After switching from real-time to batch updates I found many fewer
problems with corrupt indices and bogus lock files. Since there aren't
any locks in contention, things work much better.
So, do you just dump messages to a spool directory in realtime and then run
mhonarc over that spool periodically?
Hmm, how do you archive over large periods of time?
I use procmail to copy the messages to a separate folder (constant name,
Then once an hour (via cron) I run a script
(http://www.armadillo.com/~zoo/contrib/newmail) which processes the mail
with MHonArc, and then saves the file:
Once a day I run another script
(http://www.armadillo.com/~zoo/contrib/daily) which cleans up and
combines all of these hourly message groups.
Finally, once a month the last script
(http://www.armadillo.com/~zoo/contrib/newmonth) creates the current
monthly archive (again, with a constant name of ~/www/LDsig/current/)
and life progresses automatically.
A top level page (http://www.armadillo.com/~zoo/LDsig/) remains
constant, since I haven't bothered to write a script to keep the date
current (the month and year buttons should be kept up to date, but I do
this whenever I remember to think about it, which is about every other
Things to keep in mind. Scripts like to deal with constant names, so I
make good use of symlinks. Second, date(1) has some powerful
capabilities to create a useful string. And running like this means
that I don't have to worry about locks.
david d zuhn | Montana Rail Link in N scale
zoo(_at_)armadillo(_dot_)com | http://www.armadillo.com/zoo/