feat: Add support for deletion delay when using PodGC (#6168)

Signed-off-by: Stefan Sedich <[email protected]>

config/config.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 	"math"
 	"path"
+	"time"
 
 	apiv1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@@ -106,12 +107,16 @@ type Config struct {
 	// PodSpecLogStrategy enables the logging of podspec on controller log.
 	PodSpecLogStrategy PodSpecLogStrategy `json:"podSpecLogStrategy,omitempty"`
 
-	// PodGCGracePeriodSeconds specifies the duration in seconds before the pods in the GC queue get deleted.
-	// Value must be non-negative integer. A zero value indicates that the pods will be deleted immediately
-	// as soon as they arrived in the pod GC queue.
-	// Defaults to 30 seconds.
+	// PodGCGracePeriodSeconds specifies the duration in seconds before a terminating pod is forcefully killed.
+	// Value must be non-negative integer. A zero value indicates that the pod will be forcefully terminated immediately.
+	// Defaults to the Kubernetes default of 30 seconds.
 	PodGCGracePeriodSeconds *int64 `json:"podGCGracePeriodSeconds,omitempty"`
 
+	// PodGCDeleteDelayDuration specifies the duration in seconds before the pods in the GC queue get deleted.
+	// Value must be non-negative integer. A zero value indicates that the pods will be deleted immediately.
+	// Defaults to 5 seconds.
+	PodGCDeleteDelayDuration *metav1.Duration `json:"podGCDeleteDelayDuration,omitempty"`
+
 	// WorkflowRestrictions restricts the controller to executing Workflows that meet certain restrictions
 	WorkflowRestrictions *WorkflowRestrictions `json:"workflowRestrictions,omitempty"`
 
@@ -144,6 +149,14 @@ func (c Config) GetResourceRateLimit() ResourceRateLimit {
 	}
 }
 
+func (c Config) GetPodGCDeleteDelayDuration() time.Duration {
+	if c.PodGCDeleteDelayDuration == nil {
+		return 5 * time.Second
+	}
+
+	return c.PodGCDeleteDelayDuration.Duration
+}
+
 // PodSpecLogStrategy contains the configuration for logging the pod spec in controller log for debugging purpose
 type PodSpecLogStrategy struct {
 	FailedPod bool `json:"failedPod,omitempty"`

test/e2e/pod_cleanup_test.go

@@ -21,7 +21,7 @@ type PodCleanupSuite struct {
 	fixtures.E2ESuite
 }
 
-const enoughTimeForPodCleanup = 3 * time.Second
+const enoughTimeForPodCleanup = 10 * time.Second
 
 func (s *PodCleanupSuite) TestNone() {
 	s.Given().

workflow/controller/controller.go

@@ -768,7 +768,8 @@ func (wfc *WorkflowController) processNextItem(ctx context.Context) bool {
 		}
 		if doPodGC {
 			for podName := range woc.completedPods {
-				woc.controller.queuePodForCleanup(woc.wf.Namespace, podName, deletePod)
+				delay := woc.controller.Config.GetPodGCDeleteDelayDuration()
+				woc.controller.queuePodForCleanupAfter(woc.wf.Namespace, podName, deletePod, delay)
 			}
 		}
 	}

workflow/controller/operator.go

@@ -596,10 +596,12 @@ func (woc *wfOperationCtx) persistUpdates(ctx context.Context) {
 			switch woc.execWf.Spec.PodGC.Strategy {
 			case wfv1.PodGCOnPodSuccess:
 				if podPhase == apiv1.PodSucceeded {
-					woc.controller.queuePodForCleanup(woc.wf.Namespace, podName, deletePod)
+					delay := woc.controller.Config.GetPodGCDeleteDelayDuration()
+					woc.controller.queuePodForCleanupAfter(woc.wf.Namespace, podName, deletePod, delay)
 				}
 			case wfv1.PodGCOnPodCompletion:
-				woc.controller.queuePodForCleanup(woc.wf.Namespace, podName, deletePod)
+				delay := woc.controller.Config.GetPodGCDeleteDelayDuration()
+				woc.controller.queuePodForCleanupAfter(woc.wf.Namespace, podName, deletePod, delay)
 			}
 		} else {
 			// label pods which will not be deleted