Hi everyone,
We are experiencing an issue with the health check while using ARR.
It starts with a crashed w3wp process. After this crash one out of four pools is stopped. Now I would expect the health check for this pool to switch to unhealthy so that new requests wouldn't lead to the stopped pool. At first it does, but a few seconds later it automatically switches back to healthy while the pool still stays stopped. Again a few seconds later it is unhealthy and then healthy, and then unhealthy, and then healthy... you get it. Caused by the not working health check users sometimes are led to the stopped pool and get the HTTP 503 "Service Unavailable" error.
This keeps going until I start the stopped pool manually. Then everything is fine again. Until next time a pool will be stopped due to a crashed w3wp process. Unfortunately I cannot reproduce this issue. It occurred two times within the last three months.
I'm aware we have got two problems here: the crashing w3wp process and the health check constantly switching between healthy and unhealthy.
In this thread I would like to analyse the health check problem so that it could fulfil its purpose to move requests correctly when a pool doesn't work. The crashing
processes will be analysed separately.
So what is the right procedure to catch the error? Which information do we need?
I really appreciate your help.