Proactive Policies

​

Scope Mapping

• ProactivePolicy is namespaced and references an AutomationStrategy in the same namespace.
• ClusterProactivePolicy is cluster-scoped and references a ClusterAutomationStrategy; that resource is documented separately in Cluster Proactive Policies.

​

Field Reference

​

ProactivePolicy.spec

​

Example: Namespaced Proactive Policy

apiVersion: rightsizing.kubex.ai/v1alpha1
kind: ProactivePolicy
metadata:
  name: team-a-proactive
  namespace: team-a
spec:
  scope:
    containers:
      - api
      - worker
    labelSelector:
      matchLabels:
        app.kubernetes.io/part-of: storefront
      matchExpressions:
        - key: tier
          operator: In
          values:
            - api
            - worker
    workloadTypes:
      - Deployment
      - StatefulSet
  automationStrategyRef:
    name: team-a-balanced
  weight: 100
  safetyChecks:
    maxAnalysisAgeDays: 3

​

Ephemeral Storage

1. Turn it on for the controller in your Helm values. Do this only if your Kubex instance is on a rapid release that supports ephemeral-storage automation. On any other release, leave it off.
   controllerManager:
     extraEnv:
       - name: EPHEMERAL_STORAGE_AUTOMATION_ENABLED
         value: "true"
   
2. Allow storage changes in each AutomationStrategy or ClusterAutomationStrategy that should make them. Every storage change is off in a strategy until you set it to true:
   spec:
     enablement:
       ephemeralStorage:
         requests:
           upsize: true
           downsize: true
           setFromUnspecified: true
         limits:
           upsize: true
           downsize: true
           setFromUnspecified: true
   

• Kubernetes cannot change ephemeral storage on a running pod. By default, Kubex applies storage changes the next time the pod is recreated (for example a rollout, scale-up, or node drain). A container crash restart does not apply them. They never cause an eviction.
• CPU and memory rightsizing for the same pod is unaffected and happens as usual.
• Set restartStrategy: Evict to let Kubex restart the pod to apply storage changes. This requires podEviction.enabled: true in the strategy. A pod with CPU or memory changes together with storage changes is then restarted once to apply all of them.

spec:
  podEviction:
    enabled: true
  enablement:
    ephemeralStorage:
      restartStrategy: Evict

• Storage changes are checked against node ephemeral-storage capacity, namespace ResourceQuota, and LimitRange.
• Kubex does not resize PVCs or manage emptyDir sizeLimit.

​

Notes

• Use namespaced proactive policies when teams own their own namespaces.
• Container filtering happens before recommendation age and KubexAutomation checks, so summaries cover only containers owned by the policy.
• By default, only one same-kind proactive policy contributes to a workload. With globalConfiguration.multiPolicyContainerRightsizingEnabled: true, policies can contribute to separate container/resource targets; PolicyEvaluation resolves overlapping targets. See Multi-Policy Container Rightsizing.
• Model is included in default workload types. Use spec.scope.workloadTypes: [Model] when you want to restrict scope to KubeAI Model objects only. Recommendations and rollback state are stored on the Model owner, then inherited by model-owned pods.
• For cluster-scoped examples and field references, see Cluster Proactive Policies.
• EXPERIMENTAL: Recommendations can now include gpu.gpuOverallOptimal, which is applied as a proactive requests.gpu target. See the GPU Sharing with KAI guide for more information.