I can’t figure out what mechanism in the ALB is limiting the load. Increasing the capacity (in this case, ECS containers) doesn’t help; if the load is reduced to 50,000 requests, there are no fluctuations. What could be causing this behavior? What are the possible solutions to the problem?
[Graphics]
[Graphics]

I have a load balancer with two target groups split evenly (50/50). The issue arises with one of the target groups, which may not have any registered targets throughout the day. During this time, the load balancer forwards requests to this target group without any targets, resulting in 5xx errors.