Hi everybody, off-topic AWS question. Any of you have experienced terrible response times after putting your EC2 through an application LB? We had this thing running fine in a t2 micro. As soon as we put it behind the LB performance dropped. We had another case were we just removed availability zones from the LB that had no instances on them. But this time I just can't get what's going on.

it could be anything

you could have run out of cpu credits

you need some data to know definitively

profiling the application/JVM profiling the machine monitor the load balancer (in rough priority order of things to investigate)

What's weird is this happens as soon as we put this thing behind the LB. Otherwise things run just fine.

"performance dropped" means "latency increased"? How much?

Yup, from 1-3 seconds to up to a minute. Seems like the first requests you make "warm up" something, after that the application works fine for a while and then takes a dive again.

my $0.02; it is too early to speculate -- find some data about what is going on, then make a hypothesis

CPU normal?

Loadbalancer target group instances healthy?

t2/t3 instances use CPU credits -- are you still running with t2 instances?

starting up the application could have used up all the cpu credits -- load balancer might be a total red herring

CPU normal. Instances show as healthy. No CPU credits since these are being paid on demand. Everything looks fine until I plug this thing to the LB. Direct access to the instances works just fine.