feat(chart): add health probes to api-rest

Signed-off-by: Niladri Halder <[email protected]>
openebs · Dec 10, 2024 · 1e28ff3 · 1e28ff3
1 parent ece2e0b
commit 1e28ff3
Show file tree

Hide file tree

Showing 3 changed files with 59 additions and 0 deletions.
diff --git a/chart/README.md b/chart/README.md
@@ -93,6 +93,17 @@ This removes all the Kubernetes components associated with the chart and deletes
 | agents.&ZeroWidthSpace;ha.&ZeroWidthSpace;node.&ZeroWidthSpace;resources.&ZeroWidthSpace;requests.&ZeroWidthSpace;cpu | Cpu requests for ha node agent | `"100m"` |
 | agents.&ZeroWidthSpace;ha.&ZeroWidthSpace;node.&ZeroWidthSpace;resources.&ZeroWidthSpace;requests.&ZeroWidthSpace;memory | Memory requests for ha node agent | `"64Mi"` |
 | agents.&ZeroWidthSpace;ha.&ZeroWidthSpace;node.&ZeroWidthSpace;tolerations | Set tolerations, overrides global | `[]` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;liveness.&ZeroWidthSpace;enabled | Toggle liveness probe. | `true` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;liveness.&ZeroWidthSpace;failureThreshold | No. of failures the liveness probe will tolerate. | `1` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;liveness.&ZeroWidthSpace;initialDelaySeconds | No. of seconds of delay before checking the liveness status. | `0` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;liveness.&ZeroWidthSpace;periodSeconds | No. of seconds between liveness probe checks. | `30` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;liveness.&ZeroWidthSpace;timeoutSeconds | No. of seconds of timeout tolerance. | `5` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;readiness.&ZeroWidthSpace;agentCoreProbeFreq | Frequency for the agent-core liveness probe. | `"1m"` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;readiness.&ZeroWidthSpace;enabled | Toggle readiness probe. | `true` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;readiness.&ZeroWidthSpace;failureThreshold | No. of failures the readiness probe will tolerate. | `2` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;readiness.&ZeroWidthSpace;initialDelaySeconds | No. of seconds of delay before checking the readiness status. | `0` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;readiness.&ZeroWidthSpace;periodSeconds | No. of seconds between readiness probe checks. | `20` |
+| apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;healthProbes.&ZeroWidthSpace;readiness.&ZeroWidthSpace;timeoutSeconds | No. of seconds of timeout tolerance. | `5` |
 | apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;logLevel | Log level for the rest service | `"info"` |
 | apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;priorityClassName | Set PriorityClass, overrides global. If both local and global are not set, the final deployment manifest has a mayastor custom critical priority class assigned to the pod by default. Refer the `templates/_helpers.tpl` and `templates/mayastor/apis/rest/api-rest-deployment.yaml` for more details. | `""` |
 | apis.&ZeroWidthSpace;rest.&ZeroWidthSpace;replicaCount | Number of replicas of rest | `1` |

diff --git a/chart/templates/mayastor/apis/api-rest-deployment.yaml b/chart/templates/mayastor/apis/api-rest-deployment.yaml
@@ -53,6 +53,9 @@ spec:
             - "--core-grpc=https://{{ .Release.Name }}-agent-core:50051"
             - "--ansi-colors={{ .Values.base.logging.color }}"
             - "--fmt-style={{ include "logFormat" . }}"
+            {{- if .Values.apis.rest.healthProbes.readiness.enabled }}
+            - "--core-health-freq={{ .Values.apis.rest.healthProbes.readiness.agentCoreProbeFreq }}"
+            {{- end }}
           ports:
             - containerPort: 8080
             - containerPort: 8081
@@ -63,3 +66,23 @@ spec:
             - name: RUST_LOG_SILENCE
               value: {{ default .Values.base.logging.silenceLevel .Values.apis.rest.logSilenceLevel }}
             {{- end }}
+          {{- if .Values.apis.rest.healthProbes.readiness.enabled }}
+          readinessProbe:
+            httpGet:
+              path: /ready
+              port: 8081
+            failureThreshold: {{ .Values.apis.rest.healthProbes.readiness.failureThreshold }}
+            initialDelaySeconds: {{ .Values.apis.rest.healthProbes.readiness.initialDelaySeconds }}
+            periodSeconds: {{ .Values.apis.rest.healthProbes.readiness.periodSeconds }}
+            timeoutSeconds: {{ .Values.apis.rest.healthProbes.readiness.timeoutSeconds }}
+          {{- end }}
+          {{- if .Values.apis.rest.healthProbes.liveness.enabled }}
+          livenessProbe:
+            httpGet:
+              path: /live
+              port: 8081
+            failureThreshold: {{ .Values.apis.rest.healthProbes.liveness.failureThreshold }}
+            initialDelaySeconds: {{ .Values.apis.rest.healthProbes.liveness.initialDelaySeconds }}
+            periodSeconds: {{ .Values.apis.rest.healthProbes.liveness.periodSeconds }}
+            timeoutSeconds: {{ .Values.apis.rest.healthProbes.liveness.timeoutSeconds }}
+          {{- end }}
diff --git a/chart/values.yaml b/chart/values.yaml
@@ -254,6 +254,31 @@ apis:
   rest:
     # -- Log level for the rest service
     logLevel: info
+    healthProbes:
+      readiness:
+        # -- Toggle readiness probe.
+        enabled: true
+        # -- Frequency for the agent-core liveness probe.
+        agentCoreProbeFreq: "20s"
+        # -- No. of failures the readiness probe will tolerate.
+        failureThreshold: 2
+        # -- No. of seconds of delay before checking the readiness status.
+        initialDelaySeconds: 0
+        # -- No. of seconds between readiness probe checks.
+        periodSeconds: 20
+        # -- No. of seconds of timeout tolerance.
+        timeoutSeconds: 5
+      liveness:
+        # -- Toggle liveness probe.
+        enabled: true
+        # -- No. of failures the liveness probe will tolerate.
+        failureThreshold: 1
+        # -- No. of seconds of delay before checking the liveness status.
+        initialDelaySeconds: 0
+        # -- No. of seconds between liveness probe checks.
+        periodSeconds: 30
+        # -- No. of seconds of timeout tolerance.
+        timeoutSeconds: 5
     # -- Number of replicas of rest
     replicaCount: 1
     resources: