> First, I want to thank the people who helped me compile and understand
> bulk_mailer and its use when combined with Majordomo.
>
> I thought the following would be of interest, especially to SAs just
> setting up lists or for SAs with lists growing large (5000+). The
> following message is intended to be a warning to keep an eye on system
> resources and list size.
>
> --
>
> Hardware:
>
> Sun SPARC 20, 64 Mbs RAM, 128 Mbs swap, 2 2.1 GB Drives
>
> Software:
>
> Solaris 2.4 (with all the patches), sendmail 8.8.5, majordomo 1.94.3,
> bulk_mailer 1.5, procmail 3.10, Apache 1.2
>
> --
>
> Among other "Mars Project" lists that I own, the two lists that see the
> most traffic are the lists that are open to the public: "mpf-status" (Mars
> Pathfinder) and "mgs-status" (Mars Global Surveyor).
>
> Currently, mpf-status has close to 12000 subscribers, while mgs-status has
> close to 8000. As you may guess, keeping these lists relatively clean
> (free of dead accoounts and such) is becoming a full time job. Both of
> these lists are set to "restrict-post" so that only a few relavent people
> are able to send to the list. The list is meant to be "read-only" ...
>
Wow :) I'm running bulk_mailer, but the highest number of subscribers I have
is around 200. Wish I had so many :) !
> These two status lists were working fine, until we reached about 3000+
> subscribers - then mail started to crawl, and sendmail would eat up a bunch
> of swap space (majordomo did not care, since it was just handing things off
> to sendmail). Anwyays, I kept looking for a way to increase mail speed.
>
> Well as many of you know, Mars Pathfinder successfully landed on the
> surface of Mars ... astonishing if I do say so myself :) . Back to the
> point ... last week (June 27) mpf-status contained 6587 members. That
> number jumped to over 10000 in just a few days. I was on shift (8PM to 8AM
> Thursday July 3) when the list machine came to a screeching halt. All of
> the swap space had been allocated, and I could not even telnet in! I was
> able to login through the console, and through the "hundreds" of sendmail
> errors (swap space gone, etc.) I was able to issue a "sendmail stop"
> command and then a warm boot (reboot).
>
> It appears as though sendmail was trying to deliver all the mail
> immediately, which not only included outbound mail to the list, but
> outbound mail to the hundreds of people subscribing to it! (all with
> carrying out regular email demands of several other Project personnel) I
> needed to find out a way to work this problem through ... I was a bit
> nervous.
>
> With the help of a few people on this list (majordomo-users) I was able to
> set up bulk_mailer and integrate it with my majordomo.aliases file. So far
> so good ... mail seems to be moving MUCH faster than ever before. I have
> changed the sendmail flags in the majordomo.cf file to queue up mail
> (-odq), and increased the "-maxdomains" flag in bulk_mailer to 40. Mail is
> flying. Very cool.
>
> So the point is, mail to mailing lists in excess of 10000 members can be
> done (even on a very busy machine) ... oh yeah, I did not mention that the
> machine is also the main website for MGS (which yesterday took 467000
> hits!).
>
> Kelly
>
> ps: Sorry, did not mean to be so chatty!
>
>
> ============================================================================
> Kelly Feagans Jet Propulsion Laboratory phone (818) 393-0501
> Network Engineer 4800 Oak Grove Dr MS 264-214 fax (818) 393-5247
> Mars Global Surveyor Pasadena, CA 91109 kfeagans@mgsw3.jpl.nasa.gov
> ============================================================================
>
>
|
|
