[Sevii]

Don't use it. It's crapware. We have 100+ hosts. Problems include the scheduler assigning tasks to nodes that report their ecs agent as crashed. Just use kubernetes it's going to be more stable and have more support. I wish execs didn't take aws seriously when they promised features 6 months down the line.

[NathanKP]

I'd love to hear more about this problem and see if we can get to a root cause and help you resolve it, because it does not sound like standard ECS behavior. Please email [e-mail] and I can connect you to the right people to figure out what is going on.

[Sevii]

ECS doesn't stop trying to schedule tasks ever. So you can ddos and crash your entire cluster if one if your containers fails on startup.

[Sevii]

Issue is still open on GitHub, you can easily blow through your entire iops budget and have cpu pegged at 95% iowait.