The mailing list for listmasters using Sympa

[Patrick von der Hagen <address@concealed>]

Am Montag, den 20.11.2006, 13:22 +0100 schrieb Olivier Salaün - CRU:
> Hi Patrick,
Hi Olivier,

> Patrick von der Hagen wrote:
> > looks like sympa-daemon died when my harddisk was full Friday evening.
> > Actually, that's strange since sympa has its own partition which was not 
> > affected.
> >   
> Sympa might have died because of a message it failed to process. If so, 
> sympa.pl might die again.
So far sympa.pl is running fine, however I moved most of the messages
out of my spool as a form of priorization. After the important lists
have been processed, I'll move them back in smaller batches.

[...]
> The queue processing deals with messages priorities (defined on a per 
> list basis). This is very time consuming because after each message has 
> been processed, sympa.pl scans the spool again to find out about new files.
Hm. Does sympa start from scratch on every scan, so it processes every
file again and finally select just one message, or does it keep a list
of files in memory which is only updated during each scan?

Are there any other criteria involved, besides priority? E.g. "messages
in spool with highest priority are processed first, ordered by arrival"
or something like that?

> The process could be optimized if sympa.pl scans its spool once only. 
> But currently you'd have to patch sympa.pl.
So I could try to do something like "select the most urgent 10% of
spool-messages and process all of them before doing the next check of
the incoming-spool" or I might create an additional spool for each
priority, moving messages from the incoming spool to the priority spools
once, so sympa has to do less continuous calculations.
Some loop like 
"
frequency = take_from_sympa.conf
while true do
    distribute incoming-spool to priority-spools
    for up to frequency messages do
        process message which has highest priority
    done
done"

Sheduling is very tricky business.... :-(

-- 
CU,
   Patrick.