Question

A DevOps team manages runtime feature flags and dynamic configuration for a fleet of microservices running on ECS Fargate. Currently, all configuration is stored as plaintext values in SSM Parameter Store, and applications poll the parameters every 60 seconds. Recently, a bad configuration value was pushed that instantly propagated to all tasks, causing a widespread outage before anyone noticed. The team wants a solution that validates configuration before deployment, gradually rolls out changes across the fleet, and automatically rolls back if a CloudWatch alarm fires during the rollout. Which approach BEST meets these requirements?

Accepted Answer

Migrate the configuration to AWS AppConfig, define validators (JSON Schema or Lambda), and use a deployment strategy with a bake time and associated CloudWatch alarms for automatic rollback.

Answer

Continue using SSM Parameter Store but enable Parameter Store change history and create an EventBridge rule that alerts on parameter updates so the team can manually revert.

Answer

Store configuration in a versioned S3 bucket and have each ECS task download the latest object on a schedule, using S3 versioning to roll back manually if needed.

Answer

Use SSM Parameter Store advanced tier with parameter policies that set an expiration, forcing tasks to re-fetch and validate values before they take effect.