Reliability Intelligence: Ensure zone failover capacity

When a system fails over an availability zone, other availability zones need to pick up the slack. Make sure you have enough replicas and other resources in all available zones to account for sudden increases in traffic.

‍

Considerations for Kubernetes

• Use Pod Topology Constraints to ensure pods are distributed evenly across availability zones
• Configure a high enough minimum replica count for your service to ensure each availability zone can handle a sudden increase in traffic in response to zone fail over.
• Ensure that horizontal or vertical autoscaling can increase capacity in a given availability zone when demand in that zone increases in response to fail over.

‍

Related Resources

• Best Practices for Testing Zone Redundancy
• Failover network calls to healthy availability zones to reduce errors when a zone becomes unavailable.

‍