[comp.sys.apollo] several sendmails using the same /usr/spool/mail?

has@ztivax.UUCP (Hans-Albert Schneider) (08/27/90)

Configuration: we are one of several APOLLO rings connected via
ethernet, and administered centrally. One machine of the ring is at
SR10.2, the others are at 9.7.3.
sendmail runs at the master machines (one for SR9, one for SR10) in
the computing centre as well as on our gateway (an other gateways; but
I don't know whether their daemons get mail at all). As far as I know,
we are the only ring having a /usr/spool/mail of our own (on the
gateway), forwarding mail for users outside our ring to the master
machine.

I remember that SR9 sendmail has some problems with two or more
sendmails using the same /usr/spool/mail directory (i.e.,
//machine2/bsd4.2/usr/spool/mail is a link to
//machine1/bsd4.2/usr/spool/mail).

I don't remember exactly what goes wrong, but I seem to remember that
mail may be *returned* with a "cannot lock file" message instead of
retrying delivery later. There may also be a problem with the
lock() [maybe flock()] call when the file is on another machine.

However, changing our machines (with one exception) to SR10.2, the
consideration of having ONE /usr/spool/mail for all machines in ALL
rings (as the computing centre would prefer) comes up again, so are
there still any reasons to have separate mail directories at SR10?
One of those mentioned above? Problems when the machine holding the
mail directory is down? Others?

Please email to one of my addresses given below, as I currently do not
read news very regularly. If there is interest, I'll summarize.
-- 
Hans-Albert Schneider			ARPA: has@ztivax.siemens.com
Siemens AG, ZFE IS INF 21		or:   has%ztivax@siemens.siemens.com
Otto-Hahn-Ring 6, Munich, W.Germany	EUnet & UUCP:
phone: (+49) 89 636 45 445			has@ztivax.UUCP