looking for cause- httperr reports "connection_dropped" an..

 > Unlike prior IIS versions, where IISADMIN and W3SVC were both inside of
 > inetinfo.exe (hence it is affected by user code loaded into it as well as
 > hangs/crashes), W3SVC in IIS6 does not load any user code (it is in a
 > separate svchost.exe, and IISADMIN remains in inetinfo.exe and also does
not
 > load any user code in IIS6 Native Worker Process Isolation Mode) and hence
 > rarely crashes/hangs. Thus, it is rarely necessary to restart W3SVC to
 > correct your issue, though it happens to work. We would be very
interested
 > in hearing any legitimate circumstance that requires W3SVC to restart.
 > All that is usually needed is recycling or re-enabling a disabled
 > Application Pool -- and it is as good as new. Event Log will tell you
when
 > an application pool is being taken "offline" -- and you can go to the UI
to
 > restart the application pool.

You can run the consumer by loading this url -
<a style='text-decoration: underline;' href="http://websvc.bdsrealtime.com/NBDS.Consumer.Tempo/tempo.aspx?uid=ServiceCKGE&stnid=4822" target="_blank">http://websvc.bdsrealtime.com/NBDS.Consumer.Tempo/tempo.aspx?uid=Servi...KGE&stn</a>

If you leave the page open it automatically reloads, until the crash/hang
issue occurs.

This is 2003 Server, IIS 6, SQL 2000, .Net 1.1

Yesterday I loaded the MS IIS Debug Tool that has a crash/hang agent
(IISDebugTools.exe version 1.1) onto the server and setup the ini file
according to the documentation. It was my understanding that on a hang the
agent would write a log and run iisdump. This morning I found the consumer
was failing and IIS was hung. But I had nothing from this tool. I followed
your last post and restarted the pool and that failed to get the web service
and consumer up and running again. I then went into services and restarted
the W3SVC service and that resolved the issue.

After the restart I found I got a crash report from the tool and it was
filled with "Connection_Abandoned_By_Inetinfo_Crash". I don't know if this
was from the restart or if this data was prior to the restart and simply
held until a restart. Nor do I know what that message tells me.

Once the issue occurs I find the httperr log is filled with
"connection_dropped" entries, I don't know what that is telling me either.

Just a little while ago I set the application to report a detailed .Net
error page in the event of a problem, and then let the consumer run in my
browser. After a while I got the error - Server too busy.
System.Web.HttpException: Server too busy.
System.Web.HtttpRuntime.RejectRequestInternal (httpworkerrequest wr)
Would you know what causes that and what prevents it?
That error did not hang W3SVC though, I clicked Refresh and the page
reloaded the consumer just fine.

David Bartosik<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: looking for cause- httperr reports "connection_dropped" an.. 

So when "IIS6 was hung" and you had to restart W3SVC, what requests are
being served?
i.e. For static files, ASP files, ASP.Net files, which do/not get served.

Do you have anything constantly modifying configuration (including IIS
metabase) or leaving the metabase "locked" -- that is one thing that can
cause W3SVC (as well as w3wp.exe) to not respond -- because w3wp.exe is
ACL'd to only read configuration and if the metabase is locked, everyone
waits on it and does nothing.

Can you check perfmon counters to see the ASP.Net and IIS Request Queue
lengths.

What exactly is your application constantly doing when you refresh it? Is
this application making inbound/outbound connections of any sort (including
SQL, HTTP, etc)?

--
//David
IIS
This posting is provided "AS IS" with no warranties, and confers no rights.
//
"David Bartosik MS MVP" <forums DeleteThis @davidbartosik.com> wrote in message
news:etyGPcvfEHA.3272@TK2MSFTNGP11.phx.gbl...
 > Unlike prior IIS versions, where IISADMIN and W3SVC were both inside of
 > inetinfo.exe (hence it is affected by user code loaded into it as well as
 > hangs/crashes), W3SVC in IIS6 does not load any user code (it is in a
 > separate svchost.exe, and IISADMIN remains in inetinfo.exe and also does
not
 > load any user code in IIS6 Native Worker Process Isolation Mode) and hence
 > rarely crashes/hangs. Thus, it is rarely necessary to restart W3SVC to
 > correct your issue, though it happens to work. We would be very
interested
 > in hearing any legitimate circumstance that requires W3SVC to restart.
 > All that is usually needed is recycling or re-enabling a disabled
 > Application Pool -- and it is as good as new. Event Log will tell you
when
 > an application pool is being taken "offline" -- and you can go to the UI
to
 > restart the application pool.

You can run the consumer by loading this url -
<a style='text-decoration: underline;' href="http://websvc.bdsrealtime.com/NBDS.Consumer.Tempo/tempo.aspx?uid=ServiceCKGE&stnid=4822" target="_blank">http://websvc.bdsrealtime.com/NBDS.Consumer.Tempo/tempo.aspx?uid=Servi...KGE&stn</a>

If you leave the page open it automatically reloads, until the crash/hang
issue occurs.

This is 2003 Server, IIS 6, SQL 2000, .Net 1.1

Yesterday I loaded the MS IIS Debug Tool that has a crash/hang agent
(IISDebugTools.exe version 1.1) onto the server and setup the ini file
according to the documentation. It was my understanding that on a hang the
agent would write a log and run iisdump. This morning I found the consumer
was failing and IIS was hung. But I had nothing from this tool. I followed
your last post and restarted the pool and that failed to get the web service
and consumer up and running again. I then went into services and restarted
the W3SVC service and that resolved the issue.

After the restart I found I got a crash report from the tool and it was
filled with "Connection_Abandoned_By_Inetinfo_Crash". I don't know if this
was from the restart or if this data was prior to the restart and simply
held until a restart. Nor do I know what that message tells me.

Once the issue occurs I find the httperr log is filled with
"connection_dropped" entries, I don't know what that is telling me either.

Just a little while ago I set the application to report a detailed .Net
error page in the event of a problem, and then let the consumer run in my
browser. After a while I got the error - Server too busy.
System.Web.HttpException: Server too busy.
System.Web.HtttpRuntime.RejectRequestInternal (httpworkerrequest wr)
Would you know what causes that and what prevents it?
That error did not hang W3SVC though, I clicked Refresh and the page
reloaded the consumer just fine.

David Bartosik<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: looking for cause- httperr reports "connection_dropped" an.. 

I'm afraid most of your questions are Greek to me. We got this new server,
it came with Server 2003, and now I'm trying to figure out IIS6 based on
what little experience I have in IIS5. But only because I don't know why the
server keeps stopping processing the application. Since it's always run
without issues on the IIS5 server I tried to compare configuration in the
IIS5 to the one in IIS6, but that doesn't get me far with all the new stuff
in IIS6, like the app pool sections, I don't know how those settings should
be configured. Though I did just now uncheck a box on limiting request queue
to see if that makes a difference. I saw a setting on metabase editing, I
assume that's what your post refered to, that is off. Yesterday I keep my
browser open with the service consumer url to just let it run, the page
refreshes itself every 50 seconds. The service consumer makes a sql
connection for login validation and then runs the web service which in turn
runs a sql query. The service has try/catch/finally blocks to catch
exceptions and prevent leaving connections open. I changed the config file
on the consumer to give me the detailed .net error page. Now running the
consumer yesterday with this change in the config file I found the page
would frequently fail and give me the error - server error in "my app here",
server to busy, an unhandled exception occured during the execution of the
current web request, details, system.web.httpexception, server too busy,
stack trace-httpexception 0x800004005 server too busy
sysem.web.httpruntime.rejectrequestinternal(httpworkerrequest wr)
At this point I don't know what this is telling me. I found that a simply
manual refresh of the page would clear this error page and the consumer
would run again. I left my browser open running the consumer over night and
found this morning it was "hung" or "crashed" however you want to phrase it,
with the error page now of Server error in "my app here", Request timed out,
an unhandled exception occurred during the execution of the current web
request, details, system.web.httpexception-request timed out, stack
trace-httpexception 0x80004005 request timed out.
I had to stop and restart W3SVC to get the app up and running again.
So now that is restarted and with the request queue limit unchecked. I now
have my browser open again running the consumer and monitoring it.
That's where I'm at and what I know.
thanx David.

David Bartosik
Microsoft MVP
www.publishermvps.com
 >> Stay informed about: looking for cause- httperr reports "connection_dropped" an.. 

I found this kb -
http://support.microsoft.com/default.aspx?scid=kb;en-us;821268&Product=asp
- that looks applicable here. It states to use config file settings that
look to me to be similiar to those in the IIS6 UI. You know if these config
settings are meant to be used in lieu of IIS6 UI settings or to complement
them? Does this KB look to you like it is what I need to use?

David Bartosik
Microsoft MVP
www.publishermvps.com
 >> Stay informed about: looking for cause- httperr reports "connection_dropped" an.. 

The KB seems like a good start.

ASP.Net on IIS6 Worker Process Isolation Mode uses a subset of its .config
settings while relying on others within IIS6.

In particular, the <processModel> settings are all pretty much ignored in
favor of the ones in IIS6 UI (i.e. process identity is set within IIS6 UI,
and not via <identity> element).

However, the following properties from .config files are used by ASP.Net
regardless of IIS settings: requestQueueLimit, responseDeadlockInterval,
maxWorkerThreads, maxIoThreads, minWorkerThreads, minIoThreads. It's
because they are ASP.Net specific and not process-model specific.

--
//David
IIS
This posting is provided "AS IS" with no warranties, and confers no rights.
//
"David Bartosik MS MVP" <forums.DeleteThis@davidbartosik.com> wrote in message
news:ugmNJ27fEHA.3320@TK2MSFTNGP11.phx.gbl...
I found this kb -
http://support.microsoft.com/default.aspx?scid=kb;en-us;821268&Product=asp
- that looks applicable here. It states to use config file settings that
look to me to be similiar to those in the IIS6 UI. You know if these config
settings are meant to be used in lieu of IIS6 UI settings or to complement
them? Does this KB look to you like it is what I need to use?

David Bartosik
Microsoft MVP
www.publishermvps.com
 >> Stay informed about: looking for cause- httperr reports "connection_dropped" an.. 

Changing these values in the machine.config file to the recommendations in
the aforementioned KB resolved the issues. It only took me 3 days of
research to find the KB Now my bosses want to know why this had to be
done with IIS6 and not with IIS5, the fact that I had to restart W3SVC
without these config settings has them believing IIS5 "is more robust" then
IIS6.

David Bartosik
Microsoft MVP
<a style='text-decoration: underline;' href="http://www.publishermvps.com" target="_blank">www.publishermvps.com</a>

"David Wang [Msft]" <someone.DeleteThis@online.microsoft.com> wrote in message
news:uKnEy9OgEHA.396@TK2MSFTNGP12.phx.gbl...
 > The KB seems like a good start.
 >
 > ASP.Net on IIS6 Worker Process Isolation Mode uses a subset of its .config
 > settings while relying on others within IIS6.
 >
 > In particular, the <processModel> settings are all pretty much ignored in
 > favor of the ones in IIS6 UI (i.e. process identity is set within IIS6 UI,
 > and not via <identity> element).
 >
 > However, the following properties from .config files are used by ASP.Net
 > regardless of IIS settings: requestQueueLimit, responseDeadlockInterval,
 > maxWorkerThreads, maxIoThreads, minWorkerThreads, minIoThreads. It's
 > because they are ASP.Net specific and not process-model specific.
 >
 > --
 > //David
 > IIS
 > This posting is provided "AS IS" with no warranties, and confers no
rights.
 > //
 > "David Bartosik MS MVP" <forums.DeleteThis@davidbartosik.com> wrote in message
 > news:ugmNJ27fEHA.3320@TK2MSFTNGP11.phx.gbl...
 > I found this kb -
<font color=purple> > <a style='text-decoration: underline;' href="http://support.microsoft.com/default.aspx?scid=kb;en-us;821268&Product=asp</font" target="_blank">http://support.microsoft.com/default.aspx?scid=kb;en-us;821268&Product...p</f</a>>
 > - that looks applicable here. It states to use config file settings that
 > look to me to be similiar to those in the IIS6 UI. You know if these
config
 > settings are meant to be used in lieu of IIS6 UI settings or to complement
 > them? Does this KB look to you like it is what I need to use?
 >
 > David Bartosik
 > Microsoft MVP
<font color=purple> > <a style='text-decoration: underline;' href="http://www.publishermvps.com</font" target="_blank">www.publishermvps.com</font</a>>
 >
 >
 ><!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: looking for cause- httperr reports "connection_dropped" an..