NFS / MBUFs, oh my!

[Eric Anderson]

On a FreeBSD 4.7-RELEASE machine, running as a super heavily loaded NFS 
server, I occasionally get messages like:

All mbuf clusters exhausted, please see tuning(7).
All mbufs exhausted, please see tuning(7).
All mbufs exhausted, please see tuning(7).
All mbuf clusters exhausted, please see tuning(7).
All mbuf clusters exhausted, please see tuning(7).
All mbuf clusters exhausted, please see tuning(7).
All mbuf clusters exhausted, please see tuning(7).
All mbuf clusters exhausted, please see tuning(7).
All mbuf clusters exhausted, please see tuning(7).
(from dmesg - /var/log/messages has been pushed out already)

Here's my current netstat -m:
199/53536/262144 mbufs in use (current/peak/max):
         198 mbufs allocated to data
         1 mbufs allocated to packet headers
155/37152/65536 mbuf clusters in use (current/peak/max)
87688 Kbytes allocated to network (1% of mb_map in use)
156 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines

It's the "156 requests for memory denied" that bothers me, and I think 
is causing my problems, but I'm not sure what more I can do. I read 
tuning, but it doesn't say much more that I have already done (unless I 
read over it).

What can I do to fix this, and does anyone have any good NFS tweaks to 
increase performance (yes, I already have a RAID, I'm talking about 
FreeBSD tweaks).

Eric


-- 
------------------------------------------------------------------
Eric Anderson	   Systems Administrator      Centaur Technology
Attitudes are contagious, is yours worth catching?
------------------------------------------------------------------