkg/basecontroller/factory"
"open-cluster-management.io/ocm/pkg/common/queue"
)
// ManifestWorkGarbageCollectionController is to delete the manifestworks when it has the completed condition.
type ManifestWorkGarbageCollectionController struct {
workClient workclientset.Interface
workLister worklisters.ManifestWorkLister
}
// NewManifestWorkGarbageCollectionController creates a new ManifestWorkGarbageCollectionController
func NewManifestWorkGarbageCollectionController(
workClient workclientset.Interface,
manifestWorkInformer workinformers.ManifestWorkInformer,
) factory.Controller {
controller := &ManifestWorkGarbageCollectionController{
workClient: workClient,
workLister: manifestWorkInformer.Lister(),
}
return factory.New().
WithInformersQueueKeysFunc(
queue.QueueKeyByMetaNamespaceName,
manifestWorkInformer.Informer(),
).
WithSync(controller.sync).
ToController("ManifestWorkGarbageCollectionController")
}
// sync is the main reconcile loop for completed ManifestWork TTL
func (c *ManifestWorkGarbageCollectionController) sync(ctx context.Context, controllerContext factory.SyncContext, key string) error {
logger := klog.FromContext(ctx)
logger.V(4).Info("Reconciling ManifestWork for TTL processing", "key", key)
namespace, name, err := cache.SplitMetaNamespaceKey(key)
if err != nil {
utilruntime.HandleError(err)
return nil
}
manifestWork, err := c.workLister.ManifestWorks(namespace).Get(name)
switch {
case apierrors.IsNotFound(err):
return nil
case err != nil:
return err
}
if manifestWork.DeletionTimestamp != nil {
return nil
}
// skip if manifestwork is generated by manifesworkreplicaset, which should not be deleted by GC since
// manifestworkreplicaset controller is responsible for managing it.
if len(manifestWork.Labels) > 0 {
if _, ok := manifestWork.Labels[workapiv1alpha1.ManifestWorkReplicaSetControllerNameLabelKey]; ok {
return nil
}
}
// Check if ManifestWork has TTLSecondsAfterFinished configured
if manifestWork.Spec.DeleteOption == nil || manifestWork.Spec.DeleteOption.TTLSecondsAfterFinished == nil {
return nil
}
// Find the Complete condition
completedCondition := meta.FindStatusCondition(manifestWork.Status.Conditions, workapiv1.WorkComplete)
if completedCondition == nil || completedCondition.Status != metav1.ConditionTrue {
return nil
}
ttlSeconds := *manifestWork.Spec.DeleteOption.TTLSecondsAfterFinished
if ttlSeconds > 0 {
// Calculate time elapsed since completion
// Compute deadline precisely using durations and handle clock skew.
completedTime := completedCondition.LastTransitionTime.Time
ttl := time.Duration(ttlSeconds) * time.Second
deadline := completedTime.Add(ttl)
now := time.Now()
if now.Before(deadline) {
requeueAfter := time.Until(deadline)
logger.V(4).Info("ManifestWork completed; will be deleted after remaining TTL",
"namespace", namespace, "name", name, "remaining", requeueAfter)
controllerContext.Queue().AddAfter(key, requeueAfter)
return nil
}
}
// Time to delete the ManifestWork
logger.Info("Deleting completed ManifestWork after TTL expiry",
"namespace", namespace, "name", name, "ttlSeconds", ttlSeconds)
err = c.workClient.WorkV1().ManifestWorks(namespace).Delete(ctx, name, metav1.DeleteOptions{})
if err != nil && !apierrors.IsNotFound(err) {
return fmt.Errorf("failed to delete completed ManifestWork %s/%s: %w", namespace, name, err)
}
return nil
}