[comp.mail.sendmail] "Locked" error messages in syslog ?

pj@isatis.ensmp.fr (Pierre JOUVELOT <jouvelot@ensmp.fr>) (04/10/91)

Hi,

I'm getting a lot (I mean *a lot*) of messages like:

Apr  5 10:24:48 linz sendmail[13333]: AA02370: locked
Apr  5 10:24:48 linz sendmail[13333]: AA03647: locked
Apr  5 10:24:48 linz sendmail[13333]: AA04590: locked
Apr  5 10:24:48 linz sendmail[13333]: AA05227: locked
Apr  5 10:24:48 linz sendmail[13333]: AA05259: locked
Apr  5 10:24:48 linz sendmail[13333]: AA05263: locked
Apr  5 10:24:48 linz sendmail[13333]: AA05274: locked
Apr  5 10:24:49 linz sendmail[13333]: AA05283: locked
Apr  5 10:24:49 linz sendmail[13333]: AA06411: locked
Apr  5 10:24:49 linz sendmail[13333]: AA06911: locked

in the syslog file of my mail server (linz is a workstation connected on
our local ethernet to the mail server). The mail files are, however,
correctly sent and received; I'm only concerned about this flood of
error (warning) messages.

It seems to be somewhat related to the use of GNU Emacs since mail sent
from linz, but outside of the GNU mail mode, seems to behave properly.

Any idea ?

Thanks,

Pierre

vixie@decwrl.dec.com (Paul A Vixie) (04/11/91)

[Pierre JOUVELOT]
>> Apr  5 10:24:48 linz sendmail[13333]: AA02370: locked
>> Apr  5 10:24:48 linz sendmail[13333]: AA03647: locked
>> Apr  5 10:24:48 linz sendmail[13333]: AA04590: locked

These come up when a sendmail "queue run" finds a locked envelope.
(If you are running 5.65, this means that a qf* file was flock()'d;
if you are running 5.61 or below, this means that an lf* file was
found.)  Basically it means that a delivery attempt is in progress
and the "queue run" (think "/usr/lib/sendmail -q30m" or "-q -v") is
skipping over it.

Since 5.61 (and earlier) used "lock files", it was possible for a delivery
attempt to core dump and leave the envelope locked.  This is why
/etc/rc.local removes all the /usr/spool/mqueue/lf* files on
Berkeley-derived systems.  If you see a lot of "locked" messages for the
same QID ("AA02370", for example), and you are running pre-5.65, you might
look and see if you've got a stuck lock.

If you don't have any stuck locks, it means that deliveries are taking
longer than your queue interval ("-q30m" means 30 minutes).  This is very
unusual except on very large mailing lists, and probably means that your
name server is timing out a lot or that something else is slowing your
mail down.  Or perhaps you've got a short queue interval, like five minutes?

Lastly, you might consider changing your "Od" line from "Odb" or "Odi" to
"Odq".  This slows down your mail, since it means that only "queue runs"
can actually attempt delivery.  But it keeps you from trying to do delivery
on each piece of mail as it is submitted, thus serializing the delivery
process.

All of this matters only because you say you're seeing it on a workstation.
On a mail gateway, you tend to run multiple "queue run" processes, which
results in a lot of "locked" messages in syslog since the queue runners
will lock eachother out as they leapfrog down the queue.  On DECWRL, we
have 20 queue runners per queue, and eight queues; we changed the sendmail
source code to only log "locked" messages if OL was set very high -- other-
wise we would be overrun by megabyte after megabyte of "locked" messages.

Cheers,
--
Paul Vixie
DEC Western Research Lab	<vixie@pa.dec.com>	<paul@vixie.sf.ca.us>
Palo Alto, California, USA	...!decwrl!vixie	...!vixie!paul