Debugging IIS random crazy memory usage

Question

I've an ASP.NET MVC 5 website hosted as a Windows Azure Cloud Service. Most of the time, the website is running very smoothly (I'm caching things outside of the instance, using Azure Cache). But from time to time the website is going completly crazy, eating all the memory. I tried restarting IIS and even rebooting the instance, the memory stays full (w3wp.exe eating everything), lot of hard faults, etc. It lasts like 30min or 1h, where pages are very slow to display (or even timeout), then goes back to normal "by itself". I'm not able to reproduce this on my development machine. I tried memory analysis (analyze > performance and diagnostics in VS2013) and wasn't able to locate anything crazy. Even New Relic doesn't tell anything.

The question is: is there a way to debug that kind of problem directly "in production" by performing something like a constant memory profiling? I'm able to access the instance using Remote Desktop.

NB: it's not a big deal if there is a performance impact, it's a one time thing and nothing could be worst than a 1 hour unavailability 4-5 times a day...

Answer 1

If you can RDP the box, than you might want to install one of the following:

Profiling tools:

• Ants profiler
• Ants memory profiler

APM:

• New Relic (they have a two week PRO trial)