Date: Sat, 19 Jul 1997 14:06:26 -0700 (PDT) From: Dmitry Kohmanyuk <dk@dog.farm.org> To: gilham@csl.sri.com (Fred Gilham) Cc: freebsd-hackers@freebsd.org Subject: Re: NFS wierdness? Message-ID: <199707192106.OAA21295@dog.farm.org>
next in thread | raw e-mail | index | archive | help
In article <199707191439.HAA12910@japonica.csl.sri.com> you wrote:
> I have this user who runs programs that generate lots of output to
> stdout. The output sometimes contains very long `lines'---long
> stretches of output without a newline. The programs are also supposed
> to be very CPU intensive.
[...]
> Unfortunately when he does this on a FreeBSD box where the directory
> he's writing to is NFS mounted, he gets almost no CPU time; the
> process spends most of its time in the `nfsfsy' state (according to
> top).
[...]
First, some questions:
- what is your NFS server?
- is NFS mount v3 or v2?
On last FreeBSD-SFUG meeting, I talked to Dyson and he confirmed
that NFS appends are broken in FreeBSD; I do not know does this relate
to 3.0 or 2.2 as well.
I have got full-zero blocks written to mail folders opened in append mode,
sometimes (folders were written by mail user agent, mutt, not by MTA),
so there were no concurrency / locking problems - only one process involved.
Also, NFSv3 mount on FreeBSD 2.2.2 is simply broken - it is a sure way to
crash thew system by simple `rm -rf /some/dir' when /some/dir has
many files (i.e., /usr/src/sys directory mounted from NFS server).
This is infamous readdirplus() bug. Same on 2.2-GAMMA, or 2.2-STABLE
past 2.2.2-RELEASE (which I run now).
NFS v2 mounts do not show these problems (crashes / zero writes).
So, please try `mount -t nfs -o nonfsv3 ...' and see whether your problem
goes away.
My NFS server is NetApp F330, running ONTAP 4.0-something. (this is
not a Unix box). It works absolutely flawlessly with Solaris clients,
which also use v3 mounts as the default.
I think that changing default mount from v2 to v3 was a bad idea.
> Both of these boxes are on the same network segment as the directory
> they're writing to. They were invoked as follows:
> tester >> dummy
> and
> tester.sol >> dummy
> Another symptom of the problem is that it's not possible to get any
> output from
> tail -f dummy
> on the same box that's running on tester program under FreeBSD.
> However, solaris has no problems.
again, this looks like file append problem. Have you tried
writing numbers separated by newlines? Does this make any difference??
--
"The only problem with troubleshooting is that
sometimes trouble shoots back." --joez@earthlink.net
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199707192106.OAA21295>