I have an autoscaling group that is fronted by an NLB. It currently takes about 4 minutes and 30 seconds for an instance to come up and start taking traffic from the NLB. Updates and services start up within the first 2 minutes on the instance, and when the instance goes in the initial state on the target group, it takes about 2 minutes and 30 seconds to run the health check and start receiving traffic. I’ve reduced the health check intervals and timeouts, but it still takes the full 2 minutes. Since this is an NLB, it’s a TCP check on port 80, and the service is up and running even before the instance is added to the target group. My goal is to reduce the time it takes for an instance to come up and start taking traffic.
Current health check settings –
Protocol – TCP
Port – traffic port (80)
Healthy threshold – 2 consecutive health check successes
Unhealthy threshold – 3 consecutive health check failures
Timeout – 5 seconds
Interval – 10 seconds
- Is this an expected behaviour or am I missing something?