Elastic Load Balancers are configured to send traffic to EC2 instances (for Classic type) or target groups (for Application and Network type). When there are fewer available targets that can receive traffic than expected, the load balancer is considered unhealthy. When this happens, the remaining targets will handle the full amount of requests, which can quickly overwhelm your application.


Effects


  • Increased load on other instances backing the ELB
  • Cascading failures as other instances are unable to keep up with the load to the ELB

 

Fixes


Blue Matador detects when the hosts in your load balancer are unhealthy and alerts you. To troubleshoot unhealthy hosts try the following:

  • Check the health check endpoint for errors
  • Check general host health. Look at application logs, server metrics, or APM metrics to determine what is making your server unhealthy.
  • Use an autoscaling group with your ELB to ensure that you always have enough servers to handle load

 

Resources