IIS 5 & IIS 6
Runaway worker process
Last post Jan 09, 2009 10:23 PM by lyndonholmes
Aug 20, 2008 10:44 AM|hnugz|LINK
We have an asp.net 2.0 application on 4 win2k3 sp2 clustered servers using IIS 6. Every so often, we will see our application "freeze" at certain points. When this happens, one of our servers will have a w3wp.exe process that is consuming almost all of
the cpu and lots of memory. I have perfmon setup and when we see this problem, the Web Service -> Current Connections will start climbing. Normally each server has less than 10 at any one time, but when this happens it will just constantly go up.
I also checked the IIS logs on the server when this happens. I noticed two entries where the "time taken" is over 500000. The next top 100 or so time taken entries are around 300000. The problem is these don't tell me the exact issue since the reports
of the app freezing started about 30 minutes prior to these entries.
We believe there is a problem in the code where it will cause a looping of some sort that causes this issue. The problem is finding it. This is a large app and I need some place to be able to focus on.
Any reccomendations of tools to use? Is there a way to take a dump of what exactly this worker process is doing when this is happening? Thanks.
Aug 21, 2008 04:39 PM|joelangley|LINK
Have you tried debugdiag
Aug 21, 2008 05:03 PM|ma_khan|LINK
I would double on DebugDiag for taking IIS dumps... but I am a little confused as to how can your APP create concurrent connections as you say it could be a possibility?
Anyways if you don't want a because of this issue.... for the time being do this:
Right-Click website -> Properties -> Performance. Under Website connections change the value of connections limited to a value a little higher than your Peak value... that will atleast help you keep going ...
Another thing you can use as a workaround is to shutdown the w3wp after it reached 85-95% CPU. This can be done after it reaches such a state say 5 times... but this will disable your AppPool ....
Anyways.. is this scenario only on 1 of your clustered server or on all of them ... and that too at the same time or different time...
Aug 22, 2008 10:08 AM|hnugz|LINK
Thanks for the responses,
I will take a look at debugdiag.
We currently have connections set to unlimited.
It happens on different servers in the cluster at different times.
Jan 09, 2009 10:23 PM|lyndonholmes|LINK
Check out IISPeek. Tell them Lyndon sent you.