[Mailman-Users] list too big?

[Dave Stern - Former Rocket Scientist]

On 11 Feb 2004, Jon Carnes wrote:

> On Wed, 2004-02-11 at 07:26, Dave Stern - Former Rocket Scientist wrote:
> > We have a tiny sun ultra 10 handling 100-200 lists with general success.
> > It's running an older version of MM (2.0.13) as I haven't gotten around
> > to upgrading and don't see many new features we'd want.  Most of our lists
> > have 3-400 users and work fine. I just recently added one with over 11,000
> > addresses. Coincidentally, or not, the machine periodically hangs when this
> > list tries to send out. This is despite having added some failsafes in
> > sendmail eg load average refuse/delay queuing, connect and bad rcpt
> > throttling etc.
> >
> > One thing I've see in the qrunner log is:
> >   Could not acquire qrunner lock
> >
> > Suggestions? Has anyone had any experience with really large lists on
> > really small machines?
> >
> My guess - RAM.  Add more Ram - lots more RAM.
>
> Mailman loads the entire list into memory when processing it. Your
> larger list requires a larger hunk of memory. Look at the servers
> utilization when trying to send to that list.  You should be able to
> pinpoint the limiting factor.
>
> What sort of disk subsystem do you have?  SCSI or IDE?  If your moving a
> lot of traffic or have really large lists, SCSI is a must.
>
> We've discussed large lists a lot in the archives.  If you can not add
> more RAM then I suggest you break the larger list down into several
> smaller list.
>
> Good Luck -
> Jon Carnes
>

Thanks for your help. The memory on this beast is maxed out at a screaming
512Meg. It's an IDE structure but we just put in an emergency PR for a 15Krpm
scsi disk for /var/spool/mqueue and found a source for cheap scsi cards.
I've also made it a cache-only nameserver.

If these things don't resolve it, I'm gonna have to break it up into perhaps
26 sublists (A-Z) and make the main list an umbrella list.

Ultimately, we're gonna need a faster machine but layer 8 of the OSI model
(political) works slowly here ;)



 =-=-=-=-=-=-=-=-=-=-=-=-  generated by /dev/dave -=-=-=-=-=-=-=-=-=-=-=-=-=-=
 David Stern                                            University of Maryland
                Institute for Advanced Computer Studies