One of our ASG’s uses ELB as it’s health check. This is because sometimes a new instance is started but not completely set-up causing, among other problems, the average cpu utilization to drop which will eventually avoid new autoscaling activities.
Currently, we withdraw the instance from the ELB before restarting it’s services and put it back after, we thought doing this would avoid send requests for the machine that is currently being re-deployed. But we have another problem, as soon as the ASG tries to health-check an instance and it’s not registered on ELB, the ASG kills that instance.
The only solution I see for that is to suspend auto-scaling activities before doing any other thing and resume it after. Actually, I’ve tested it using aws CLI and it worked fine this way.
So I would like to know if you guys also think this could be a feature for the ec2_asg module?