Tools that help diagnose managed memory related issues
I was writing an internal wiki page on performance and thought this info is useful to many external readers as well so here it goes.
vadump is a good start. It’s an mstools tool – meaning you can find it on your NT CD under bin\mstools. You can take a snapshot of the process and see if the GC heap is an issue or not. It was created a long time ago before managed code was popular so the GC heap is part of “Other data” in the output of vadump (heh…:P). If it’s a small portion of your working set it’s better to look elsewhere. It shows a lot of useful info though and I’ve worked with many external customers that didn’t know about it so I thought it’d be useful to mention it.
Type vadump -? for help. Below is an example using vadump:
E:\clr\ndp\clr\src\VM>tlist | findstr /i devenv
1812 devenv.exe Microsoft Development Environment [design] – gcpriv.h
E:\clr\ndp\clr\src\VM>vadump -so -p 1812
Catagory Total Private Shareable Shared
Pages KBytes KBytes KBytes KBytes
Page Table Pages 116 464 464 0 0
Other System 127 508 508 0 0
Code/StaticData 3470 13880 1000 7828 5052
Heap 6521 26084 26084 0 0
Stack 51 204 204 0 0
Teb 14 56 56 0 0
Mapped Data 943 3772 0 444 3328
Other Data 6595 26380 26376 4 0
Total Modules 3470 13880 1000 7828 5052
Total Dynamic Data 14124 56496 52720 448 3328
Total System 243 972 972 0 0
Grand Total Working Set 17837 71348 54692 8276 8380
Module Working Set Contributions in pages
Total Private Shareable Shared Module
10 4 6 0 devenv.exe
75 4 0 71 ntdll.dll
64 3 5 56 kernel32.dll
… [more modules omitted]
20 2 0 18 NETAPI32.dll
Heap Working Set Contributions
2564 pages from Process Heap (class 0x00000000)
0x00240000 – 0x00340000 242 pages
0x02A80000 – 0x02B80000 153 pages
0x02E10000 – 0x03010000 482 pages
0x1AF10000 – 0x1B310000 911 pages
0x207D0000 – 0x20FD0000 776 pages
13 pages from Private Heap 0 (class 0x00001000)
0x00350000 – 0x00360000 13 pages
… [more heap data omitted]
Stack Working Set Contributions
33 pages from stack for thread 00000DE8
2 pages from stack for thread 0000071C
0 pages from stack for thread 00000CEC
1 pages from stack for thread 00000B3C
… [more stack data omitted]
Usually when we investigate issues, a perfmon log is essential. It shows you the size of each generation in the GC heap, the number of collections done on each generation, the number of GC handles and etc. And most importantly it shows you a trend so you know how your app behaves over a period of time which is often extremely important for discovering problems. See my GC perf counter blog entry for an explanation of these counters.
When people have trouble finding out what’s holding onto memory CLRProfiler is a great tool to help figure that out. The CLRProfiler is a visual tool that shows you a tremendous amount of info on the managed memory usage such as the types of the objects that were allocated and relocated, when GCs were triggered and how much memory is allocated by managed API calls. Since it comes with a great in-depth document I won’t repeat stuff here.
The SoS(Son of Strike) debugger extension is also a very useful and powerful tool. It gives you insights that are not easily found in other tools. It comes with the latest Windows Debugger Package (sos.dll in the clr10 directory). Michael Stanton has a nice blog entry about using it.
Abhi Khune has a blog entry that illustrates using some of the tools mentioned above to solve memory related problems. Please take a look at it here.