Subject: Re: increasing NMBCLUSTERS
To: None <tech-net@netbsd.org>
From: Manuel Bouyer <bouyer@antioche.eu.org>
List: tech-net
Date: 09/29/2005 20:00:54
On Thu, Sep 29, 2005 at 11:08:42AM -0400, Jan Schaumann wrote:
> Jan Schaumann <jschauma@netmeister.org> wrote:
> > Jan Schaumann <jschauma@netmeister.org> wrote:
> >  
> > > I'll collect some more output as the machine is running...
> > 
> > Alrighty then, now I have some more input:
> > 
> > http://www.cs.stevens.edu/~jschauma/netbsd/netstats
> > http://www.cs.stevens.edu/~jschauma/netbsd/vmstats
> 
> I now notice that netstat -a shows me a large number of nfs connection
> that sit in CLOSE_WAIT:

This could be a problem if the sendq or recvq were not empty. But it doens't
seem to be the case here. ~70000 mbufs would mean about 1000 mbufs per
connection, this is huge.
Can mbufs stay in a socket somewhere else than in the send or receive queues ?

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--