Ralf Hildebrandt:
> When sending lots of mails (mass mailings) via many machines, one
> quickly realizes that the current concept of smtp_fallback_relay is
> a bit problematic:
>
> * The mass mailing gets (randomly) distributed to n machines
> * these n machines try sending out the mails assigned to them and will
> encounter mechanisms like postscreen or greylisting mechnisms
> * thus the initial delivery fails
> * the mail gets passed to the smtp_fallback_relay
>
> I wonder if a time based smtp_fallback_relay ("pass the mail to the
> smtp_fallback_relay after a certain period of time has expired") or
> using the smtp_fallback_relay after a certain number of delivery
> attempts (>=1) would be a worthwile addition to Postfix.
I suppose you have lots of logfile information. Does the evidence
support the idea that delaying smtp_fallback_relay would improve
the delivery life cycle?
FYI, Postfix does not count the number of delivery attempts. That
would require one counter field per recipient in the queue file.
It also does not record the time of the last delivery attempt. That
would require one time stamp field per recipient in the queue file.
All that currently exists is the message arrival time (in the file),
and the "next" time to deliver the message (outside the queue file
so that Postfix can quickly skip over files in a large queue).
Before making invasive changes to the scheduling algorithm, there
should be data that supports the change.
Wietse
