Liveness Readyness probes #65

enonic · Mar 27, 2023 · feb1eaa · feb1eaa
1 parent 370b6a3
commit feb1eaa
Show file tree

Hide file tree

Showing 5 changed files with 108 additions and 43 deletions.
diff --git a/java-operator/src/main/helm/v1/xp7deployment/templates/nodeGroup_statefulset.yaml b/java-operator/src/main/helm/v1/xp7deployment/templates/nodeGroup_statefulset.yaml
@@ -170,25 +170,34 @@ spec:
               value: {{ $env.value }}
           {{- end }}
 
-          readinessProbe:
-            failureThreshold: 2
+
+          {{- if semverCompare "gt" $.Values.deployment.spec.xpVersion "1.0.0" }}
+          startupProbe:
             httpGet:
-              {{- if not $.Values.deployment.clustered }}
-              # Single nodes
-              path: {{ $.Values.deployment.preInstalledCloudUtils | ternary "/cloud-utils/health?state=" "/_cluster/health?timeout=1s&wait_for_status=" }}green
-              {{- else if and (and $nodeGroup.data $.Values.deployment.clustered) $.Values.deployment.hasDedicatedFrontendNodes }}
-              # Cluster with dedicated frontend nodes
-              path: {{ $.Values.deployment.preInstalledCloudUtils | ternary "/cloud-utils/health?state=" "/_cluster/health?timeout=1s&wait_for_status=" }}green
-              {{ else }}
-              # Cluster with no dedicated cluster nodes
-              path: {{ $.Values.deployment.preInstalledCloudUtils | ternary "/cloud-utils/health?state=" "/_cluster/health?timeout=1s&wait_for_status=" }}yellow
-              {{- end }}
-              port: {{ $.Values.deployment.preInstalledCloudUtils | ternary 4848 9200 }}
+              path: "/healthz"
+              port: 2609
               scheme: HTTP
+            failureThreshold: 20
             initialDelaySeconds: 30
-            periodSeconds: 8
-            successThreshold: 3
+            periodSeconds: 5
             timeoutSeconds: 2
+          livenessProbe:
+            httpGet:
+              path: "/healthz"
+              port: 2609
+              scheme: HTTP
+            failureThreshold: 10
+            timeoutSeconds: 2
+          readinessProbe:
+            httpGet:
+              path: "/ready"
+              port: 2609
+              scheme: HTTP
+            failureThreshold: 2
+            timeoutSeconds: 2
+          {{ else }}
+          {{- end }}
+
 
           {{- if $.Values.deployment.clustered }}
           lifecycle:

diff --git a/...c/test/resources/com/enonic/kubernetes/operator/helm/v1/xp7deployment/cluster_result.yaml b/...c/test/resources/com/enonic/kubernetes/operator/helm/v1/xp7deployment/cluster_result.yaml
@@ -808,6 +808,13 @@ spec:
               command:
               - "/bin/bash"
               - "/enonic-xp/home/extra-config/preStop.sh"
+        livenessProbe:
+          failureThreshold: 10
+          httpGet:
+            path: "/healthz"
+            port: 2609
+            scheme: "HTTP"
+          timeoutSeconds: 2
         name: "exp"
         ports:
         - containerPort: 2609
@@ -825,12 +832,9 @@ spec:
         readinessProbe:
           failureThreshold: 2
           httpGet:
-            path: "/_cluster/health?timeout=1s&wait_for_status=yellow"
-            port: 9200
+            path: "/ready"
+            port: 2609
             scheme: "HTTP"
-          initialDelaySeconds: 30
-          periodSeconds: 8
-          successThreshold: 3
           timeoutSeconds: 2
         resources:
           limits:
@@ -845,6 +849,15 @@ spec:
             - "ALL"
           runAsNonRoot: true
           runAsUser: 1337
+        startupProbe:
+          failureThreshold: 20
+          httpGet:
+            path: "/healthz"
+            port: 2609
+            scheme: "HTTP"
+          initialDelaySeconds: 30
+          periodSeconds: 5
+          timeoutSeconds: 2
         terminationMessagePath: "/dev/termination-log"
         terminationMessagePolicy: "File"
         volumeMounts:
@@ -1068,6 +1081,13 @@ spec:
               command:
               - "/bin/bash"
               - "/enonic-xp/home/extra-config/preStop.sh"
+        livenessProbe:
+          failureThreshold: 10
+          httpGet:
+            path: "/healthz"
+            port: 2609
+            scheme: "HTTP"
+          timeoutSeconds: 2
         name: "exp"
         ports:
         - containerPort: 2609
@@ -1085,12 +1105,9 @@ spec:
         readinessProbe:
           failureThreshold: 2
           httpGet:
-            path: "/_cluster/health?timeout=1s&wait_for_status=green"
-            port: 9200
+            path: "/ready"
+            port: 2609
             scheme: "HTTP"
-          initialDelaySeconds: 30
-          periodSeconds: 8
-          successThreshold: 3
           timeoutSeconds: 2
         resources:
           limits:
@@ -1105,6 +1122,15 @@ spec:
             - "ALL"
           runAsNonRoot: true
           runAsUser: 1337
+        startupProbe:
+          failureThreshold: 20
+          httpGet:
+            path: "/healthz"
+            port: 2609
+            scheme: "HTTP"
+          initialDelaySeconds: 30
+          periodSeconds: 5
+          timeoutSeconds: 2
         terminationMessagePath: "/dev/termination-log"
         terminationMessagePolicy: "File"
         volumeMounts:
@@ -1351,6 +1377,13 @@ spec:
               command:
               - "/bin/bash"
               - "/enonic-xp/home/extra-config/preStop.sh"
+        livenessProbe:
+          failureThreshold: 10
+          httpGet:
+            path: "/healthz"
+            port: 2609
+            scheme: "HTTP"
+          timeoutSeconds: 2
         name: "exp"
         ports:
         - containerPort: 2609
@@ -1368,12 +1401,9 @@ spec:
         readinessProbe:
           failureThreshold: 2
           httpGet:
-            path: "/_cluster/health?timeout=1s&wait_for_status=yellow"
-            port: 9200
+            path: "/ready"
+            port: 2609
             scheme: "HTTP"
-          initialDelaySeconds: 30
-          periodSeconds: 8
-          successThreshold: 3
           timeoutSeconds: 2
         resources:
           limits:
@@ -1388,6 +1418,15 @@ spec:
             - "ALL"
           runAsNonRoot: true
           runAsUser: 1337
+        startupProbe:
+          failureThreshold: 20
+          httpGet:
+            path: "/healthz"
+            port: 2609
+            scheme: "HTTP"
+          initialDelaySeconds: 30
+          periodSeconds: 5
+          timeoutSeconds: 2
         terminationMessagePath: "/dev/termination-log"
         terminationMessagePolicy: "File"
         volumeMounts:

diff --git a/...rc/test/resources/com/enonic/kubernetes/operator/helm/v1/xp7deployment/simple_result.yaml b/...rc/test/resources/com/enonic/kubernetes/operator/helm/v1/xp7deployment/simple_result.yaml
@@ -625,6 +625,13 @@ spec:
             \ -Dhazelcast.shutdownhook.policy=GRACEFUL -Dhazelcast.graceful.shutdown.max.wait=600"
         image: "enonic/xp:7.7.0-ubuntu"
         imagePullPolicy: "IfNotPresent"
+        livenessProbe:
+          failureThreshold: 10
+          httpGet:
+            path: "/healthz"
+            port: 2609
+            scheme: "HTTP"
+          timeoutSeconds: 2
         name: "exp"
         ports:
         - containerPort: 2609
@@ -642,12 +649,9 @@ spec:
         readinessProbe:
           failureThreshold: 2
           httpGet:
-            path: "/_cluster/health?timeout=1s&wait_for_status=green"
-            port: 9200
+            path: "/ready"
+            port: 2609
             scheme: "HTTP"
-          initialDelaySeconds: 30
-          periodSeconds: 8
-          successThreshold: 3
           timeoutSeconds: 2
         resources:
           limits:
@@ -662,6 +666,15 @@ spec:
             - "ALL"
           runAsNonRoot: true
           runAsUser: 1337
+        startupProbe:
+          failureThreshold: 20
+          httpGet:
+            path: "/healthz"
+            port: 2609
+            scheme: "HTTP"
+          initialDelaySeconds: 30
+          periodSeconds: 5
+          timeoutSeconds: 2
         terminationMessagePath: "/dev/termination-log"
         terminationMessagePolicy: "File"
         volumeMounts:

diff --git a/kubernetes/example-simple.yaml b/kubernetes/example-simple.yaml
@@ -15,17 +15,26 @@ metadata:
   namespace: my-namespace
 spec:
   enabled: true
-  xpVersion: 7.10.3
+  xpVersion: 7.13.0-SNAPSHOT
 
   # Preinstall snapshotter on all nodes
   nodesPreinstalledApps:
     - name: snapshotter
       url: https://repo.enonic.com/public/com/enonic/app/snapshotter/3.0.2/snapshotter-3.0.2.jar
 
+  # List of disks shared by all nodes (see description of disks below)
+  nodesSharedDisks:
+    - name: blobstore
+      size: 1Gi
+    - name: snapshots
+      size: 1Gi
+    - name: export
+      size: 1Gi
+
   # Create one node
   nodeGroups:
     - name: main
-      replicas: 1
+      replicas: 3
 
       data: true
       master: true
@@ -36,12 +45,6 @@ spec:
 
         # Volumes private to the node
         disks:
-          - name: blobstore
-            size: 1Gi
-          - name: snapshots
-            size: 1Gi
-          - name: export # Dumps and other data
-            size: 1Gi
           - name: deploy # Apps installed in the deploy folder
             size: 1Gi
           - name: index # Node ES index

diff --git a/kubernetes/kind/Makefile b/kubernetes/kind/Makefile
@@ -3,6 +3,7 @@ KIND_MTU:=$(shell docker network inspect bridge -f '{{ index .Options "com.docke
 KIND_CTX:=kind-kind
 NFS_SERVER_NAME:=kind-nfs-server
 NFS_SERVER_VOLUME:=kind-nfs-server
+MY_IP:=192.168.0.30
 
 .PHONY: kind-setup kind-bootstrap kind-ingress kind-up kind-down help
 .DEFAULT_GOAL:=help