mail woes (fwd)

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

mail woes (fwd)

Craig Carey-2
This will explain many of our problems on election-methods-list...
---------- Forwarded message ----------
Date: Fri, 12 Apr 1996 05:07:36 -0700 (PDT)
From: Joel Boring <[hidden email]>
To: Multiple recipients of list <[hidden email]>
Subject: mail woes

We are currently having _huge_ problems with out mail machine, it is working on
a 10000 message mailqueue.  This all started when a program "ypbind" decided it
was going to die on the mail machine.  Because of this, mail was unable to
resolve any hostnames in mail messages so it queued them all.  We found the
problem and fixed it, but the machine is still backlogged.  We have moved the
old messages to a different queue so they shouldn't be dragging mail to its
knees anymore...but some mail sent in the last 3 or 4 days (mostly list mail
because of the greater number or recipients) is still being processed.

About 2000 messages were processed from the old queue tonight, out of 7000
(there were 10000 but it managed to deliver some of them before we moved the
queue).  I just took a look at the queue, and here are the results:

/<1>spool/mqueue.old (ttyp6) dwild@mail# ls -1 q* | wc -l
/<1>spool/mqueue.old (ttyp6) dwild@mail# grep -ci "list" q* | grep -v ":0" | wc -l

in english:  there are 5059 messages yet to be processed, 4791 of these are
either going to a list or coming from a list.  This is why your lists have been
stalled for the past few days, I apologize for not finding this sooner.

By moving the queue, we have made a tradeoff.  New mail is being delivered in a
more timely manner, but the mail already in the queue is actually being
processed slower.  We hope to be able to send all these messages out by this
Sunday, I assure you we are making this process go as fast as possible, but we
do need to be careful not to load the machine so much that new mail is queued
rather than processed.

There has been another problem stemming from this one, when sendmail passes
messages off to a program, like "catmail" (listproc) or "wrapper" (majordomo),
if the process does not have enough file descriptors available the command
fails and the message is bounced.  This is a sendmail bug (the fact that it
bounces the message instead of queueing it and trying again), we are in contact
with the program maintainers to try to get this fixed.  Until we can get
sendmail fixed, the only thing that can be done about this is to re-send the
message at a later time, usually it will go through.  List owners, you may want
to notify your users of the problems we are having so they will not be as
confused when they get a message back saying something about "Mailer error 14"
or "Mailer error 7", tell them if they re-send their message it should go
through.  The problem with running out of file descriptors should diminish when
the load on mail goes down (after this weekend), and with it, the catmail and
wrapper messages.

Thank you all for your patience through all this, it's very frustrating not
only for you, but for me because I'm in the process of making listproc work for
everyone, then things like this happen =/

    Joel Boring aka Derek Wildstar <[hidden email]>
This message is not to be sent over the Microsoft Network.