

How It Works
Automation Engine operates within your Kubernetes cluster and works with Kubex Analytics in a continuous cycle:- Kubex Analytics analyzes workload utilization and generates rightsizing recommendations
- Automation Engine retrieves recommendations through the Gateway API and evaluates them against your policies
- Policy matching identifies applicable AutomationStrategy and ProactivePolicy resources for each workload
- Safety validation checks Kubernetes constraints (ResourceQuota, HPA, node capacity) before proceeding
- Resource changes are applied using in-place resize, pod eviction, or admission-time mutation
- Continuous monitoring watches for rollback conditions and updates recommendations as workload patterns evolve
Key Capabilities
- Continuous optimization: Automatically applies rightsizing recommendations as workload patterns evolve
- Safety-first approach: Pre-flight validation for ResourceQuota, LimitRange, HPA/VPA conflicts, and node capacity
- Policy-driven control: Declarative CRD-based configuration defines what, when, and how automation happens
- Zero-downtime updates: In-place resize on Kubernetes 1.33+ with eviction fallback for older versions
- Rollback protection: Automatic rollback detection and backoff for changes that cause pod failures
- Resource dimension control: Configure which resources can be changed (CPU, memory, ephemeral storage, GPU) and by how much
- Team autonomy: Namespaced and cluster-scoped policies support both platform-wide and team-specific controls
Architecture
Automation Engine runs within your Kubernetes cluster and consists of:- Automation Engine controller: Evaluates policies, performs safety checks, and applies changes
- Mutating admission webhook: Enables admission-time pod mutation for immediate enforcement
- Gateway sidecar: Retrieves recommendations from Kubex Analytics
- Custom Resources: Declarative CRDs define automation behavior, including core policies, specialized policies for GPU and node optimization, and rollback rules
Policy Model
Policies separate behavior from scope for flexible control:- AutomationStrategy: Defines how automation happens (resource dimensions, safety thresholds, execution methods)
- ProactivePolicy: Defines where automation applies (namespace/workload selectors)
- RollbackPolicy: Defines when to revert changes (failure conditions, backoff behavior)
Supported Workloads
Automation Engine optimizes Deployment, StatefulSet, CronJob, DaemonSet, Job, AnalysisRun, Rollout (Argo), and StrimziPodSet resources.Getting Started
Installation
Install Automation Engine with Helm and configure basic policies
Automation Strategies
Define how automation happens with resource dimensions and safety controls
Proactive Policies
Configure which workloads are automated using selectors and policies
Safety Controls
Understand pre-flight checks and validation rules
Additional Resources
- Configuration Reference: Complete Helm chart options and Custom Resources
- Troubleshooting: Diagnose common issues
- FAQ: Frequently asked questions
- Helm Chart Repository: Official installation source
The older Kubex Automation Controller chart has been deprecated. If you’re using it, see Migrating from Helm-Managed Config.

