On 16/10/2022 16:34, Lennart Poettering wrote:
On Fr, 14.10.22 10:59, lejeczek (peljasz@xxxxxxxxxxx) wrote:
Hi guys.
I'm on Centos 8 S with systemd 239.
Users homes are mounted at later (latest?) stage off NFS so when such a user
logs in then:
-> $ systemctl --user status -l xyz.service
Unit xyz.service could not be found.
-> $ systemctl --user daemon-reload
-> $ systemctl --user status -l xyz.service
● xyz.service - Podman container-xyz.service
Loaded: loaded (/apps/appownia/.config/systemd/user/xyz.service; enabled;
vendor preset: enabled)
Active: inactive (dead)
Docs: man:podman-generate-systemd(1)
Is it possible and if so then how, to make "systemd" account for such a
"simple" case - where home dir is net mounted very late?
I don't get this scenario. You talk to the systemd --user instance,
which is the per-user instance, so $HOME of that user should be
mounted at that time. But then you issue a reload and new stuff
appears and you appear to suggest that now the user's $HOME was
mounted?
Yes, it appears that systemd 'misses' user's units - @ boot
time I guess and later does not do anything about them
'homes' getting mounted later(too late for systemd?)
When I login to that user it seems that systemd sees no
unit's - at that point homes are mounted of course - and I
have to poke systemd manually with:
-> $ systemctl --user daemon-reload
at which point, it seems, system re-check user's home
It is PCS service/daemon which is a part of or rather
manages, a couple of other daemons which together do a HA.
For 'pcs' I've tried
[Unit]
Before=systemd-logind.service
Before=systemd-user-sessions.service
but do not avail.
So what now? Usually, the assumption is that first the user logs in,
which is the point where $HOME must be mounted at the latest, and then
systemd --user gets started off it and the user's login session is
allowed to begin.
I really do not know what to assume - I can only tell you
what happens.
Ideal scenario - which I'm hoping it is possible to make
work - is when a given user does not have to log in at at.
As you can see in this case, it is a container(podman) unit
which I hope a 'lingered' session shall manage without user
interventions.
many thanks, L>
Lennart
--
Lennart Poettering, Berlin
