On Wed, 8 Aug 2007, Mike Brudenell wrote:
> 1. Send "Message 1" to an address on the test server.
> Delivery is attempted. Message gets deferred.
>
> Queue runs within the 5 minute maximum retry period attempt
> delivery
> and defer the message again.
>
> A queue run after the 5 minute period attempts delivery then
> constructs
> a delivery failure notification which is returned to the sender,
> and the
> incoming message removed from the queue.
...
> 2. Now send "Message 2" to the same recipient on the test server.
> Delivery is attempted. Exim then constructs a delivery failure
> notification
> which is returned to the sender, and the incoming message
> removed from the
> queue.
That is correct behaviour. Exim does not operate a "per message" retry
scheme. It operates a "per problem" retry scheme. (At least for this and
many other temporary errors.) Your 5 minutes starts from the time the
error is first discovered; after 5 minutes, there was be no retrying
until a delivery succeeds (which resets the clock).
More typically, one might set a retry rule for quota errors to a day.
After a day of full mailbox, Exim doesn't bother trying more than once
for each message before bouncing it.
--
Philip Hazel University of Cambridge Computing Service
Get the Exim 4 book: http://www.uit.co.uk/exim-book
