[14891] in Athena Bugs
more on queued mail on suns
daemon@ATHENA.MIT.EDU (Jonathon Weiss)
Tue Dec 24 03:58:07 1996
From: Jonathon Weiss <jweiss@MIT.EDU>
To: bugs@MIT.EDU
Cc: jweiss@MIT.EDU
Date: Tue, 24 Dec 1996 03:58:00 EST
So I spent some time tonight trying to learn more about the problem
that causes mail to end up locked in the mail queues on suns. I found
several interesting things.
1) when running the q sendmail seems to lock a message by making a
hard link from the lockfile to the queue file, however in all cases
where I've seen messages locked on a workstation, the lockfile is a
seperate file not a hardlink. Thus if we are willing to assume that
it always takes sendmail less than 1 day to get a piece of mail from
the MUA and drop it into the q, the following command should detect
bogus lockfiles:
find /usr/spool/mqueue -name "lf*" -links 1 -mtime +0 -print
the "-print" may be replaced wiht a "-exec /bin/rm -f {} \;" if you so
desire. (Note that this hasn't been extensively tested.)
Of course getting a working sendmail is really the right answer.
2) I noticed that on a couple of suns in w20 there was mail that was
not locked, and was 1.5 to 2 weeks old sitting in the queue. running
the queue cleared the mail. On the one machine where I looked at this
more carefully, it appeared that none of root's (I didn't check other
users) cron jobs were being run, although cron appeared to be running.
When I added a job to touch /tmp/foo every minute, it did so. Weird,
and somewhat disturbing.
3) On an only vaguely related note, I noticed that it looks like there
are still some public suns that don't always syslog when they should.
Or at lease the logs don't seem to make it anywhere. This is also
disturbing.
Jonathon
