fix: Mutex not being released on step completion

workflow/controller/operator.go

@@ -1691,6 +1691,9 @@ func (woc *wfOperationCtx) executeTemplate(nodeName string, orgTmpl wfv1.Templat
 					woc.computeMetrics(processedTmpl.Metrics.Prometheus, localScope, realTimeScope, false)
 				}
 			}
+			if processedTmpl.Synchronization != nil {
+				woc.controller.syncManager.Release(woc.wf, node.ID, processedTmpl.Synchronization)
+			}
 			return retryParentNode, nil
 		}
 		lastChildNode := getChildNodeIndex(retryParentNode, woc.wf.Status.Nodes, -1)

workflow/controller/operator_concurrency_test.go

@@ -392,3 +392,86 @@ func TestMutexInDAG(t *testing.T) {
 		}
 	})
 }
+
+const RetryWfWithSemaphore = `
+apiVersion: argoproj.io/v1alpha1
+kind: Workflow
+metadata: 
+  name: script-wf
+  namespace: default
+spec: 
+  entrypoint: step1
+  retryStrategy:
+    limit: 10 
+  templates:
+    - name: step1
+      steps:
+        - - name: hello1
+            template: whalesay
+        - - name: hello2
+            template: whalesay 
+    - name: whalesay
+      daemon: true
+      synchronization: 
+        semaphore: 
+          configMapKeyRef: 
+            key: template
+            name: my-config
+      container: 
+        args: 
+          - "hello world"
+        command: 
+          - cowsay
+        image: "docker/whalesay:latest"
+`
+
+func TestSynchronizationWithRetry(t *testing.T) {
+	assert := assert.New(t)
+	cancel, controller := newController()
+	defer cancel()
+	controller.syncManager = sync.NewLockManager(GetSyncLimitFunc(controller.kubeclientset), func(key string) {
+	})
+	var cm v1.ConfigMap
+	err := yaml.Unmarshal([]byte(configMap), &cm)
+	assert.NoError(err)
+	_, err = controller.kubeclientset.CoreV1().ConfigMaps("default").Create(&cm)
+	assert.NoError(err)
+	t.Run("WorkflowWithRetry", func(t *testing.T) {
+		wf := unmarshalWF(RetryWfWithSemaphore)
+		wf, err := controller.wfclientset.ArgoprojV1alpha1().Workflows(wf.Namespace).Create(wf)
+		assert.NoError(err)
+		woc := newWorkflowOperationCtx(wf, controller)
+		woc.operate()
+		for _, node := range woc.wf.Status.Nodes {
+			if node.Name == "hello1" {
+				assert.Equal(wfv1.NodePending, node.Phase)
+			}
+		}
+
+		// Updating Pod state
+		makePodsPhase(woc, v1.PodSucceeded)
+
+		// Release the lock from hello1
+		woc = newWorkflowOperationCtx(woc.wf, controller)
+		woc.operate()
+		for _, node := range woc.wf.Status.Nodes {
+			if node.Name == "hello1" {
+				assert.Equal(wfv1.NodeSucceeded, node.Phase)
+			}
+			if node.Name == "hello2" {
+				assert.Equal(wfv1.NodePending, node.Phase)
+			}
+		}
+		// Updating Pod state
+		makePodsPhase(woc, v1.PodSucceeded)
+
+		// Release the lock  from hello2
+		woc = newWorkflowOperationCtx(woc.wf, controller)
+		woc.operate()
+		// Nobody is waiting for the lock
+		assert.Empty(woc.wf.Status.Synchronization.Semaphore.Waiting)
+		// Nobody is holding the lock
+		assert.Empty(woc.wf.Status.Synchronization.Semaphore.Holding[0].Holders)
+
+	})
+}

workflow/sync/sync_manager.go

@@ -155,6 +155,7 @@ func (cm *Manager) Release(wf *wfv1.Workflow, nodeName string, syncRef *wfv1.Syn
 
 	if syncLockHolder, ok := cm.syncLockMap[lockName.EncodeName()]; ok {
 		syncLockHolder.release(holderKey)
+		syncLockHolder.removeFromQueue(holderKey)
 		log.Debugf("%s sync lock is released by %s", lockName.EncodeName(), holderKey)
 		lockKey := lockName.EncodeName()
 		wf.Status.Synchronization.GetStatus(syncRef.GetType()).LockReleased(holderKey, lockKey)