On Thu, Jan 13, 2022 at 10:03 AM Michael Biebl <bi...@debian.org> wrote:
> Am 13.01.22 um 08:36 schrieb Jérémy Lal: > > Package: systemd-oomd > > Version: 250.2-2 > > Severity: normal > > > > Hi, > > > > i've installed systemd-oomd 250.2-1 and on upgrade, > > noticed this: > > > > systemd-oomd.socket: Socket service systemd-oomd.service already active, > refusing. > > Failed to listen on Userspace Out-Of-Memory (OOM) Killer Socket. > > > > I'm not sure what should be done. Stop the service, disable the socket ? > > > The behaviour of dh_installsystemd is to treat systemd-oomd.service and > systemd-oomd.socket as completely separate services. > So it generates code for both units to enable/disable and restart those > units (more or less independently). > > You can't really restart a socket unit though which is bound to a > running service. > > What you see is basically a duplicate of > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=955483 > > We don't have a good answer here. This probably needs support from > upstream as well. > > As a workaround, we could probably drop the restart/start of > systemd-oomd.socket and only keep the enable/disable/stop-on-removal bits. > > For that we'd need separate calls to the dh_installsystemd override. One > for systemd-oomd.service and one for systemd-oomd.socket. > > This is untested: > > override_dh_installsystemd: > ... > dh_installsystemd -Psystemd-oomd systemd-oomd.service > dh_installsystemd -Psystemd-oomd --no-start --no-stop-on-upgrade > systemd-oomd.socket > > (not quite sure if we need --no-restart-after-upgrade as well). > > This has the downside that the socket is no longer restarted on > upgrades, so changes to the socket unit are not applied. > It has the upside, that it doesn't interfere with the currently running > .service unit. > > Ideally, this would be handled better by systemctl and/or debhelper by > knowing that those units belong together and restarting them in the > correct order. > > I believe the ultimate fix is to actually fix https://github.com/systemd/systemd/issues/8102 upstream. AFAICT, that would resolve everything? > Balint had some thoughts in #955483, but nothing really happened since then > > > > > -- Saludos, Felipe Sateler
