Make DaemonSet metrics scrapable (#215)

We do this by adding the prometheus service label + port to the container. Signed-off-by: Alex Kalenyuk <akalenyu@redhat.com> Co-authored-by: Alex Kalenyuk <akalenyu@redhat.com>
kubevirt · Jan 28, 2022 · 6a0f1ab · 6a0f1ab
1 parent 794666f
commit 6a0f1ab
Show file tree

Hide file tree

Showing 3 changed files with 45 additions and 0 deletions.
diff --git a/pkg/controller/hostpathprovisioner/daemonset.go b/pkg/controller/hostpathprovisioner/daemonset.go
@@ -473,6 +473,7 @@ func (r *ReconcileHostPathProvisioner) createCSIDaemonSetObject(cr *hostpathprov
 	pathMounts := buildVolumeMountsFromStoragePoolInfo(storagePoolPaths)
 	biDirectional := corev1.MountPropagationBidirectional
 	labels := getRecommendedLabels()
+	labels[PrometheusLabelKey] = PrometheusLabelValue
 	ds := &appsv1.DaemonSet{
 		TypeMeta: metav1.TypeMeta{
 			APIVersion: "apps/v1",
@@ -557,6 +558,11 @@ func (r *ReconcileHostPathProvisioner) createCSIDaemonSetObject(cr *hostpathprov
 									Name:          "healthz",
 									Protocol:      corev1.ProtocolTCP,
 								},
+								{
+									ContainerPort: 8080,
+									Name:          "metrics",
+									Protocol:      corev1.ProtocolTCP,
+								},
 							},
 							LivenessProbe: &corev1.Probe{
 								FailureThreshold:    5,

diff --git a/pkg/controller/hostpathprovisioner/daemonset_test.go b/pkg/controller/hostpathprovisioner/daemonset_test.go
@@ -218,6 +218,31 @@ var _ = Describe("Controller reconcile loop", func() {
 			table.Entry("longNamecr", createStoragePoolWithTemplateLongNameCr(), hashedLongStoragePoolNameDataDir),
 		)
 
+		table.DescribeTable("DaemonSet should have prometheus labels, port", func(cr *hppv1.HostPathProvisioner, volumeMountName string) {
+			cr, r, cl = createDeployedCr(cr)
+			ds := &appsv1.DaemonSet{
+				ObjectMeta: v1.ObjectMeta{
+					Name:      fmt.Sprintf("%s-csi", MultiPurposeHostPathProvisionerName),
+					Namespace: testNamespace,
+				},
+			}
+			dsNN := client.ObjectKeyFromObject(ds)
+			err := cl.Get(context.TODO(), dsNN, ds)
+			Expect(err).NotTo(HaveOccurred())
+			port := corev1.ContainerPort{
+				ContainerPort: 8080,
+				Name:          "metrics",
+				Protocol:      corev1.ProtocolTCP,
+			}
+			Expect(ds.Spec.Template.Spec.Containers[0].Ports).To(ContainElement(port))
+			Expect(ds.Labels[PrometheusLabelKey]).To(Equal(PrometheusLabelValue))
+			Expect(ds.Spec.Template.Labels[PrometheusLabelKey]).To(Equal(PrometheusLabelValue))
+		},
+			table.Entry("legacyCr", createLegacyCr(), legacyDataDir),
+			table.Entry("legacyStoragePoolCr", createLegacyStoragePoolCr(), legacyStoragePoolDataDir),
+			table.Entry("storagePoolCr", createStoragePoolWithTemplateCr(), localDataDir),
+		)
+
 		table.DescribeTable("Should create daemonset with node placement", func(dsName string) {
 			req := reconcile.Request{
 				NamespacedName: types.NamespacedName{

diff --git a/pkg/controller/hostpathprovisioner/prometheus.go b/pkg/controller/hostpathprovisioner/prometheus.go
@@ -235,6 +235,20 @@ func getAlertRules() []promv1.Rule {
 				componentAlertLabelKey: componentAlertLabelValue,
 			},
 		),
+		generateAlertRule(
+			"HPPSharingPoolPathWithOS",
+			"kubevirt_hpp_pool_path_shared_with_os == 1",
+			"1m",
+			map[string]string{
+				"summary":     "HPP pool path sharing a filesystem with OS, fix to prevent HPP PVs from causing disk pressure and affecting node operation",
+				"runbook_url": runbookURLBasePath + "HPPSharingPoolPathWithOS",
+			},
+			map[string]string{
+				severityAlertLabelKey:  "warning",
+				partOfAlertLabelKey:    partOfAlertLabelValue,
+				componentAlertLabelKey: componentAlertLabelValue,
+			},
+		),
 	}
 }