[jcifs] jcifs.smb.SmbException: No description available[ERRS RV/90] - 15 Minutes per day

[Allen, Michael B (RSCH)]

> -----Original Message-----
> From:	Paul Gutenbrunner [SMTP:pagu at pagu.at]
> 
> > I wish there was a way for me to test this. I don't have a domain
> > controller handy that I can use to create hundreds of accounts on.
> 
> we also don't have a test-environment in which we can produce hundreds of
> accounts. during the test-phase, this error hasn't occured, and we cannot
> use the production environment for testing...
> 
> how can multiplexing be disabled? is there a possibility to determine how
> many
> sessions are multiplexed over the same socket?
> 
	Well I don't think we want to "disable" it. It's a huge waste of resources to
	open additional sockets. In fact your server would probably run out of memory
	after a few thousand logons.

> >
> > Having said that, ideally if you're using HTTP GET requests and the
> > NtlmPasswordAuthentication object (NPA) is being cached in the session
> > properly the Filter sets jcifs.smb.client.soTimeout to 10 minutes. Meaning
> > someone would have to log in every 10 minutes to keep the client from
> > closing the transport and releasing all sessions over it. So in theory you
> > should be able to handle more than 150 users at a time provided there is
> > an occasional 10 minute period between logins so that the transport is
> > dropped and users get by on the cached NPA in the HttpSession.
> >
> > If you could tell me more details taking into consideration the above
> > perhaps I can tell you what the ideal solution is or make a change that
> > will  make this smoother.
> >
> 
> here are a few more facts which could be helpfull:
> - during the 15 minutes "error-window", there are many successfull
> smb-logons.
>   why do they succeed while others fail?
> 
	I don't know. Maybe the domain controller is getting a free slot for another logon.

> - if a user gets the errsrv/90 error, a simple reload in the browser is
> enough
>   to get the things working.
> 
	Ditto. Or maybe it's finding another domain controller. How are you specifying the
	domain controller? Are you using jcifs.smb.client.domain and jcifs.netbios.wins?
	Maybe if you set jcifs.netbios.cachePolicy = 0 it will not cache the domain name
	and WINS will give you different server for a crude load balancing.

> - after the 15 minutes "error-window", the application works perfectly
> (until
>   the next "error-window" in the afternoon), with hundreds and thousands of
>   additional authentications.
> 
	Hundreds of thousands! Are you sure the NtlmPasswordAuthentication object is
	being cached in the HttpSession? Or are you using HTTP POST?

	I'll look into a proper solution for this. Ideally I think the session to the domain
	controller should just timeout after a certain period of time if the session has been
	idle. Disabling multiplexing over transport or explicitly closing transport periodically
	is not the correct way to deal with the problem.

	Mike