Fargate Spot Tasks Not Restarting After Interruption: Is Auto‑Rescheduling Enabled?

Goal

Deploy a low‑traffic, bursty microservice on Fargate Spot to achieve up to 70% cost savings while maintaining continuous availability.

Constraints

• Tasks receive a 2‑minute interruption notice and may terminate abruptly.
• No EBS volumes are attached; persistent state must be on EFS or external stores.
• The workload must be idempotent to tolerate repeated starts.

Uncertainty

It is unclear whether the ECS service scheduler automatically re‑launches a Spot task after an interruption, or if manual intervention or a custom event‑driven workflow is required. The documented behavior for Spot interruption handling is not fully specified, leaving a gap in understanding the default rescheduling policy.

Questions

1. When a Fargate Spot task receives an interruption notice, does the ECS service scheduler automatically replace it to maintain the desired count?
2. If automatic replacement does not occur, what event or API call must be triggered to restart the task?
3. Are there any service‑level configuration flags or IAM permissions needed to enable or observe this auto‑rescheduling behavior?