[Mailman-Users] Deleting old msgs?
Richard Haas
rhaas at rhaas.us
Sat Dec 3 17:50:41 CET 2011
Mark Sapiro <mark <at> msapiro.net> writes:
>
> Frank Bell wrote:
>
> >Is there any easier way? I just took over our mailman server and we have
> >several years worth of messages in 190+ mboxs totaling approx 130 gig
> >and a few 100K msgs.
>
> You inspired me. I've created a script for pruning archives. See "NOTE
> ON PRUNING OLD MESSAGES:" in the FAQ at <http://wiki.list.org/x/2YA9>
> for links.
>
> Since this is a brand new process, I suggest you make backup copies of
> the LISTNAME.mbox files before starting. The script has a --backup
> option, but I would make separate backups to be sure until you've run
> the script successfully.
>
:-) Timing is everything ... just finished integrating mbox-purge.pl
(http://www.argon.org/~roderick/mbox-purge.html) with a withlist
callable module to do the same thing.
Having fewer layers would be welcome though, so thanks for this script,
Mark.
One idea/request: Would you be willing to add the logic to write
the pruned message data to a supplied path+filename? That would let
the script dump the pruned data where it could be retained or aged via
another scheme.
Our site (and maybe this is more common) periodically prunes the
archived .mbox messages when they are a year old, rebuilding the
pipermail hierarchy, but keeps a compressed copy of the pruned data
for another year (or longer), in case it is needed. The compressed
pruned .mbox text is considerably smaller (like 1/20th or better on
average) when compared to the uncompressed .mbox plus the
associated pipermail HTML hierarchy -- so keeping a copy is a
relatively trivial insurance policy or "nice to have" for our lists.
We've found that pruning is essential once archives become
multi-gigabyte, not for the .mbox archives themselves, but due to
the pipermail HTML files that result (particularly for archives with
many small messages). We've seen 3 GB .mbox archives with
approaching 1 million files in the pipermail hierarchy. Traversing
that many files or rebuilding them, particularly for hundreds
of such lists, is non-trivial even on modern hardware and file
systems.
Thanks in advance for considering adding a way to save the pruned
data.
Richard
More information about the Mailman-Users
mailing list