I think I'll try a dump, it's the simplest way of dealing with it because
it is not necessary to also install the symbols. I am somewhat familiar with
windbg :P Ahhh.. the old days of softice & windbg...
> You can debug a production process with common tools like windbg (no
> major installs). I would take a look what is going on using windbg ..
> you can find out alot more about what is happenning as opposed to
> using process explorer. If windbg is too much, atleast make a dump ..
> you can load up symbols on another box.
>
>
http://www.eggheadcafe.com/articles/20060114.asp includes some basic
> instructions. If you are not familiar with windbg it can be quite a
> learning curve. John Robbins has a great book on the subject and there
> is alot of documentation sprinkled around the internet but it will
> take a bit of time until you are comfortable in what you are looking
> at.
>
> Also as a warning .. windbg's interface is a umm acquired taste :)
>
> Cheers,
>
> Greg Young
> MVP - C#
>
http://codebetter.com/blogs/gregyoung > I would not think it likely though to have the JIT "mclaurentiu"
> <mclaurentiu@yahoo.com> wrote in message
> news:b97bd44c16f0a8c86f0c663268a8@msnews.microsoft.com...
>
>> Hi.
>>
>> I have a weird problem with an ASP.NET application which uses Crystal
>> Reports.
>>
>> Sometimes (I hate this), when the ASP.NET application attempts to
>> generate
>> a report, the number of threads of the w3wp increases dramatically
>> and
>> then comes to a stop (probably because the thread pool is full).
>> We've
>> been able to somewhat inspect the w3wp using Process Explorer (its a
>> build
>> environment, no debug assemblies, no debuggers installed and no
>> permission
>> to install almost anything, etc...) and we've seen an interesting
>> behaviour:
>> - when pressing the button that starts the report generation, a new
>> thread
>> is created (by the framework ?!) which runs GetCompileInfo;
>> - then, about twice per second, two or more threads are created, all
>> of
>> them stuck in GetCompileInfo;
>> - the step above repeats itself until a critical number of threads is
>> reached (in my case around 440, I suppose because of the fact that
>> maxworkerthreads = 100 multiplied by 2 processors multiplied by
>> hyperthreading on each processor) and the entire w3wp comes to a
>> halt.
>> there is a lot of I/O and context switches, but no noticeable CPU
>> usage
>> whatsoever;
>> - also, "% Time in JIT" is around 93% - 98%;
>> Some more details: .NET Framework has version 1.1.4322; the report
>> generates fine multiple times before the crashes (with the same
>> report settings); we do not use dynamic code generation;
>>
>> Could you suggest anything? Some documentation, some tips, maybe?
>>
>> Thanks a lot,
>> Laur.