1 instance of autoscale group always fails health check even though the application is started.

Question

Using Elastic **Bean Stalk**, **SpringBoot** configured with an autoscaling group using **Elastic Load Balancer** health checks (700 second grace period).

When deploying new version of the application via Elastic Bean Stalk console, only 2 of 3 instances report as healthy via the /health check. 
Health check is positive for 2 healthy instances, but NGINX returns 404 for the unhealthy instance, which is eventually terminated and replaced, over and over.

The logs for the unhealthy instance show the SpringBoot app start just as the others, but the app is unreachable, as if NGINX is unable to proxy requests to new instance. NGINX access logs show failed /health attempts, etc. The fact NGINX starts indicates there is nothing wrong with the EC2 container itself, and it is reachable via it's public IP.

With the exact same code deployed to 2 working instances, what could be wrong with the 3rd?

Answer

Hi, please ensure that the subnets in the Load Balancer and the subnets in the Auto Scaling Group (ASG) match. If they don't match, the load balancer will have no way to reach the instances in those particular subnets.

1 instance of autoscale group always fails health check even though the application is started.

Relevant content