Problem with unmanaged Memory

b.hemmerb.hemmer Posts: 5
edited December 4, 2014 8:36AM in ANTS Memory Profiler 8
Hello,

we have a huge problem with unmanaged memory in our .NET Application. So we decided to analyze the problem with the ants memory profiler. After the startup of the application we took a snapshot and before closing the application. Between that we did more Snapshots but i´m refering the first and last one.

On the first snapshot we had that parameters:
  1. Managed Memory: 129 MB
  2. Unused Memory: 33 MB
  3. Unmanaged Memory 265 MB

On the second snapshot we had that parameters:
  1. Managed Memory: 122 MB
  2. Unused Memory: 40 MB
  3. Unmanaged Memory 324 MB

A look on the unmanaged memory breakdown by module showed us that every module used approximately the same memory. The only exception is the user32 module. That used 4 MB more memory due to the first snapshot. But that does not declare the 60 MB unmanaged memory difference between snapshot 1 and 2. I´m guessing that we have a problem with a leak on the stack and not on the heap. Are there some tools by ants memory profiler to detect a leak like that?

with best regards

bernd hemmer

Comments

  • Jessica RJessica R Posts: 1,319 Rose Gold 4
    Hi Bernd,

    It does look like the stack memory is what's growing, but I'm afraid we don't have a tool that will show more details about the stack. :/

    You may find windbg useful--you can use ~*k to dump call stacks for all threads.

    Hope that might help!

    Kind regards,

    Jessica Ramos

    Jessica Ramos | Product Support Engineer | Redgate Software

    Have you visited our Help Center?


  • Hello,

    i tried that. I found that we have problems with our handles. When i´m looking in the tskmgr the handle count grows staticly and is not going back. I got the information via windbg that the thread and event handles are growing. But i dont get the information at which point they are allocated. Do you have some idea on that?

    with best regards
    bernd
Sign In or Register to comment.