> On Fri, Jan 16, 2004 at 06:14:29PM -0600, Stephen M. Jones wrote:
> > I've experienced vnlock deadlockish behaviour twice today since increase
> > kern.maxvnodes to ~25% (250000) of system memory (1GB).  Both clients
> 
> I would not expect increasing the total number of vnodes to solve a
> problem caused by a locking bug.

Is it a locking bug though?  Are vnodes not being reclaimed because
there is a locking bug?  Are new vnodes not being allocated because
there are none to be had? 

I'm not 100% sure that it is infact a locking problem.  The reason
why I've increased kern.maxvnodes was that it seemed to be a resource
issue where all available vnodes seemed to have been exhausted.  My
previous numbers showed that the current vnodes in use was very close
to the number of maxvnodes available, which is what led me to believe
that it could have just been a kernel resource issue.

I showed numbers for a machine that has been up for 91 days with a
consistent load as well as a fileserver with a low load and of course
the nfs clients that are just getting hammered.

After reading Perry Metzger's suggestion and hearing from David Brownlee,
I thought that it definately could be because I am running loaded machines
with a previous limit of around 21900 per client (including the nfs server)