I have an ECS cluster configured with ELB healthchecks and autoscaling policy on average CPU. When there is a spike in traffic, some containers stop responding to ELB healthchecks and are marked as unhealthy. The problem is that ECS starts draining connections and killunhealthy containers before starting new ones, so it happens that I can have zero containers running before new ones are started. The autoscaling policy is not quick enough in these situations to start new containers. Ideally I would like ECS to start a new task as soon as it marks a task as unhealthy. Any idea how to solve this issue?
Asked
Active
Viewed 37 times
