> Is it a "bad idea" to let more than one server append mail to a shared
> NFS-mounted /var/spool/mail/user mail spool directory.?
I believe that when possible it is preferable to avoid it, as it gives a
"warm glow".
My understanding is that the belt&braces technique that exim uses (if so
configured) are safe across NFS (i.e. lock file hitching posts etc).
If anyone has any info that indicates that NFS access is *not* safe, please
let me know, as we rely on it heavily !
> I want to introduce the new exim-based server gradually, so if I could let
> it deliver "local" mail to the NFS mount, as well as let the creaking old
> mail server do it too (on its local drive), my transition could be more
> gradual.
AHH -- there *MAY* be a problem there !
The critical things is that *ALL* programmes which lock the file do so in the
*SAME* way(s).
> I gather that NFS is a bit weak in Linux,
NFS locking is -- yes.
> it's too late to change the Linux basis of the LAN.
We have lots of Linux clients locking NFS served mailboxes to "inc" the mail.
> The mailboxes in /var/spool/mail are accessed by users on Pine over IMAP
> during the course of things too.
How does the IMAP server and the old MTA lock ?
> How can be sure that dot-locking (which I presume is what I need) or whatever
> is a) working and b) being used by everything?
What I do is to RTFM, check the configuration (i.e. that the required locking
is compiled in and enabled), and then use a programme which "claims the lock"
and try all the operations.
e.g. claim the lock, then try IMAP reading, old MTA delivery and new NTA
delivery -- if any work (i.e. do not "fail" or "block", you know for sure that
you have a problem !
--
*** Exim information can be found at
http://www.exim.org/ ***
