fixes #423: Populate pod list in webhook notification (#464)

* fixes #423: Populate pod list in webhook notification

* fixes #423: stop PostDrainTask incase cordon/drain fails

* fixes #423: initialize drainHelper.Client even if dry-run is set to true

* fixes #423: Not fetching pod list if dry-run is set to true

Co-authored-by: Abhay Verma <[email protected]>

Author: vermaabhay

cmd/node-termination-handler.go

@@ -318,32 +318,35 @@ func drainOrCordonIfNecessary(interruptionEventStore *interruptioneventstore.Sto
 		runPreDrainTask(node, nodeName, drainEvent, metrics, recorder)
 	}
 
+        podNameList, err := node.FetchPodNameList(nodeName)
+        if err != nil {
+                log.Err(err).Msgf("Unable to fetch running pods for node '%s' ", nodeName)
+        }
+        drainEvent.Pods = podNameList
+        err = node.LogPods(podNameList, nodeName)
+        if err != nil {
+                log.Err(err).Msg("There was a problem while trying to log all pod names on the node")
+        }
+
 	if nthConfig.CordonOnly || (drainEvent.IsRebalanceRecommendation() && !nthConfig.EnableRebalanceDraining) {
 		err = cordonNode(node, nodeName, drainEvent, metrics, recorder)
 	} else {
 		err = cordonAndDrainNode(node, nodeName, metrics, recorder, nthConfig.EnableSQSTerminationDraining)
 	}
-
-	if err != nil {
-		podNameList, err := node.FetchPodNameList(nodeName)
-		drainEvent.Pods = podNameList
-		if err != nil {
-			log.Err(err).Msgf("Unable to fetch running pods for node '%s' ", nodeName)
-		}
-		err = node.LogPods(podNameList, nodeName)
-		if err != nil {
-			log.Err(err).Msg("There was a problem while trying to log all pod names on the node")
-		}
-	}
-
-	interruptionEventStore.MarkAllAsProcessed(nodeName)
+        
 	if nthConfig.WebhookURL != "" {
 		webhook.Post(nodeMetadata, drainEvent, nthConfig)
 	}
-	if drainEvent.PostDrainTask != nil {
-		runPostDrainTask(node, nodeName, drainEvent, metrics, recorder)
-	}
-	<-interruptionEventStore.Workers
+
+        if err != nil {
+                <-interruptionEventStore.Workers
+        } else {
+	        interruptionEventStore.MarkAllAsProcessed(nodeName)
+	        if drainEvent.PostDrainTask != nil {
+		        runPostDrainTask(node, nodeName, drainEvent, metrics, recorder)
+	        }
+	        <-interruptionEventStore.Workers
+        }
 
 }
 

pkg/node/node.go

@@ -496,6 +496,10 @@ func (n Node) fetchKubernetesNode(nodeName string) (*corev1.Node, error) {
 }
 
 func (n Node) fetchAllPods(nodeName string) (*corev1.PodList, error) {
+	if n.nthConfig.DryRun {
+		log.Info().Msgf("Would have retrieved running pod list on node %s, but dry-run flag was set", nodeName)
+		return &corev1.PodList{}, nil
+	}
 	return n.drainHelper.Client.CoreV1().Pods("").List(context.TODO(), metav1.ListOptions{
 		FieldSelector: "spec.nodeName=" + nodeName,
 	})
@@ -513,6 +517,7 @@ func getDrainHelper(nthConfig config.Config) (*drain.Helper, error) {
 		Out:                 log.Logger,
 		ErrOut:              log.Logger,
 	}
+
 	if nthConfig.DryRun {
 		return drainHelper, nil
 	}