Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Our IT team is trying to track down some memory issues we're having in production with a specific customer. Since this is a classic ASP application on Win2K3/IIS6 we've determined that DebugDiag/WinDbg are the best ways for us to take a memory dump and analyze it for root cause of the issue.

Since we can only get a TRUE representation of the issue on production, we want to generate a dump there, but in order to get the green light for ANY kind of tracing in production, we need to somehow calculate the overhead of running the tools (leaktracking dll).

If we can say "running this in production should only add 5% overhead to your system", its more likely we can proceed with our tests.

Any advice on the best way to determine this? Should we use PERFMON counters? If so, which counters would be useful?

Thanks!

jg

share|improve this question
    
Don't the memory leaks occur on test systems? –  Jim B Jan 5 '12 at 19:17
    
Probably, but we can't reliably reproduce the LIVE traffic variations -- and don't really have the expertise to write a script to replicate it. –  tresstylez Jan 5 '12 at 19:57
    
Sounds painful, I can't think of a way to predict impact of windbg on something like that. What makes you think it's a memory leak vs increased men usage during high traffic? Is this a vm (on either hyper-v or VMware) –  Jim B Jan 6 '12 at 0:39
1  
It's impossible to predict - it depends entirely on how the app behaves. If the app doesn't do much, then it'll have next to no overhead. If the app allocates memory with desperation and vigour, the overhead might be sizeable. If the app's CPU bound, there might be no noticeable difference. –  TristanK Jan 6 '12 at 1:50

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.