kutz@bgsuvax.UUCP (Kenneth J. Kutz) (01/04/90)
Configuration: Vax 11/785 running MORE/bsd (based on 4.3BSD) using
sendmail 5.61 ftp'd from Berkeley
Just recently we have started administering a large mailing list. The
machine which receives the mail message and then consequently disperses
it to the rest of the list (kept in a file with a corresponding
':include filename' entry in /usr/lib/aliases) has the above configuration.
The problem we are experiencing is that sometimes sendmail doesn't
"remember" which hosts it has sent the message to and sends some
messages more than once to the same user at the same host. Currently
there are 120 people on the mailing list. I assume that when a
message gets "deferred" or fails to be sent because of some interruption
in service, it is then that we are ripe for multiple copies to
hit the same recipient. If that message fails for the same reason
on future tries, previous successful sends get sent again succesfully
which irritates those folks who only like to read a message once :-)
Is anyone else maintaining large mail lists with sendmail and experienced
these same growing pains? I suppose the problem could be from the
original sender site and not here given the problem is rather sporadic.
It has happened 3 times in a month and a half.
Thank you.
--
Kenneth J. Kutz Internet kutz@andy.bgsu.edu
Systems Programmer BITNET KUTZ@BGSUOPIE
University Computer Services UUCP ...!osu-cis!bgsuvax!kutz
Bowling Green State Univ. US Mail 238 Math Science, BG OH 43403cfe+@andrew.cmu.edu (Craig F. Everhart) (01/04/90)
I wish I had isolated the specific thing that I did to address the problem. andrew.cmu.edu houses several 100+ recipient mailing lists. The problem in sendmail was that it would be called on to deliver mail to a whole long list; some deliveries would be successful, some not, but it wouldn't finish processing the entire list before it crashed or the machine was rebooted. Meanwhile, the qf* file hadn't been updated to remove the destination addresses for those deliveries that had been successful. What I did was to insist that sendmail update the qf* file after every successful delivery, and sync() the disk. There were only a couple of changes involved. Is there a substantial call for diffs (which might not be that useful since our local sendmail had diverged long ago)? Or is this diagnosis/treatment report sufficient? Or has the problem been solved in later versions of sendmail? Craig Everhart
paul@speedmetal.engin.umich.edu (Paul Killey) (01/05/90)
Is there still a setjmp/longjmp botch in the distributed sendmail? This is when it showed up ... on large mailing lists. There have been fixes from a couple sites ... mit and U-maryland maybe? Something about a 2 hour timer going off during sfgets() and trying to jump into some context that had been freed by another shorter timer in smtpinit()? I can't remember now. If you think that might be the problem I can send you some diffs for that, or maybe can find the older original message that discussed it. Just a thought. --paul "Don't call me baby when she's waiting in the car."