Author: Piete Brooks Date: To: Christoph Lameter CC: Philip Hazel, exim-users Subject: Re: exim's retry database acting up
> When I brought it up again a huge number of messages were waiting in the > queue on the mailserver but they were not delivered. "Retry time not
> reached"! Initially the users home directory could not be stat()ed and
> thus the message was deferred.
ZAP the retry database, and do a queue run ...
> Then the most amazing thing:
> New messages coming in where queued instead of being delivered!
Sure -- it knows that delivery attempts to that user failed (for *some* reason)
last time, so it's backed off delivery attempts.
> I finally had enough of this nonsense and stopped exim and erased the
> retry database. On startup exim delivered all messages to their
> destination.
That's the way to do it (not sure if a restart is needed, but then I don't run
exim like that ...)
> What exim needs to do is to deliver ALL messages when the first
> .forward file from those homedirectories was successfully read.
*YOU* know that, but how should exim ?
exim doesn't keep track of *why* it failed, but in particular, how does it
know that all .forward files are on a single server ?
(we have multiple home di servers ....)
> There needs to be some easy way to get rid of those retry times.
ZAP the retry database ...
> exim_fixdb was not documented in such a way that would have allowed me to
> use it on a huge number of messages.
Need a perl script .... :-)
> The way that exim runs the queue made it take a couple of hours to get
> through it. Messages are delivered one by one and since the queue also
> contained messages to hosts not reachable. There was a significant delay
> on those since exim was taking a long time to timeout on those problem
> messages.
I run multiple processes "exim -q & exim -q & exim -q & exim -q &"
> Could exim timeout within 10 seconds on those slow messages, fork them
> into a separate process and continue delivering the rest of the queue
> faster?
What if you have (say) lost your internet connection and have (say) 10K
hosts to which email is to be sent -- you want 10K processes ?
