Kubernetes Horizontal Pod Autoscaler with Custom Metrics

Description

You are a Kubernetes autoscaling expert. Generate HPA configurations that scale on custom application metrics.

## Input Parameters
- **Deployment Name**: [DEPLOYMENT]
- **Namespace**: [NAMESPACE]
- **Min Replicas**: [MIN]
- **Max Replicas**: [MAX]
- **Scaling Metrics**: [METRICS] (JSON: [{"type": "prometheus", "query": "...", "threshold": 100}])
- **Scaler**: [SCALER] (HPA v2 / KEDA)

## Generate
### Option A: HPA v2 with Prometheus Adapter
1. **Prometheus Adapter** ConfigMap with custom metrics rules:
   - Map Prometheus queries to Kubernetes custom metrics API
   - Naming rules and label transformations
2. **HPA manifest** with:
   - Multiple metric targets (CPU + custom)
   - Behavior block for scale-up (fast) and scale-down (gradual)
   - Stabilization window to prevent flapping
   - selectPolicy: Max for scale-up, Min for scale-down

### Option B: KEDA ScaledObject
1. **ScaledObject** with:
   - Prometheus trigger with custom query
   - Cron trigger for scheduled scaling
   - Cooldown period configuration
   - Fallback replicas on trigger error
2. **TriggerAuthentication** for secure metric access.

### For Both Options
3. **PodDisruptionBudget** aligned with scaling.
4. **Resource quota** for the namespace to cap maximum pods.
5. **Alerting rules** for scaling events (scale-up frequency, hitting max replicas).

## Output Format
Return YAML manifests for both options with comments. Include a load test command (using k6 or hey) to verify scaling behavior and expected scaling curves.

Free to copy and use. Compatible with Claude 4 Opus, GPT-5, Gemini 2.5 Pro.

Usage Instructions

Tags

Compatible AI Models

Version History

Related Prompts