Skip to content

installation.md

Latest commit

 

History

History
101 lines (70 loc) · 2.35 KB

installation.md

File metadata and controls

101 lines (70 loc) · 2.35 KB

Installation Guide

This guide covers installing Workload-Variant-Autoscaler (WVA) on your Kubernetes cluster.

Prerequisites

  • Kubernetes v1.32.0 or later with administrator access or namespace-level permissions
  • Helm 3.x
  • kubectl configured to access your cluster

Installation Methods

Option 1: Kustomize Installation (Recommended)

Using kustomize for more control:

# Install CRDs
make install

# Deploy the controller
make deploy IMG=quay.io/llm-d/llm-d-workload-variant-autoscaler:latest

Option 2: Helm Installation (Deprecated)

See the Helm Installation for detailed instructions.

Verify the installation:

kubectl get pods -n workload-variant-autoscaler-system

Option 3: Local Development and Testing

See the comprehensive deployment guide for detailed instructions.

Integrating with HPA/KEDA

WVA can work with existing autoscalers:

For HPA integration: See HPA Integration Guide

For KEDA integration: See KEDA Integration Guide

Verifying Installation

  1. Check controller is running:

    kubectl get deployment -n workload-variant-autoscaler-system

  2. Verify CRDs are installed:

    kubectl get crd variantautoscalings.llmd.ai

  3. Check controller logs:

    kubectl logs -n workload-variant-autoscaler-system \
  deployment/workload-variant-autoscaler-controller-manager

Uninstallation

Kustomize:

make undeploy
make uninstall  # Remove CRDs

Helm:

helm uninstall workload-variant-autoscaler -n workload-variant-autoscaler-system

Troubleshooting

Common Issues

Controller not starting:

  • Check if CRDs are installed: kubectl get crd
  • Verify RBAC permissions
  • Check controller logs for errors

Metrics not appearing:

  • Ensure Prometheus ServiceMonitor is created
  • Verify Prometheus has proper RBAC to scrape metrics
  • Check network policies aren't blocking metrics endpoint

See Also:

Next Steps