I'm using Amazon's elastic load balancing across several machines that are running nginx. The problem is, after running for fifteen minutes or so, ELB starts to mark all of the machines as dead. When we look at the nginx access logs, however, the ELB requests are returning 200's in less than 1 ms (we just have nginx serving the a static file for the ELB healthcheck). So, I'm not sure how to figure out what's going on.
Is there either some way we can get a more detailed description of why ELB thought all of the machines were bad, or any general tips for situations that might cause this? I'm really stumped trying to figure out what I can try next, to debug.