running out of mbuf clusters

[pete wright]

On 12/12/06, John Oxley <john at yoafrica.com> wrote:
> Hi,
>
> I'm running a shaper system on a FreeBSD box.  It is pushing sustained 8
> Mbps.  In the messages log I'm getting lots of
>
> Dec 11 21:03:54 ritalin /kernel: All mbuf clusters exhausted, please see
> tuning(7).
>
> When I run netstat -m however I get
> $ netstat -m
> 1766/5904/80000 mbufs in use (current/peak/max):
>         1766 mbufs allocated to data
> 1765/5900/20000 mbuf clusters in use (current/peak/max)
> 13276 Kbytes allocated to network (22% of mb_map in use)
> 30 requests for memory denied
> 0 requests for memory delayed
> 0 calls to protocol drain routines
>
>
> Why is it saying its run out of mbuf clusters when it peaked at 5900?
> The machine is running FreeBSD 4.9-RELEASE.  The reason for this is that
> it is a shaper appliance, and it was shipped to us with 4.9 on it.
>

hmm...don't have a 4.x box handy but on 6.1-RELEASE "man tuning" turns this up:
"
kern.ipc.nmbclusters may be adjusted to increase the number of network
mbufs the system is willing to allocate.  Each cluster represents approx-
imately 2K of memory, so a value of 1024 represents 2M of kernel memory
reserved for network buffers.
"

yikes, can't remember if 4.x allows you to tune this via sysctl or if
you have to define it in your kernel config.  in any event if you type
"sysctl kern.ipc.nmbclusters" that should report you maximum mbuf's
that can be allocated.  I am guessing you may have to increase this
value.

-pete



-- 
~~o0OO0o~~
Pete Wright
www.nycbug.org
NYC's *BSD User Group