Phil wrote:
>When I wrote Exim I wasn't happy with this, figuring you wouldn't want
>Exim picking up a new configuration until you wanted it to, but OTOH, of
>course, any updating of the file gets picked up by non-daemon processes
>right away, so my worry is probably a silly one.
>
>However, I still wonder whether this is a good idea for another reason.
>To make it "automatic" you'd have to have the daemon statting the file
>pretty regularly; if you left it till, say, each queue run, it would
>behave oddly from the user's point of view. Also, the daemon currently
>wakes up only on the queue run timer or the arrival of a new message, so
>implementing this would require serious surgery, and I'd rather leave
>the daemon's code alone.
I have a related wish. We have one server whose configuration changes
very rapidly (the networks/domains we allow through it varies
rapidly). Every time the configuration changes, we HUP exim and it
rereads it. So far so good.
Unfortunately the already running queue-runners don't appear tracked by
the new exim daemon. [Indeed they appear to become orphaned, as the
daemon respawns].
We had the machine close to running out of memory the other day (with over
400 exim processes running :), so keeping track of them is important.
Is this a behaviour which could be changed? Could exim reread it's
configuration without respawning?
