Max replicas reached: when autoscaling hits its ceiling
Every auto-scaling group has a maximum. When demand exceeds what the maximum can serve, autoscaling stops helping and the group saturates like a single server.
What causes it
The maximum protects your budget and your downstream dependencies — a database cannot survive unlimited app servers. But if it is set too low for real peaks, requests queue and fail at the ceiling.
How to recognise it
- Replica count flat at the maximum
- CPU saturated across all replicas
- Queueing and errors despite “autoscaling”
How to fix it
- Raise max replicas
- Use bigger replicas
- Reduce per-request cost
Reproduce it in the simulator
Open the simulator, build a small architecture and use Fault injection in the inspector, or push traffic past a component’s capacity.
Learn more
Part of the guide Load balancing and capacity planning.