CVS: mailman/Mailman ListAdmin.py,1.33,1.34
It works great! Two cents=
in order to still have a responsive server, you need qrunner to be activated every minute or every other minute, not just every half hour like it is now (usually).
nothing is logged into /logs/* when a message is approved. Very stressfull for the mailman-master trying to watch if it works. S/he has to think to look at qfiles/ to see if the message was correctly processed.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Date: Tue, 30 May 2000 22:36:43 -0700 From: Barry Warsaw <bwarsaw@slayer.i.sourceforge.net> To: mailman-checkins@python.org Subject: [Mailman-checkins] CVS: mailman/Mailman ListAdmin.py,1.33,1.34
Update of /cvsroot/mailman/mailman/Mailman In directory slayer.i.sourceforge.net:/tmp/cvs-serv13283
Modified Files: ListAdmin.py Log Message: __handlepost(): Instead of trying to deliver approved messages immediately with DeliverToList(), enqueue the message for later delivery by qrunner. This vastly improves the responsiveness of the web interface for huge lists.
The actual current cron/crontab.in says the following:
# Every 10 minutes retry failed deliveries. 02,12,22,32,42,52 * * * * /usr/bin/python /var/lib/mailman/cron/qrunner
It should be
# Every minute try pending deliveries.
- /usr/bin/python /var/lib/mailman/cron/qrunner
Fil (fil@orwell.bok.net) écrivait :
It works great! Two cents=
in order to still have a responsive server, you need qrunner to be activated every minute or every other minute, not just every half hour like it is now (usually).
nothing is logged into /logs/* when a message is approved. Very stressfull for the mailman-master trying to watch if it works. S/he has to think to look at qfiles/ to see if the message was correctly processed.
On Wed, May 31, 2000 at 11:40:24AM +0200, Fil wrote:
It works great! Two cents= __handlepost(): Instead of trying to deliver approved messages immediately with DeliverToList(), enqueue the message for later delivery by qrunner. This vastly improves the responsiveness of the web interface for huge lists.
wow this is my dream come true... now i can stop bugging Barry about implementing this ... ;) i can't wait to get home and install the new cvs... Our list relies heavily on approving messages so I'll report my experiences with it asap...
Ricardo.
"RK" == Ricardo Kustner <ricardo@rixhq.nu> writes:
RK> wow this is my dream come true... now i can stop bugging Barry
RK> about implementing this ... ;) i can't wait to get home and
RK> install the new cvs... Our list relies heavily on approving
RK> messages so I'll report my experiences with it asap...Excellent, please do. I'm on a bent to get beta3 out (and after that, clear my freakin' mailman-dev inbox ;).
Btw, it was a one line change :)
-Barry
On Wed, May 31, 2000 at 12:45:02PM -0400, Barry A. Warsaw wrote:
"RK" == Ricardo Kustner <ricardo@rixhq.nu> writes: RK> wow this is my dream come true... now i can stop bugging Barry RK> about implementing this ... ;) i can't wait to get home and RK> install the new cvs... Our list relies heavily on approving RK> messages so I'll report my experiences with it asap... Excellent, please do. I'm on a bent to get beta3 out (and after that, clear my freakin' mailman-dev inbox ;).
I've installed the patch and approved about 15 posts (had to discard a lot of crap ;() and qrunner is doing the postings now... it seems to work ok so far... but I do notice that when I go to the admin URL while qrunner is doing it's job, it takes more than a minute before the page is displayed... does this have something to do with locking or it's because my server has a load average of 3+ (which isn't very uncommon considering the hardware performance though)
Btw, it was a one line change :)
cool :)
Ricardo.
-- International Janet Jackson fanclub called MISS JANET. For more information write to: Miss Janet. P.O.Box 10016, 1001 EA Amsterdam, The Netherlands Fax/phone: +31-(0)20-7764493 Email: fanclub@miss-janet.com Or check out our website: http://miss-janet.com
"RK" == Ricardo Kustner <ricardo@rixhq.nu> writes:
RK> I've installed the patch and approved about 15 posts (had to
RK> discard a lot of crap ;() and qrunner is doing the postings
RK> now... it seems to work ok so far... but I do notice that
RK> when I go to the admin URL while qrunner is doing it's job, it
RK> takes more than a minute before the page is displayed... does
RK> this have something to do with locking or it's because my
RK> server has a load average of 3+ (which isn't very uncommon
RK> considering the hardware performance though)Maybe a bit of both. qrunner is careful to keep the lock only long enough to deliver a single message. If there are multiple messages destined for the same list, it will release and acquire the lock in between each delivery. Still, for large lists, this can take a while.
One of the things to eventually do is audit the cgi scripts so that they're locking the list only during those operations that require it. I.e. a listinfo probably doesn't need the lock, and setting user options probably only needs the lock for a very short amount of time.
Won't all this just work better with a real transaction based db underneath? ;)
-Barry
participants (3)
-
bwarsaw＠python.org -
Fil -
Ricardo Kustner