IIS suddenly becomes unresponsive

Code profiling for .NET developers.

Moderators: AndrewH, melvyn.harbour, james.billings, Laura Morley, dene.boulton

IIS suddenly becomes unresponsive

Postby JamesO » Thu Apr 26, 2012 5:54 pm

IIS: 7
Windows: Server 2008
Version: 7

We are running an ASP.NET website (.net frametwork version 4) and I am attaching the profiler to the w3wp process. It all works fine until suddenly the IIS becomes unresponsive and no pages can be loaded. This only happens when the profiler is attached.

When this happens the nice chart in the profiler does of course drop to zero activity.

To make the web site working again I can for example recycle the application pool (on which only 1 web site is running).

Has anyone else had this same problem?
JamesO
 
Posts: 3
Joined: Thu Apr 26, 2012 5:49 pm

Postby Brian Donahue » Mon Apr 30, 2012 10:53 am

Hi,

ANTS Profiler runs its' application pool outside of IIS, so more than likely what has happened is that attaching the profiler made the website crash. Unfortunately, because of the design of IIS, we are unable to advise you about why it crashed because IIS hides all of its' error information and every time this happens we end up debugging it on a case-by-case basis.

Your best bet is to check the application event log and see if w3wp.exe reports why it had to terminate.
Brian Donahue
 
Posts: 6670
Joined: Mon Aug 23, 2004 10:48 am

Postby Brian Donahue » Wed May 02, 2012 11:48 am

You may also be able to get at least some data if you use the "fastest" profiling mode - the thing that causes the most issues with IIS is profiler's "line-level" code, which intrudes into assemblies and tries to get the metadata. This causes failures when the metadata has been scrambled up by obfuscators and/or code access policy denies the reflection permission.
Brian Donahue
 
Posts: 6670
Joined: Mon Aug 23, 2004 10:48 am

Postby Brian Donahue » Wed May 09, 2012 11:17 am

You may also be able to find the source of the problem in IIS by enabling request tracing. This will create a series of logs in the IIS logs folder.

First, you go to the server manager (or programs and features) and under ths IIS role->Health and Diagnostics, tick "Tracing".

Then you can configure request tracing for the website.

You should get details about every error in a separate log file in c:\\inetpub\\logs\\FailedReqLogFiles.
Brian Donahue
 
Posts: 6670
Joined: Mon Aug 23, 2004 10:48 am

Postby JamesO » Wed May 09, 2012 11:49 am

Is there any chance that running a page request with hundreds of "Response.Flush()" in a short period of time will cause problems for the profiler? We have had such page running everytime we had crashes when profiling. Now when we have not used that page during profiling there hasn't been any problems.
JamesO
 
Posts: 3
Joined: Thu Apr 26, 2012 5:49 pm

Postby Brian Donahue » Wed May 09, 2012 12:16 pm

It wouldn't cause a problem for Profiler - profiler is agnostic to the code that runs 99.9% of the time. But the coding may be causing a problem that makes w3wp crash. I wouldn't presume anything unless there was some evidence, and like I said, most of this evidence gets buried by IIS.
Brian Donahue
 
Posts: 6670
Joined: Mon Aug 23, 2004 10:48 am

Postby JamesO » Wed May 09, 2012 12:45 pm

What about long running (> 1 minute) page requests that does a lot of chruncing, including cpu and io? Could that be a problem for the profiler?
JamesO
 
Posts: 3
Joined: Thu Apr 26, 2012 5:49 pm

Postby Brian Donahue » Fri May 11, 2012 3:32 pm

Profiler shouldn't be ended by a long-running method.
Brian Donahue
 
Posts: 6670
Joined: Mon Aug 23, 2004 10:48 am


Return to ANTS Performance Profiler 7

Who is online

Users browsing this forum: No registered users and 0 guests