Ok, I just now removed all dependencies from my filters from my controllers.
This means none of my filters tries to use a Dao of it's executing
controller. Everything gets injected via the kernel. Still I am getting the
same errors, which is quite a bummer and leaves me to suspect, that
accessing ISession in an IFilter at all creates this bug (which would be
_quite_ a drag). I am under the impression that the default implementation
of ISessionManager simply does not have a look, whether the caller comes
from a different thread or the same that opened the session.

Could anybody of the maintainers of the Nhibernate Integration Facility
confirm this? Or at least tell me where to look...

2009/10/23 Jan Limpens <[email protected]>

> We found a likely issue, but have not yet implemented it:
>
> From filters (which are probably assyncronous) we are accessing entities
> and daos from the controllers. I imagine this can lead just to our
> situations, but this is unconfirmed so far.
>
> 2009/10/23 Jan Limpens <[email protected]>
>
> Well so far I have not heard anything from Ayende, unfortunately.
>> My setup is kind of similar, I think.
>>
>> What I tried was to change the lifestyle of my daos - no result.
>> disabling AutomaticTransactionManagement did not do anything as well.
>> So my guess is on NHibernateIntegration. But I am unsure if testing the
>> same with rhino commons unitOfWork will invalidate atm...
>>
>> 2009/10/19 Artur Dorochowicz <[email protected]>
>>
>>
>>> Francisco,
>>>
>>> Sorry for such a late reply, it was a busy time recently.
>>>
>>> I have tried your suggestion and it did not help. I have completely
>>> disabled Castle Logging Facility and even removed log4net config file.
>>> Still, I get the alert in NHibernate Profiler.
>>>
>>> I see that Jan involved Ayende in the discussion on this topic on
>>> Castle Users Group. I hope he will find us a solution.
>>>
>>>
>>> Artur
>>>
>>>
>>>
>>> On 15 Paź, 09:17, "Francisco A. Lozano" <[email protected]> wrote:
>>> > Try to completely disable your own log4net logging and see if nhprof
>>> > still throws these alarms.
>>> >
>>> > Francisco A. Lozano
>>> >
>>> > On Thu, Oct 15, 2009 at 08:33, Artur Dorochowicz
>>> >
>>> > <[email protected]> wrote:
>>> >
>>> > > I have exactly the same problem, although my app is still early in
>>> > > production and I didn't (yet) see my 'server crash'.
>>> >
>>> > > I have the same setup: Castle Monorail + Windsor + NHibernate
>>> > > Integration Facility, and no manual management of threads.
>>> > > Yet, NHProf is alerting about session usage from different threads.
>>> >
>>> > > To investigate a bit more I have configured log4net to log thread ids
>>> > > for NH loggers and indeed I could see that on concurrent requests
>>> > > sessions were disposed on different thread than they had been
>>> created.
>>> > > I'm not sure if this was the only problem and that that was the only
>>> > > thing that triggered NHProf alert.
>>> >
>>> > > I'm also looking forward to seeing this resolved or explained if this
>>> > > is not a real issue.
>>> >
>>> > > Artur
>>> >
>>> > > On 14 Paź, 22:12, Jan Limpens <[email protected]> wrote:
>>> > >> Hi,
>>> >
>>> > >> NhProf keeps telling me that on _some_ requests of my webb app my
>>> > >> session/transactions seem to use multiple threads.
>>> > >> And the server sometimes enters into a race condition, where too
>>> many
>>> > >> requests kill sql server (problems with closing transactions that
>>> are
>>> > >> already disposed of, sounding just like what you would expect from
>>> > >> multiple threads per session).
>>> >
>>> > >> Now, nowhere i my code I am even coming close to managing my
>>> threads.
>>> > >> From my point of view 1 thread should be all I need. I am using
>>> > >> Monorails with castle windsor integration. My daos are created per
>>> web
>>> > >> request, sessions come from castle's NHibernateIntegrationFacility
>>> and
>>> > >> Automatic Transaction Facility, so these might need my love in some
>>> > >> way, but I cannot tell from the outside.
>>> >
>>> > >> Are there some good debug paths to find out what is wrong before I
>>> get
>>> > >> the next server crash?
>>> >
>>> > >> ....I might have posted this to the castle list, but I suppose most
>>> > >> people who could help me there read this list as well, and nh is at
>>> > >> the center of all this...
>>> >
>>> > >> --
>>> > >> Jan
>>> >
>>> >
>>> >>>
>>>
>>
>>
>> --
>> Jan
>>
>
>
>
> --
> Jan
>



-- 
Jan

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"nhusers" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/nhusers?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to