AWS Distributed Load test

Question

Hi , I was testing Distributed load test . I am also having an autoscaling group with an ALB, which was set to scale out in case of CPU utilization of 60 %.

After the test, the CPU utilization reached 10% and the web page of my EC2 instance started giving 5XX errors, so the ALB saw the machine as unhealthy. Accordingly the autoscaling group deleted this machine and created a new one.

This should not have happened. As when we try to access the ec2 web page directly without the ALB during the 10% CPU utilization it is working fine. but from the ALB DNS name it is giving us the 5XX errors

Can someone please help explain why this is happening and what should be edited in the distributed load test ?

Answer

Hello,
Do you have logs from the ALB and also logs from the application/web servers that run inside the EC2 ? I think analyzing the logs (both from ALB and from apps) will help paint a clearer picture of what happened. The CPU load might be ok, but there could be issues on other component, ex: memory, disk I/O, or even application configuration. It happened to me once, and it turned out I wrongly set up my application's concurrency, and it was clearly shown in my application logs.

AWS Distributed Load test

Conteúdo relevante