Question

A company runs a stateless order-processing microservice on Amazon ECS with the Fargate launch type. The service must remain available even if an entire Availability Zone becomes impaired. Currently the service definition specifies a desired count of 6 tasks running in a cluster spanning three AZs, but during a recent AZ outage most tasks were unavailable because they happened to be concentrated in the failed AZ. Which change most effectively ensures tasks are automatically redistributed and remain balanced across AZs after failures and scaling events?

Accepted Answer

Configure the ECS service to use the 'spread' task placement strategy across the 'attribute:ecs.availability-zone' field so tasks are evenly distributed across AZs.

Answer

Increase the desired task count to 12 and enable Service Auto Scaling based on CPU utilization to compensate for lost tasks.

Answer

Switch the launch type to EC2 and configure a 'binpack' placement strategy to consolidate tasks and improve density.

Answer

Attach an Application Load Balancer with cross-zone load balancing enabled so traffic is evenly spread regardless of task location.