On Wed, Feb 04, 2004 at 01:24:01PM -0800, Bart Schaefer wrote:
On Wed, 4 Feb 2004, Dallman Ross agreed:
Here's some timing tests on one hundred spam messages. [...]
Okay, [procmail] took just over a minute-and-a-half, and seems to have
sent the load up a bit briefly on this machine. [...]
Christ. I interrupted [spamc] after *twenty-four minutes*!!! I thought
it must be stuck. But no, we were nearly 90% through: [...]
The load was between 110 and 128 the whole time! A couple of
minutes after I bailed on the job, it is back to 6 or 7 for the time
being. Hmm.
This indicates something seriously wrong with Panix's spamd setup. My
first guess would be that they're not using the "-m" option to limit the
number of forked copies of spamd.
You may well be right. I've asked the admins.
Even so, however, I can't explain how you managed to run the load up over
100 -- with a formail loop, there should still be at most one spamc plus
one forked-off spamd at a time (not considering what other users might be
doing).
I erred in leading us on a wild goose chase with that part. I thought
the spamd stuff ran on the supplemental host that doesn't take logins
from ordinary users. However, I've since been informed that it
only will do that in rare cases. So presumably it was running on the
host where I was. The load wasn't going up. I don't know *what*
was going on, but I agree that it seems very fishy. I mean, we
do have a few thousand shell users, *many* of whom are running
spamc/spamd via procmail or their .forwards, and I can't imagine
that a measly hundred messages would change the schema *that* much.
I will send these comments to the admin folks.
I've pumped mailboxes containing several *thousand* messages through
"formail -s spamc" in much less than 24 minutes and without killing our
500MHz PIII linux server.
--
dman
_______________________________________________
procmail mailing list
procmail(_at_)lists(_dot_)RWTH-Aachen(_dot_)DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/procmail