Question

A media streaming company runs a stateful session service on an EC2 Auto Scaling group behind a Network Load Balancer. During scale-in events, users report abruptly dropped streams because instances terminate while active sessions are still in progress. Each session can last up to 8 minutes and the application can persist session state to DynamoDB if given time. The DevOps team must ensure that instances chosen for termination complete in-flight sessions before shutting down, without changing the streaming protocol. What is the MOST appropriate way to achieve this?

Accepted Answer

Add an Auto Scaling lifecycle hook for the terminating state that pauses the instance in Terminating:Wait, and use a script on the instance to drain sessions and call CompleteLifecycleAction before the heartbeat timeout expires.

Answer

Increase the NLB deregistration delay (connection draining) to 600 seconds so sessions finish before the instance is removed from the target group.

Answer

Enable instance scale-in protection on all instances in the Auto Scaling group so that no instance with active sessions is ever terminated.

Answer

Configure a target tracking scaling policy with a longer cooldown period to delay scale-in until sessions naturally complete.