K8s: Update ScaledJob scaling strategy to eager as default (#2466)

Author: VietND96

Diff for: charts/selenium-grid/CONFIGURATION.md

@@ -79,7 +79,7 @@ A Helm chart for creating a Selenium Grid Server in Kubernetes
 | serviceAccount.create | bool | `true` | Create a service account for all components. If using an external service account, set to false and provide its name in `nameOverride` below |
 | serviceAccount.nameOverride | string | `nil` | Override to use an external service account |
 | serviceAccount.annotations | object | `{}` | Annotations for the service account |
-| rbacRole | object | `{"annotations":{},"create":true,"nameOverride":null,"rules":[{"apiGroups":["keda.sh"],"resources":["scaledjobs"],"verbs":["get","list","patch","update","delete"]},{"apiGroups":["keda.sh"],"resources":["scaledobjects"],"verbs":["get","list","patch","update","delete"]},{"apiGroups":["keda.sh"],"resources":["triggerauthentication"],"verbs":["get","list","patch","update","delete"]},{"apiGroups":["autoscaling"],"resources":["horizontalpodautoscalers"],"verbs":["get","list","patch","update","delete"]}]}` | RBAC settings for patching finalizers KEDA scaled resources |
+| rbacRole | object | `{"annotations":{},"create":true,"nameOverride":null,"rules":[{"apiGroups":["keda.sh"],"resources":["scaledjobs"],"verbs":["get","list","patch","update","delete"]},{"apiGroups":["keda.sh"],"resources":["scaledobjects"],"verbs":["get","list","patch","update","delete"]},{"apiGroups":["keda.sh"],"resources":["triggerauthentications"],"verbs":["get","list","patch","update","delete"]},{"apiGroups":["autoscaling"],"resources":["horizontalpodautoscalers"],"verbs":["get","list","patch","update","delete"]}]}` | RBAC settings for patching finalizers KEDA scaled resources |
 | rbacRole.create | bool | `true` | Enable to create RBAC role to access few KEDA resources. If using an external role, set to false and provide its name in `nameOverride` below |
 | rbacRole.nameOverride | string | `nil` | Override resource name or provide an external role name |
 | rbacRoleBinding | object | `{"annotations":{},"create":true,"nameOverride":null,"roleRef":{"apiGroup":"rbac.authorization.k8s.io","kind":"Role"},"subjects":[{"kind":"ServiceAccount"}]}` | RBAC role binding settings for patching finalizers KEDA scaled resources |
@@ -326,8 +326,8 @@ A Helm chart for creating a Selenium Grid Server in Kubernetes
 | autoscaling.enabled | bool | `false` | Enable autoscaling. Implies installing KEDA |
 | autoscaling.enableWithExistingKEDA | bool | `false` | Enable autoscaling without automatically installing KEDA |
 | autoscaling.scalingType | string | `"job"` | Which type of KEDA scaling to use: job or deployment |
-| autoscaling.authenticationRef | object | `{"annotations":{"helm.sh/hook":"post-install,post-upgrade,post-rollback","helm.sh/hook-weight":"-2"},"name":""}` | Specify an external KEDA TriggerAuthentication resource is used for scaler triggers config. Apply for all browser nodes |
-| autoscaling.useCachedMetrics | bool | `true` | Enables caching of metric values during polling interval (as specified in .spec.pollingInterval). |
+| autoscaling.authenticationRef | object | `{"annotations":{"helm.sh/hook":"post-install,post-upgrade,post-rollback","helm.sh/hook-weight":"0"},"name":""}` | Specify an external KEDA TriggerAuthentication resource is used for scaler triggers config. Apply for all browser nodes |
+| autoscaling.useCachedMetrics | bool | `false` | Enables caching of metric values during polling interval (as specified in .spec.pollingInterval). |
 | autoscaling.metricType | string | `"Value"` | The type of metric that should be used (Override the default: AverageValue in KEDA) |
 | autoscaling.annotations | object | `{"helm.sh/hook":"post-install,post-upgrade,post-rollback","helm.sh/hook-weight":"1"}` | Annotations for KEDA resources: ScaledObject and ScaledJob |
 | autoscaling.patchObjectFinalizers.nameOverride | string | `nil` | Override the name of the patch job |
@@ -341,7 +341,7 @@ A Helm chart for creating a Selenium Grid Server in Kubernetes
 | autoscaling.scaledOptions.minReplicaCount | int | `0` | Minimum number of replicas |
 | autoscaling.scaledOptions.maxReplicaCount | int | `8` | Maximum number of replicas |
 | autoscaling.scaledOptions.pollingInterval | int | `10` | Polling interval in seconds |
-| autoscaling.scaledJobOptions.scalingStrategy.strategy | string | `"accurate"` | Scaling strategy for KEDA ScaledJob |
+| autoscaling.scaledJobOptions.scalingStrategy.strategy | string | `"eager"` | Scaling strategy for KEDA ScaledJob - https://keda.sh/docs/latest/reference/scaledjob-spec/#scalingstrategy |
 | autoscaling.scaledJobOptions.successfulJobsHistoryLimit | int | `0` | Number of Completed jobs should be kept |
 | autoscaling.scaledJobOptions.failedJobsHistoryLimit | int | `0` | Number of Failed jobs should be kept (for troubleshooting purposes) |
 | autoscaling.scaledJobOptions.jobTargetRef | object | `{"backoffLimit":0,"completions":1,"parallelism":1}` | Specify job target ref for KEDA ScaledJob |

Diff for: charts/selenium-grid/templates/_helpers.tpl

@@ -249,7 +249,9 @@ triggers:
     authenticationRef:
       name: {{ template "seleniumGrid.autoscaling.authenticationRef.fullname" $ }}
     useCachedMetrics: {{ $.Values.autoscaling.useCachedMetrics }}
+  {{- if eq $.Values.autoscaling.scalingType "deployment" }}
     metricType: {{ $.Values.autoscaling.metricType }}
+  {{- end }}
 {{- end }}
 {{- end -}}
 

Diff for: charts/selenium-grid/templates/patch-keda/patch-keda-objects-job.yaml

@@ -35,8 +35,8 @@ spec:
             - |
               echo "Cleaning up ScaledObjects, ScaledJobs and HPAs for {{ .Release.Name }} when upgrading or disabling autoscaling."
               kubectl get ScaledObjects,ScaledJobs,TriggerAuthentication -n {{ .Release.Namespace }} -l component.autoscaling={{ .Release.Name }} -o=json | jq '.metadata.finalizers = null' | kubectl apply -f - || true ;
-              kubectl delete ScaledObjects,ScaledJobs,TriggerAuthentication -n {{ .Release.Namespace }} -l component.autoscaling={{ .Release.Name }} --wait || true ;
-              kubectl delete hpa -n {{ .Release.Namespace }} -l component.autoscaling={{ .Release.Name }} --wait || true ;
+              kubectl delete ScaledObjects,ScaledJobs,TriggerAuthentication -n {{ .Release.Namespace }} -l component.autoscaling={{ .Release.Name }} --wait false || true ;
+              kubectl delete hpa -n {{ .Release.Namespace }} -l component.autoscaling={{ .Release.Name }} --wait false || true ;
         {{- with $.Values.autoscaling.patchObjectFinalizers.resources }}
           resources: {{ toYaml . | nindent 12 }}
         {{- end }}

Diff for: charts/selenium-grid/values.yaml

@@ -156,7 +156,7 @@ rbacRole:
     - apiGroups:
         - keda.sh
       resources:
-        - triggerauthentication
+        - triggerauthentications
       verbs: [get, list, patch, update, delete]
     - apiGroups:
         - autoscaling
@@ -835,15 +835,17 @@ autoscaling:
     name: ""
     annotations:
       "helm.sh/hook": post-install,post-upgrade,post-rollback
-      "helm.sh/hook-weight": "-2"
+      # TriggerAuthentication is used by ScaledObject/ScaledJob, hence weight should be less than those hooks
+      "helm.sh/hook-weight": "0"
   # Configuration for ScaledObject triggers https://keda.sh/docs/latest/reference/scaledobject-spec/#triggers
   # -- Enables caching of metric values during polling interval (as specified in .spec.pollingInterval).
-  useCachedMetrics: true
+  useCachedMetrics: false
   # -- The type of metric that should be used (Override the default: AverageValue in KEDA)
   metricType: Value
   # -- Annotations for KEDA resources: ScaledObject and ScaledJob
   annotations:
     "helm.sh/hook": post-install,post-upgrade,post-rollback
+    # Ensure the weight should be higher than TriggerAuthentication hook
     "helm.sh/hook-weight": "1"
   patchObjectFinalizers:
     # -- Override the name of the patch job
@@ -856,6 +858,7 @@ autoscaling:
     annotations:
       "helm.sh/hook": post-install,post-upgrade,post-rollback,pre-delete
       "helm.sh/hook-delete-policy": hook-succeeded,before-hook-creation
+      # This should be run before all other hooks (since delete action is called), so use a negative weight
       "helm.sh/hook-weight": "-1"
     # -- Define an external service account name contains permissions to patch KEDA scaled resources
     serviceAccount: ""
@@ -882,9 +885,9 @@ autoscaling:
   # Options for KEDA ScaledJobs (only used when scalingType is set to "job"). See https://keda.sh/docs/latest/concepts/scaling-jobs/#scaledjob-spec
   scaledJobOptions:
     scalingStrategy:
-      # Offer the strategy default with scaler calculation updated in https://github.com/SeleniumHQ/docker-selenium/tree/trunk/.keda/README.md
-      # -- Scaling strategy for KEDA ScaledJob
-      strategy: accurate
+      # Use `eager` strategy for utilizing all available slots up to the maxReplicaCount, ensuring that waiting request are processed as quickly as possible.
+      # -- Scaling strategy for KEDA ScaledJob - https://keda.sh/docs/latest/reference/scaledjob-spec/#scalingstrategy
+      strategy: eager
     # -- Number of Completed jobs should be kept
     successfulJobsHistoryLimit: 0
     # -- Number of Failed jobs should be kept (for troubleshooting purposes)