Strange behavior between syslogd and sendmail

[91erm at bigbird.cc.williams.edu]

I just saw a strange begavior on our workstations.

Yesterday I switched our 3/50s (Running 4.0, no less) from the old
"normal" sendmail (patched against the November worm) to use the mx
version of sendmail (also patched against the worm).  Everything seemed
okay and I didn't see any reason to think twice.

Today a user reported that none of his mail to our VMS VAX was being
delivered.  This struck me as odd so I looked into it.  My standard check
in the beginning of these things is to run a mail -v and watch the SMTP
conversation.  So I let it go, and sendmail told me that it couldn't find
the host.  In fact, due to a few omissions in our named files, it couldn't
find any of the hosts it wanted, and therefore couldn't even get a message
out to postmaster.

I won't go into the details about why the host lookups failed and why the
postmaster alias wasn't quite robust enough.  I think I've got that fixed.
The wierd thing is this: there are *NO* entries in log files of any sort,
kind or shape.  I have a line like the following the /etc/syslog.conf of
all our machines:

mail.debug		/var/log/maillog

and there's no trace of these events in the log.  Personally speaking,
stats about completed messages are great, but if the logs don't record
when sendmail cannot deliver a message (to anyone, postmaster especially),
what good are they ??

Can someone with source or close ties to Sun explain this behavior ??  Can
you also explain why, when a machine fails to appear in the nameserver,
sendmail.mx doesn't consult the YP service for its address ??  If this
worked this problem never would have occured in the first place.

Thanks for the info.

Evan R. Moore
Academic Computing Group
Williams College
91erm at bigbird.cc.williams.edu
91erm at cc.williams.edu
91erm at cs.williams.edu
91erm at williams.bitnet
Now you decide ...