faxrunq very broken on large queues

Gert Doering (gert@greenie.muc.de)
Fri, 27 Nov 1998 09:16:46 +0100


Hi,

(I took the liberty to copy mgetty@muc.de on the CC:, as I think that it
might be interesting for the archives or for other readers).

On Fri, Nov 27, 1998 at 08:39:53AM +0200, Graham Leggett wrote:
> I just tried a bulk fax run of around 400 faxes in the queue, and it
> failed spectacularly. During normal operation, I run faxrunq once a
> minute from cron. The main problem I experienced was that so many
> instances of faxrunq eventually caused the all of them to fail with file
> open errors, pipe open errors, etc. As a result of this, reporting on
> which faxes were sent and which faxes were not stopped working, as
> sendmail wouldn't run correctly.

Uh, faxrunq is not really meant to be run once per minute.  If you want
that response time, use faxrunq*d*.

> A suggested enhancement to fix this would be to change faxrunq so that
> there can be no more instances of faxrunq than there are faxmodems. We
> have 5 modems in a modembank, so in our case faxrunq should be
> executable 5 times in parallel only. Any further attempts at running
> faxrunq should cause the program to exit immediately.

*Especially* if you have multiple modems, you shouldn't use faxrunq,
because it's locking/jobs distribution scheme is pretty lame - it was
never meant for that, and doesn't handle it well.

Faxrunqd needs Perl5, runs all the time, and was written for a customer
that sends about 1000 faxes a day over 5 modem lines, all during business
hours (10:00-17:00).  It works for them very smoothly since day 1...

(Get a copy of 1.1.19, just for the latest version of faxrunqd.  I have
done lots of fine tuning on the scheduler and "kaput modem handler"
between 1.1.14 and 1.1.19.)

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@greenie.muc.de
fax: +49-89-35655025                        gert.doering@physik.tu-muenchen.de