From 53db54ce4990d73993f69201a70d81305fc55113 Mon Sep 17 00:00:00 2001
From: Mohammed <abdulmh@amazon.com>
Date: Fri, 12 Mar 2021 09:42:55 -0800
Subject: [PATCH] Add running pods to webhook notifications

---
 cmd/node-termination-handler.go |  7 ++++++-
 pkg/monitor/types.go            |  1 +
 pkg/node/node.go                | 23 ++++++++++++++++-------
 pkg/node/node_test.go           | 27 ++++++++++++++++++++++++++-
 4 files changed, 49 insertions(+), 9 deletions(-)

diff --git a/cmd/node-termination-handler.go b/cmd/node-termination-handler.go
index 26b62020..9ad5c7bc 100644
--- a/cmd/node-termination-handler.go
+++ b/cmd/node-termination-handler.go
@@ -296,7 +296,12 @@ func drainOrCordonIfNecessary(interruptionEventStore *interruptioneventstore.Sto
 			}
 		} else {
 			log.Log().Str("node_name", nodeName).Msg("Node successfully cordoned")
-			err = node.LogPods(nodeName)
+			podNameList, err := node.FetchPodNameList(nodeName)
+			if err != nil {
+				log.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.Log().Err(err).Msg("There was a problem while trying to log all pod names on the node")
 			}
diff --git a/pkg/monitor/types.go b/pkg/monitor/types.go
index e84b8c07..5e52ff95 100644
--- a/pkg/monitor/types.go
+++ b/pkg/monitor/types.go
@@ -31,6 +31,7 @@ type InterruptionEvent struct {
 	AutoScalingGroupName string
 	NodeName             string
 	NodeLabels           map[string]string
+	Pods                 []string
 	InstanceID           string
 	StartTime            time.Time
 	EndTime              time.Time
diff --git a/pkg/node/node.go b/pkg/node/node.go
index d1612be1..15023fbd 100644
--- a/pkg/node/node.go
+++ b/pkg/node/node.go
@@ -346,20 +346,29 @@ func (n Node) TaintRebalanceRecommendation(nodeName string, eventID string) erro
 }
 
 // LogPods logs all the pod names on a node
-func (n Node) LogPods(nodeName string) error {
-	podList, err := n.fetchAllPods(nodeName)
-	if err != nil {
-		return fmt.Errorf("Unable to fetch all pods from API: %w", err)
-	}
+func (n Node) LogPods(podList []string, nodeName string) error {
 	podNamesArr := zerolog.Arr()
-	for _, pod := range podList.Items {
-		podNamesArr = podNamesArr.Str(pod.Name)
+	for _, pod := range podList {
+		podNamesArr = podNamesArr.Str(pod)
 	}
 	log.Log().Array("pod_names", podNamesArr).Str("node_name", nodeName).Msg("Pods on node")
 
 	return nil
 }
 
+// FetchPodNameList fetches list of all the pods names running on given nodeName
+func (n Node) FetchPodNameList(nodeName string) ([]string, error) {
+	podList, err := n.fetchAllPods(nodeName)
+	if err != nil {
+		return nil, err
+	}
+	var podNamesList []string
+	for _, pod := range podList.Items {
+		podNamesList = append(podNamesList, pod.Name)
+	}
+	return podNamesList, nil
+}
+
 // TaintScheduledMaintenance adds the scheduled maintenance taint onto a node
 func (n Node) TaintScheduledMaintenance(nodeName string, eventID string) error {
 	if !n.nthConfig.TaintNode {
diff --git a/pkg/node/node_test.go b/pkg/node/node_test.go
index 2cfd2aa5..50b115fc 100644
--- a/pkg/node/node_test.go
+++ b/pkg/node/node_test.go
@@ -240,6 +240,31 @@ func TestMarkForUncordonAfterRebootAddActionLabelFailure(t *testing.T) {
 	h.Assert(t, err != nil, "Failed to return error on MarkForUncordonAfterReboot failing to add action Label")
 }
 
+func TestFetchPodsNameList(t *testing.T) {
+	resetFlagsForTest()
+
+	client := fake.NewSimpleClientset(
+		&v1.Pod{
+			ObjectMeta: metav1.ObjectMeta{
+				Name: "myPod",
+				Labels: map[string]string{
+					"spec.nodeName": nodeName,
+				},
+			},
+		},
+		&v1.Node{
+			ObjectMeta: metav1.ObjectMeta{
+				Name: nodeName,
+			},
+		},
+	)
+
+	tNode := getNode(t, getDrainHelper(client))
+	podList, err := tNode.FetchPodNameList(nodeName)
+	h.Ok(t, err)
+	h.Equals(t, []string{"myPod"}, podList)
+}
+
 func TestLogPods(t *testing.T) {
 	resetFlagsForTest()
 
@@ -260,7 +285,7 @@ func TestLogPods(t *testing.T) {
 	)
 
 	tNode := getNode(t, getDrainHelper(client))
-	err := tNode.LogPods(nodeName)
+	err := tNode.LogPods([]string{"myPod"}, nodeName)
 	h.Ok(t, err)
 }