0

I have been having trouble for the past week where one of my web apps running on IIS is hanging at random times at least once a day, I have to recycle the app pool to kick it back into life. CPU and memory seems normal.

When I look at the active threads in the worker processes section in IIS manager i can see threads building up as if they are stuck, when it hangs there is around 25+ threads there, when running ok there is 3-6 at any one point.

I'm guessing this could be down to a code change on my C# web app, but I can't work out where / what it is, there has only been a few minor changes made from what I can remember.

The last week I have really tried hard to diagnose with several tools such as WinDbg & dotTrace to see if I can track down anything obvious, but i'm a little over my head at this point.

I can see there are lots of threads with Timeout errors connecting to my NoSql database (RavenDb), however I think this is a red herring and is down to the threads being blocked in IIS as I can connect to the same database from another IIS app just fine, as well as use the database management tools to manage/query the database.

I have mini dumps and dotTrace snapshots to play with.

Here is the results from "~*e !clrstack" on WinDbg: https://gist.github.com/phinett/7901d82fa526696d3c92

Any help / ideas on what I should be looking out for to track down the culprit? I am happy to provide access to minidumps if need be.

Thank You!

Paul Hinett
  • 1,205
  • 3
  • 11
  • 19
  • Phew 246 threads to look at... Do `!analyze -v -hang` or SOSEX `!dlk` show something? – Thomas Weller Oct 13 '15 at 20:57
  • thank you for the reply! it looks like i'm missing some symbols?...i have MS symbols loaded and called .loadby sos clr...https://gist.github.com/phinett/fbd597c904e9a3045bae (this contains loads more of the same errors so cut it short), and !dlk output: Examining SyncBlocks... Scanning for ReaderWriterLock instances... Scanning for holders of ReaderWriterLock locks... Scanning for ReaderWriterLockSlim instances... Scanning for holders of ReaderWriterLockSlim locks... Examining CriticalSections... No deadlocks detected. – Paul Hinett Oct 13 '15 at 21:08
  • Oh yeah, that bad: [Known issue with kernel symbols](http://stackoverflow.com/questions/32278634/is-there-a-known-issue-relating-to-windows-7-kernel-symbols) – Thomas Weller Oct 13 '15 at 21:12

1 Answers1

0

Probably a lot simpler and quicker to start with DebugDiag v2 update 1 and force a process dump then run the hang analysis.

Greg Askew
  • 35,880
  • 5
  • 54
  • 82
  • I have done this also...and have the results here: http://stats.house-mixes.com/w3wp.mht (think IE is required to open?)...just not quite sure how to decipher the report. – Paul Hinett Oct 14 '15 at 00:16