Application intermittently holding requests open

Question

I have a strange case where one of my applications is causing the IIS (7.0) request queue to fill up. Requests are not being terminated after 30s as they should be. This then takes all DB connections from the pool and renders that app useless (other apps are unaffected).

I have no idea a) Why they are stalling in the first place, and b) why IIS is letting them sit there stalled rather than killing them. I would guess my app is locking something, perhaps something the GC is trying to reclaim.

My question is where do I start on debugging such an issue? I have no idea. It's currently happening only in production, but reasonably regularly (maybe once every 4 hours) on all web servers.

PS: There is potentially an argument that this question is better on serverfault than on SO, but given that I think this is a development problem with the app rather than an admin one, I have started on SO for now. I am however happy to re-post there if needed.

Answer 1

For reference using WinDbg was the solution I used. I attached WinDbg to w3wp process for the app pool once requests had queued. I could then view the call stack in each process, and although different most of them were sat waiting on a lock inside ResourceManager.

I still don't know why it was locking there, I thought ResourceManager was thread safe. I re-wrote some code to cache the ouput of ResourceManager in another class and that seems to be avoid the lock.