Make cluster-autoscaler resource request/limit configurable (#1556)

* make cluster autoscaler configurable

* Making autoscaler configurable

* Removing code that isn't required

Author: Luke Humphreys

core/controlplane/config/templates/cloud-config-controller

@@ -4143,11 +4143,11 @@ write_files:
                   name: cluster-autoscaler
                   resources:
                     limits:
-                      cpu: 100m
-                      memory: 300Mi
+                      cpu: {{ if .Addons.ClusterAutoscaler.ComputeResources.Limits.Cpu }}{{ .Addons.ClusterAutoscaler.ComputeResources.Limits.Cpu }}{{ else }}100m{{ end }}
+                      memory: {{ if .Addons.ClusterAutoscaler.ComputeResources.Limits.Memory }}{{ .Addons.ClusterAutoscaler.ComputeResources.Limits.Memory }}{{ else }}300Mi{{ end }}
                     requests:
-                      cpu: 100m
-                      memory: 300Mi
+                      cpu: {{ if .Addons.ClusterAutoscaler.ComputeResources.Requests.Cpu }}{{ .Addons.ClusterAutoscaler.ComputeResources.Requests.Cpu }}{{ else }}100m{{ end }}
+                      memory: {{ if .Addons.ClusterAutoscaler.ComputeResources.Requests.Memory }}{{ .Addons.ClusterAutoscaler.ComputeResources.Requests.Memory }}{{ else }}300Mi{{ end }}
                   command:
                     - ./cluster-autoscaler
                     - --v=4

core/root/config/templates/cluster.yaml

@@ -1352,6 +1352,14 @@ addons:
   # If you want to run CA on worker nodes, turn on `worker.nodePools[].clusterAutoscalerSupport.enabled` for the node pool.
   clusterAutoscaler:
     enabled: false
+    resources:
+      # Increase these values substantially if running a cluster of 50+ nodes
+      limits:
+        cpu: 100m
+        memory: 300Mi
+      requests:
+        cpu: 100m
+        memory: 300Mi
     # Options below can be used to inject custom settings for the autoscaler.
     # Sensible defaults are already configured in the controller-cloud-config but if you wish to override them simply
     # add them here and they'll take precedence.

model/addons.go

@@ -10,9 +10,10 @@ type Addons struct {
 }
 
 type ClusterAutoscalerSupport struct {
-	Enabled     bool              `yaml:"enabled"`
-	Options     map[string]string `yaml:"options"`
-	UnknownKeys `yaml:",inline"`
+	Enabled          bool              `yaml:"enabled"`
+	ComputeResources ComputeResources  `yaml:"resources,omitempty"`
+	Options          map[string]string `yaml:"options"`
+	UnknownKeys      `yaml:",inline"`
 }
 
 type Rescheduler struct {
@@ -33,3 +34,13 @@ type Prometheus struct {
 type APIServerAggregator struct {
 	Enabled bool `yaml:"enabled"`
 }
+
+type ComputeResources struct {
+	Limits   ResourceQuota `yaml:"limits,omitempty"`
+	Requests ResourceQuota `yaml:"requests,omitempty"`
+}
+
+type ResourceQuota struct {
+	Cpu    string `yaml:"cpu"`
+	Memory string `yaml:"memory"`
+}

test/integration/maincluster_test.go

@@ -1642,7 +1642,7 @@ experimental:
     image:
       repo: quay.io/uswitch/kiam
       tag: v2.6
-    sessionDuration: 30m	
+    sessionDuration: 30m
     serverAddresses:
       serverAddress: localhost
       agentAddress: kiam-server