My original message didn't make it through moderation yet, but I ended
up finding the solution on this list by accident.
tip: use -odf when calling exim from a systemd oneshot service
https://lists.exim.org/lurker/message/20220614.201406.278f227f.en.html
Sorry for the inconvenience. Since I did not expect systemd to be the
culprit here, this thread didn't come up when I looked into the problem.
I came across it now by just browsing the list.
What ends up happening is that systemd terminates an exim4 process. This
amazingly also solves a different issue I was looking at completely
unrelated to Exim. To at least provide some value here, the argument
required for s-nail is -S mta-arguments="-odf".
Thanks to **Ian Kelling for the hint.
On 06.07.22 16:49, Kim-Alexander Brodowski wrote:
> Hello fellow postmasters,
>
> I'm experiencing a weird issue I hope somebody here can help me make
> sense of.
>
> Background:
>
> We're using a small bash script, see below, to inform us when certain
> systemd units fail. This is essentially because stuff runs in the
> background and not everyone interested in the results of these tasks
> can work with a terminal.
>
> #!/bin/bash
> if [ "$#" -ne 2 ]
> then
> echo "Usage: $(basename "$0") RECIPIENT UNIT" 1>&2
> exit 1
> fi
>
> . /usr/lib/iserv/cfg
>
> echo "Sending status report of $2 to $1 ..."
>
> HOSTNAME="$(hostname)"
> HOSTNAME="${HOSTNAME:-localhost}"
> MAIL_DOMAIN="${Domain:-$HOSTNAME}"
> MAIL="root@${MAIL_DOMAIN}"
>
> s-nail --batch-mode -s "$2" -r "$MAIL" -C "X-IServ-systemd-Unit:
> $2" -. "$1" <<ERRMAIL
>
> $(systemctl status -n 0 --full "$2")
>
> $(journalctl -u "$2" --since "$(systemctl show "$2" --property
> ExecMainStartTimestamp --value)")
>
> This message was automatically generated by $0 on $HOSTNAME.
> ERRMAIL
>
> #sleep 10
>
> If anyone is interested in the corresponding unit file and config for
> other units, I can provide those as well, but I don't think they
> provide additional context here.
>
> When this script is called interactively, emails generated are pretty
> much instantly delivered to Cyrus, our MDA. When the script is called
> by systemd, the mail instead is queued for some time by exim4 and
> eventually delivered. This is where our problem lies: We prefer
> instantaneous delivery in this instance.
>
> s-nail sends its mail using sendmail, which on our platform of choice,
> Debian Bullseye, is part of the exim4-daemon-heavy package.
>
> I've traced s-nails call to sendmail and there is no difference in
> arguments when being called interactively. The problem appears to be
> somewhere on Exim's side.
>
> I've enabled debug logging, but looking at the output, I don't see
> anything of particular interest. I've attached an excerpt.
>
> Now to the oddity justifying the odd subject: In the script above,
> there is a sleep statement commented out. This shouldn't really do
> anything, except for increasing the runtime of the bash script.
> However, uncommenting that actually makes email delivery instantaneous
> again. Yeah ... It almost seems like exim4 somehow speeds things up if
> the calling process sticks around for some time. Maybe it feels it's
> being ... pressured?
>
> Joking aside, I've looked at stuff like queue_only_load, but I don't
> think these options are to blame here.
>
> Maybe somebody has an idea of what is going wrong here. Thanks in
> advance. We love Exim ;)
>
--
Kind regards,
Kim Brodowski
Telefon: +49 531 38821-01
Fax: +49 531-38821-999
E-Mail: info@???
Internet:
https://iserv.de
Zentrale: IServ GmbH, Bültenweg 73, 38106 Braunschweig
Niederlassungen:
IServ GmbH, Girardetstr. 6, 45131 Essen
IServ GmbH, Fanningerstraße 13, 10365 Berlin
USt-IdNr. DE265149425 | Amtsgericht Braunschweig | HRB 201822
Geschäftsführer: Benjamin Heindl, Martin Hüppe, Jörg Ludwig
Grundsätze zum Datenschutz:
https://iserv.de/privacy
IServ - aus der Schule für die Schule
