Session disruption when downscaling ECS Fargate tasks

Agents
1

This question originally came up in our Slack community and the thread has been consolidated here for long-term reference.

I’m self-hosting LiveKit agents on AWS ECS Fargate. Everything works except that when tasks are downscaling, some active sessions get disrupted.

What’s the recommended workaround for this? Should I switch to ECS with EC2 launch type?

2

Fargate is not recommended for LiveKit agents because ECS Fargate has a hard shutdown limit of 120 seconds. Sessions longer than 2 minutes can be terminated abruptly.

Solution: Use ECS with EC2 launch type instead.

For EC2 launch type, configure the drain timeout:

  1. Set drain_timeout in your worker configuration: Server options | LiveKit Documentation
  2. Set stopTimeout in your ECS task definition to be larger than the drain_timeout

This allows active sessions to complete before the instance terminates.

See this KB article for additional tips on AWS deployments: