Re: [exim] Exim with SQL message store

On Mon, May 10, 2010 at 10:33 AM, Marten Lehmann <lehmann@???> wrote:
> Hello,
>
> exim is our MTA of choice for more then 10 years now, but we are very
> unhappy with the message queue. Header, body, delivery attempt logs and
> delivery hints are spread throughout dozends of subdirectories, text
> files and berkeleydb files, which make it very hard to look at the queue
> in depth, like "show subject and queue-in date of all messages by
> authenticated_id x and freeze them". The exim tools are a nice start for
> easy tasks, but it is almost impossible to work with the queue and query
> or operate on it directly since it is mostly text based and only exim is
> able to get the structure with its parsing routines.
>
> I have not looked into the queue source in detail, but does exim keep an
> abstract layer to access to queue, so it would be possible to replace
> the current text file queue to an sql backend? Or does each exim twiddle
> directly in the files, adding delivery attempt counts, new headers etc.?

This sort of stuff is almost exactly why I wrote exipick, have you
looked at it? Taking your example, I'd do it in two passes, but it's
still fairly do-able:

display:
exipick --show-vars h_Subject,received_time --sort authenticated_id

freeze:
exipick -i --show-vars h_Subject,received_time --sort authenticated_id
| xargs exim -Mf

exipick has been pretty stable for a while because it does what I want
it to do and no one's asked for it to do more. If you think the above
id interesting and have ideas for refinements I'd be interested in
hearing them.

As for a queueing abstraction layer, you're out of luck with what's
currently in the code.

--John