On Wed, Jun 25, 2014 at 03:30:26PM -0700, Julian Mehnle wrote:
> > Yes, this is two deliveries, for two recipient records in the queue
> > file, that are in fact the same recipient address? Why does this
> > message have to identical envelope recipients? The pipe(8) pid
> > would be the same for a single delivery with two recipients, as
> > would "delays".
>
> The original RCPT TO addresses are actually different, but they
> resolve to the same destination internally. If there was a way to
> make Postfix collapse them into a single delivery, that would solve
> my problem as well.
And the reason you don't detail how this is done is to annoy me,
and delay the solution to your problem, right? :-) Fortunately,
the picture becomes more clear below...
> > What is the address class of the recipient domain (etl.qa.agari.com)?
> > What configuration settings result in the use of the pipe transport?
>
> Address class? I assume you're referring to this, but I haven't configured
> any classes explicitly.
>
> We have this in master.cf:
>
> > dmarc-aggregate unix - n n - - pipe
> > flags=RXq user=automation argv=/usr/bin/ruby1.9.1
> > /usr/local/agari/hawking/qa/current/bin/dmarc_receive -e production -p
> > /data/pipeline/queue automation
>
> and in main.cf:
>
> > mailbox_transport_maps = hash:/etc/postfix/mailbox
Well, "mailbox_transport_maps" is a recipient-dependent version of
mailbox_transport, which is in-fact unavoidably one recipient at
a time.
You'd need to use virtual(5) aliases and transport(5), not local
aliases(5) and mailbox_transport to multiple recipients in one
delivery.
--
Viktor.
