Add option for disabling leader election (#2423)

* Add option for disabling leader election

Signed-off-by: Yi Chen <github@chenyicn.net>

* Remove related RBAC rules when disabling leader election

Signed-off-by: Yi Chen <github@chenyicn.net>

---------

Signed-off-by: Yi Chen <github@chenyicn.net>

charts/spark-operator-chart/README.md

@@ -84,6 +84,7 @@ See [helm uninstall](https://helm.sh/docs/helm/helm_uninstall) for command docum
 | image.pullPolicy | string | `"IfNotPresent"` | Image pull policy. |
 | image.pullSecrets | list | `[]` | Image pull secrets for private image registry. |
 | controller.replicas | int | `1` | Number of replicas of controller. |
+| controller.leaderElection.enable | bool | `true` | Specifies whether to enable leader election for controller. |
 | controller.workers | int | `10` | Reconcile concurrency, higher values might increase memory usage. |
 | controller.logLevel | string | `"info"` | Configure the verbosity of logging, can be one of `debug`, `info`, `error`. |
 | controller.driverPodCreationGracePeriod | string | `"10s"` | Grace period after a successful spark-submit when driver pod not found errors will be retried. Useful if the driver pod can take some time to be created. |
@@ -127,6 +128,7 @@ See [helm uninstall](https://helm.sh/docs/helm/helm_uninstall) for command docum
 | controller.workqueueRateLimiter.maxDelay.duration | string | `"6h"` | Specifies the maximum delay duration for the workqueue rate limiter. |
 | webhook.enable | bool | `true` | Specifies whether to enable webhook. |
 | webhook.replicas | int | `1` | Number of replicas of webhook server. |
+| webhook.leaderElection.enable | bool | `true` | Specifies whether to enable leader election for webhook. |
 | webhook.logLevel | string | `"info"` | Configure the verbosity of logging, can be one of `debug`, `info`, `error`. |
 | webhook.port | int | `9443` | Specifies webhook port. |
 | webhook.portName | string | `"webhook"` | Specifies webhook service port name. |

charts/spark-operator-chart/templates/controller/deployment.yaml

@@ -89,9 +89,13 @@ spec:
         - --metrics-prefix={{ .Values.prometheus.metrics.prefix }}
         - --metrics-labels=app_type
         {{- end }}
+        {{ if .Values.controller.leaderElection.enable }}
         - --leader-election=true
         - --leader-election-lock-name={{ include "spark-operator.controller.leaderElectionName" . }}
         - --leader-election-lock-namespace={{ .Release.Namespace }}
+        {{- else -}}
+        - --leader-election=false
+        {{- end }}
         {{- if .Values.controller.pprof.enable }}
         - --pprof-bind-address=:{{ .Values.controller.pprof.port }}
         {{- end }}

charts/spark-operator-chart/templates/controller/rbac.yaml

@@ -85,6 +85,7 @@ metadata:
     {{- toYaml . | nindent 4 }}
   {{- end }}
 rules:
+{{- if .Values.controller.leaderElection.enable }}
 - apiGroups:
   - coordination.k8s.io
   resources:
@@ -100,6 +101,7 @@ rules:
   verbs:
   - get
   - update
+{{- end }}
 {{- if has .Release.Namespace .Values.spark.jobNamespaces }}
 {{ include "spark-operator.controller.policyRules" . }}
 {{- end }}

charts/spark-operator-chart/templates/webhook/deployment.yaml

@@ -74,9 +74,13 @@ spec:
         - --metrics-prefix={{ .Values.prometheus.metrics.prefix }}
         - --metrics-labels=app_type
         {{- end }}
+        {{ if .Values.webhook.leaderElection.enable }}
         - --leader-election=true
         - --leader-election-lock-name={{ include "spark-operator.webhook.leaderElectionName" . }}
         - --leader-election-lock-namespace={{ .Release.Namespace }}
+        {{- else -}}
+        - --leader-election=false
+        {{- end }}
         ports:
         - name: {{ .Values.webhook.portName | quote }}
           containerPort: {{ .Values.webhook.port }}

charts/spark-operator-chart/templates/webhook/rbac.yaml

@@ -107,6 +107,7 @@ rules:
   verbs:
   - get
   - update
+{{- if .Values.webhook.leaderElection.enable }}
 - apiGroups:
   - coordination.k8s.io
   resources:
@@ -122,6 +123,7 @@ rules:
   verbs:
   - get
   - update
+{{- end }}
 {{- if has .Release.Namespace .Values.spark.jobNamespaces }}
 {{ include "spark-operator.webhook.policyRules" . }}
 {{- end }}

charts/spark-operator-chart/tests/controller/deployment_test.yaml

@@ -242,6 +242,16 @@ tests:
       - contains:
           path: spec.template.spec.containers[?(@.name=="spark-operator-controller")].args
           content: --leader-election-lock-namespace=spark-operator
+  
+  - it: Should disable leader election if `controller.leaderElection.enable` is set to `false`
+    set:
+      controller:
+        leaderElection:
+          enable: false
+    asserts:
+      - contains:
+          path: spec.template.spec.containers[?(@.name=="spark-operator-controller")].args
+          content: --leader-election=false
 
   - it: Should add metric ports if `prometheus.metrics.enable` is true
     set:

charts/spark-operator-chart/tests/webhook/deployment_test.yaml

@@ -173,6 +173,16 @@ tests:
           path: spec.template.spec.containers[?(@.name=="spark-operator-webhook")].args
           content: --leader-election-lock-namespace=spark-operator
 
+  - it: Should disable leader election if `webhook.leaderElection.enable` is set to `false`
+    set:
+      webhook:
+        leaderElection:
+          enable: false
+    asserts:
+      - contains:
+          path: spec.template.spec.containers[?(@.name=="spark-operator-webhook")].args
+          content: --leader-election=false
+
   - it: Should add webhook port
     set:
       webhook:

charts/spark-operator-chart/values.yaml

@@ -45,6 +45,10 @@ controller:
   # -- Number of replicas of controller.
   replicas: 1
 
+  leaderElection:
+    # -- Specifies whether to enable leader election for controller.
+    enable: true
+
   # -- Reconcile concurrency, higher values might increase memory usage.
   workers: 10
 
@@ -220,6 +224,10 @@ webhook:
   # -- Number of replicas of webhook server.
   replicas: 1
 
+  leaderElection:
+    # -- Specifies whether to enable leader election for webhook.
+    enable: true
+
   # -- Configure the verbosity of logging, can be one of `debug`, `info`, `error`.
   logLevel: info