Graeme Fowler via Exim-users <exim-users@???> writes:
> Howdy
>
> Firstly - please subscribe to the list (or post from your subscribed address) so we don’t have to keep allowing your emails out of the mod queue.
Sorry, I have read list via gmane. Now I am subscribed.
>
> On 5 Oct 2018, at 19:11, Kamil Jońca via Exim-users <exim-users@???> wrote:
>> After discussion on systemd list we have conclusions:
>> (https://lists.debian.org/debian-user/2018/10/msg00054.html
>> and thread
>> https://lists.freedesktop.org/archives/systemd-devel/2018-September/041395.html
>> )
>>
>> 1. exim forks and creates background process to deliver mail
>> 2. systemd, after main process exits, kill all remaining proceses, so
>> ...
>> 3. background exim process are killed during delivery, and message remains in
>> queue.
>
> …how often do you run the queue?
30min
>
> It doesn’t address the scenario you raise, but if you’re running it
> frequently enough then the death of a delivery subprocess from within
> systemd shouldn’t matter one jot.
Not quite. It is possible (and I observed it) another scenario:
1. exim starts deliver message and put in queue (yes?)
2. background process starts deliver message and deliver it, but ...
3. backround process is killed just before it can report delivery
success, so message remains in queue
4. queue process deliver message for second time ...
We end up with duplicate ...
>
> That said: setting KillMode=none seems like an easy fix to me.
Yes, although it should be done for every service which might send mails.
>
> Graeme
KJ
--
http://wolnelektury.pl/wesprzyj/teraz/
I'd probably settle for a vampire if he were romantic enough.
Couldn't be any worse than some of the relationships I've had.
-- Brenda Starr
