Update leader label only at pods with label 'app=kubemacpool' (#212)

Otherwise it iterate over all the pods at the same namespace, this is not a critical bug is kind of cosmetics but it reduce the time to update leader label. Signed-off-by: Quique Llorente <[email protected]>
k8snetworkplumbingwg · Jun 28, 2020 · 68a684e · 68a684e
1 parent 77858f3
commit 68a684e
Show file tree

Hide file tree

Showing 2 changed files with 14 additions and 5 deletions.
diff --git a/pkg/manager/leaderelection.go b/pkg/manager/leaderelection.go
@@ -71,7 +71,13 @@ func (k *KubeMacPoolManager) setLeadershipConditions(status corev1.ConditionStat
 func updateLeaderLabel(kubeClient client.Client, leaderPodName, managerNamespace string) error {
 	logger := logf.Log.WithName("UpdateLeaderLabel")
 	podList := corev1.PodList{}
-	err := kubeClient.List(context.TODO(), &podList, &client.ListOptions{Namespace: managerNamespace})
+
+	byNamespaceAndApp := &client.ListOptions{Namespace: managerNamespace}
+	client.MatchingLabels{
+		"app": "kubemacpool",
+	}.ApplyToList(byNamespaceAndApp)
+
+	err := kubeClient.List(context.TODO(), &podList, byNamespaceAndApp)
 	if err != nil {
 		return errors.Wrap(err, "failed to list kubemacpool manager pods")
 	}

diff --git a/pkg/manager/manager_test.go b/pkg/manager/manager_test.go
@@ -52,7 +52,9 @@ var _ = Describe("leader election", func() {
 			table.DescribeTable("Should update the leader label in all pods", func(leaderPodFormerLabels, looserPodFormerLabels string) {
 				By("Adding the initial label state of the pods prior to winning the election")
 				initiatePodLabels(&leaderPod, leaderPodFormerLabels, leaderLabelValue)
+				initiatePodLabels(&leaderPod, "app", "kubemacpool")
 				initiatePodLabels(&looserPod, looserPodFormerLabels, leaderLabelValue)
+				initiatePodLabels(&looserPod, "app", "kubemacpool")
 
 				By("Initiating the Environment")
 				kubeClient := createEnvironment(&leaderPod, &looserPod)
@@ -65,14 +67,13 @@ var _ = Describe("leader election", func() {
 				checkLeaderPod := v1.Pod{}
 				err = kubeClient.Get(context.TODO(), types.NamespacedName{Namespace: names.MANAGER_NAMESPACE, Name: leaderPodName}, &checkLeaderPod)
 				Expect(err).ToNot(HaveOccurred(), "should successfully get the kubemacpool leader pod")
-				Expect(checkLeaderPod.Labels).To(HaveLen(1), "leader pod should have only 1 label")
-				Expect(checkLeaderPod.Labels[names.LEADER_LABEL]).To(Equal(leaderLabelValue), "leader pod should have the leader label value")
+				Expect(checkLeaderPod.Labels).To(HaveKeyWithValue(names.LEADER_LABEL, leaderLabelValue), "leader pod should have the leader label value")
 
 				By("checking the non-leader pod has no leader label")
 				checkLooserPod := v1.Pod{}
 				err = kubeClient.Get(context.TODO(), types.NamespacedName{Namespace: names.MANAGER_NAMESPACE, Name: loosingPodName}, &checkLooserPod)
 				Expect(err).ToNot(HaveOccurred(), "should successfully get the kubemacpool non-leader pod")
-				Expect(checkLooserPod.Labels).To(HaveLen(0), "non-leader pod should not have any labels")
+				Expect(checkLooserPod.Labels).ToNot(HaveKeyWithValue(names.LEADER_LABEL, leaderLabelValue), "non leader pod should not have the leader label value")
 			},
 				table.Entry("all pods don't have a former leader label", "", ""),
 				table.Entry("leader pod already has leader label from former election", names.LEADER_LABEL, ""),
@@ -85,7 +86,9 @@ var _ = Describe("leader election", func() {
 
 func initiatePodLabels(pod *v1.Pod, label string, labelValue string) {
 	if len(label) != 0 {
-		pod.Labels = make(map[string]string)
+		if len(pod.Labels) == 0 {
+			pod.Labels = make(map[string]string)
+		}
 		pod.Labels[label] = labelValue
 	}
 }