Guardrails That Stop It

• Multi-Replica Baseline: staging and production start with a minimum of 2 replicas for all critical services.
• Scaling Bounds: Every service must define minReplicas and maxReplicas and target resource utilization (CPU/Memory).
• Disruption Budgeting: Every service must have a Pod Disruption Budget (PDB) to prevent unsafe disruption during maintenance.
• Pre-Drain Verification: Never execute a node drain or rollout without first checking the PDB and HPA state.

Expected Baseline

• HPA (autoscaling/v2): Configured for both backend and frontend in all environments.
• PDB (policy/v1): Configured to allow exactly 1 disruption at a time for multi-replica services.

Safe Workflow (Step-by-Step)

1. Preflight Check: Verify the current HPA status and its defined bounds.
2. Confirm PDB Allowance: Ensure ALLOWED DISRUPTIONS is greater than 0.
3. Trigger Disruption: Perform the maintenance action (e.g., node drain).
4. Observe Scaling: Watch the HPA react if load shifts to the remaining pods.
5. Final Verification: Confirm the workload remains available and returns to its desired replica count after the action is complete.

This builds on: Resource management (Chapter 08) — HPA and PDB build on resource contracts. This enables: Observability (Chapter 10) — availability signals feed the monitoring stack.