I have resent this message to just Wietse... but I wanted to send it
to whole mailing list. So I do.
When you are running the policy service under the spawn daemon:
1) The time limit BEGINS after an SMTP server connects to the
policy server port, when a new policy daemon process is created.
Thus, an SMTP server always "mates" with a policy server that
has the full time limit left.
First of all perhaps should explain the working mode of Postfix
quota reject :
I'm going to explain this because perhaps it's a thing I should
change but Postfix Quota Reject behavior is basically : start, wait
to attributes, proccess and answer to postfix the action, later
clean the policy attributes hash (in perl) and wait attributes again
and so... it doesn't die never unless max_use or max_idle time limit
is exceeded, or exceds the policy_time_limit; I have done this for
being a bit faster because the created policy server is running and
if some of them is free you don't need to reswpan one and run it
from the beggining (storing variables in memory and so...). I mean
my policy doesn't die never by it's own... it preserves running
waiting someone to connect to it send attributes and the pqr
(postfix quota reject) to give an answer.
So knowing this, can never happen that a smtpd proccess starts a
policy server (connects to and reswpans a new one policy server,
stating here the begin time) later the smtpd proccess perhaps is
killed by spawn because max_idle or max_use and the policy server
created to be available still to be used by another smtpd
proccess?? and now let's say other different smtpd proccess
connects to the existing policy server created by the previous smtpd
proccess (now died by spawn for exceeding max_use or max_idle) but
when it (this different smtpd proccess) connects to this now
available policy server has the resting time_limit because it has
connected to the policy server "created" by the previous smtpd
proccess (starting begin time when previous smtpd proccess created
it) that now doesn't exist???; Should I do the policy server to die
after giving action? instead of just cleaning the hash table and
start the proccess again?.
2) The time limit ENDS after an SMTP server process disconnects,
because the policy daemon terminates immediately.
so should I make my policy to exit after giving action?
If you still think the time limit is too short, make it longer.
Wietse
Thanks a lot Wietse.
Bye!!
