How do you make a NestJS API with PostgreSQL and Redis resilient to the loss of a whole Availability Zone on AWS?

Spread every layer across at least two Availability Zones (AZs), ideally three. An AZ is one or more separate data centres inside an AWS Region.

Load balancer and app:

• The ALB (Application Load Balancer) spans public subnets in each AZ.
• The ECS service runs tasks in private subnets across those AZs.
• Keep enough spare capacity so the remaining AZs can carry the full load, and use autoscaling to add more tasks.
• Tasks must be stateless: keep sessions in Redis or in tokens, not in the task's memory. Then any task can serve any request.

Data:

• Run RDS or Aurora in Multi-AZ mode, so the database fails over to another AZ automatically.
• Make the app reconnect and retry when it gets connection errors, because a failover drops open connections.
• Use ElastiCache for Redis with replicas in other AZs, and turn on automatic failover.

Network: use one NAT gateway per AZ. Then an outage in one AZ does not cut outbound internet traffic for the other AZs.

Finally, test it. Trigger a database failover and stop the tasks in one AZ. Confirm that the system recovers.