Date: Wed, 14 Jan 2004 16:07:48 +0000 From: Ollie Cook <ollie@uk.clara.net> To: Doug White <dwhite@gumbysoft.com> Cc: freebsd-stable@freebsd.org Subject: Re: nfs send errors 32 and 35 on RELENG_4 Message-ID: <20040114160748.GJ27744@mutare.noc.clara.net> In-Reply-To: <20040113114525.L63732@carver.gumbysoft.com> References: <20040113154932.GE354@mutare.noc.clara.net> <20040113114525.L63732@carver.gumbysoft.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Jan 13, 2004 at 11:50:41AM -0800, Doug White wrote:
> > Jan 13 14:02:02 mese /kernel: nfs server 192.168.1.1:/vol/vol1/claramail: not responding
> > Jan 13 14:02:03 mese /kernel: nfs server 192.168.1.1:/vol/vol1/claramail: is alive again
>
> There's some tuning options for this, which I don't immediately recall.
> Under heavy load these are somewhat normal.
Hi Doug,
I have set the kernel to auto-scale nmbclusters based on the memory in the host
in question. I think it's not worth hard-coding these values since the peak
value seems not to get near the maximum:
root@mese:[conf] (10) # netstat -m
397/1200/34816 mbufs in use (current/peak/max):
294 mbufs allocated to data
103 mbufs allocated to packet headers
215/676/8704 mbuf clusters in use (current/peak/max)
1652 Kbytes allocated to network (6% of mb_map in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines
If peak does get close to max, I will increase the number of nmbclusters, but
it doesn't look necessary at present.
All sysctls are at default values.
Are there other things I can be looking at tuning? These hosts do approximately
500 NFS operations each per second (appx 5Mbit/s).
> > Jan 13 14:09:37 mese /kernel: nfs send error 35 for server 192.168.1.1:/vol/vol1/claramail
> > Jan 13 14:09:53 mese /kernel: nfs send error 32 for server 192.168.1.1:/vol/vol1/claramail
>
> These errors tend to imply resource shortages. Monitor netstat -m output
> and make sure you aren't running out of mbuf or mbuf clusters. Also check
> for network errors and dropped packets (netstat -s, switch statistics).
Prior to power cycling the box I was able to look at netstat -m, and netstat
-i, neither of which showed anything to worry about. Next time it happens I'll
be sure to take a copy of the output, in case there's something I'm missing.
Is there any way of finding out what error 35 actually means?
> Are you running rpc.lockd?
The server is an Network Appliance F-Series filer which runs a locking manager:
root@metis:[conf] (13) # rpcinfo -p 192.168.1.1 | grep lock
100021 4 tcp 607 nlockmgr
100021 3 tcp 607 nlockmgr
100021 1 tcp 607 nlockmgr
100021 4 udp 606 nlockmgr
100021 3 udp 606 nlockmgr
100021 1 udp 606 nlockmgr
Thank you for your help so far.
Cheers,
Ollie
--
Oliver Cook Systems Administrator, Claranet UK
ollie@uk.clara.net +44 20 7903 3065
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040114160748.GJ27744>