Sporadical NFS errors with NetApp backend

[Lars Köller]

--------

Hallo,

I'm a FreeBSD user for years with machines in production since V2.X.
At the moment I have a really strange problem in our environment.

  - NetApp Cluster NFS file server
  - HPUX-11, Solaris-8, FreeBSD-4.11 Clients

  - 60000 Virus/Spam scans a day, spread over 5 different FreeBSD Machines,
    two different hardware bases, 1GBit Network

A problem occurs with the Sophos PMX (PureMessage Anti-Spam/Anti-Virus)
Application. 1-3 times a day the following happens:

  - pmx_log:3 2005-06-23T12:26:58 [36968,milter] pmx_file(): 
    Can't append message info to '/usr/local/pmx4/var/qdir/tmp/1119522418-36968:135299072-ZN0IIE00EX1IKSAE.00-738-vscan1.hrz.uni-bielefeld.de': Bad file descriptor at /usr/local/pmx4/lib/site_perl/5.6.1/PureMessage/MessageFile.pm line 124.

  - /usr/local/pmx4/var is located on the NetApp Filer (NFSv3)

  - Line 124 referenced above looks like (close($f)):

sub make {
    my $class = shift;
    my $fname = shift;
    my $minfo = @_ == 1 ? shift : do {
        require PureMessage::MessageInfo;
        PureMessage::MessageInfo->new(@_);
    };
    open(my $f, ">>", $fname) || die "Can't append to '$fname': $!\n";
    print $f _minfo_block($minfo);
    close($f) || die "Can't append message info to '$fname': $!";
    1;
}

  - BUT: The message-info is absolutely correct appended!

This does not occur if I move /usr/local/pmx4/var on a local disk.

No errors in any logfile!

No problems with other Servers!

Any ideas? Is there a kernel NFS or interoperability problem with the 
NetApp Filer.

How can I debug this on FreeBSD?

Best regards

Lars

-- 
 Lars Köller     \   Raum   : V0-318 (Tel: 4964, FAX: 154964)
  Leitung Systeme \  E-Mail : Lars.Koeller at Uni-Bielefeld.DE
  u. Serverdienste \ PGP-Key: http://www.uk.pgp.net/pgpnet/wwwkeys.html
---------- FreeBSD, what else? ---- http://www.de.freebsd.org ---------