Thanks to Ben Goodwin, Steve VanDevender, and Tim McCarthy for responses on
this. The consensus was that this is a locking problem. I tried several
things, and finally moved the mail spool to a local raid filesystem on my
DEC mail server. This eliminated the stale nfs file handle errors (of
course).
However, I am still seeing a problem. Many times each day, binmail reports
via syslog that it is waiting for a lock. A few times each day, it reports
that it is overriding a lock. It appears to be dropping mail at least some
of the times that it reports waiting for a lock but not overriding it. It
looks like binmail is quietly giving up and not delivering the mail.
Any ideas why?
Thanks,
David Ritch
-----Original Message-----
From: David B. Ritch <dbritch_at_earthlink.net>
To: Alpha Managers List <alpha-osf-managers_at_ornl.gov>
Date: Thursday, February 19, 1998 9:48 PM
Subject: Strange nfs problem when delivering mail
>My site is using Berkeley Sendmail version 8.8.7 on our mail exchanger,
>using binmail for local delivery. Our mail spool is nfs mounted on the
>server from a Network Appliance, which has been extremely reliable. We are
>seeing an occasional problem that has me baffled. A few times a day,
>binmail is reporting a failure while fclosing a file. It claims to be
>getting nfs error 70 - stale nfs file handle.
>
>The file handle is *not* stale. The filesystem continues to be mounted
>throughout the day. At the suggestion of a Network Appliance engineer, we
>switched from nfs 3.0 to 2.0, and are mounting the filesystem through
>/etc/fstab instead of automount. However, the problem continues.
>
>Does anyone have a suggestion of what might be causing this?
>
>The mail exchange host is a DEC 4000 running DU 4.0B, and the NetApp is an
>F540 running DataOntap 4.3 (and had the same problems under 4.2).
>
>Thanks in advance for any help!
>
>David Ritch
>
Received on Fri Mar 06 1998 - 07:04:47 NZDT