Subject: Re: Simple thought...
To: Wolfgang Rupprecht <wolfgang+gnus20020607T164810@wsrcc.com>
From: Andrew Gillham <firstname.lastname@example.org>
Date: 06/07/2002 17:46:23
On Fri, Jun 07, 2002 at 04:53:15PM -0700, Wolfgang Rupprecht wrote:
> email@example.com (Andrew Gillham) writes:
> > Yes. Try:
> > gdb --write /netbsd
> > set bufpages=80000
> > quit
> > reboot
> Ok. That worked a bit better and grabbed quite a bit of memory.
> total memory = 767 MB
> avail memory = 433 MB
> using 6144 buffers containing 312 MB of memory
> A "du /usr/pkgsrc" still doesn't run that blindingly fast the second
> time around though. There is still plenty of disk activity. I am
> guessing that this is due to the disk not being mounted "noatime".
You might need some other tweaks to adjust the number of buffers as well.
I'm not completely sure, I just know that the bufpages trick is needed when
you have more than a 1GB or ram or so. :(
Look at the number of vnodes in use with pstat -T, and at the maxvnodes
in the kernel. With sysctl kern.maxvnodes. You may be hitting a limit
there as well.