[SOLVED] PMG in a LXC container: 'timedatectl show' fails (used by UI)

[sthielo]

TOPIC 1) timedatectl failure (UI ... and shell)
first time I noticed after recent update:
UI-Configuration-System shows an error "command 'timedatectl show' --property=Timezone' failed: exit code 1 (500)

and actually in shell [of PMG LXC as root]: timedatectl show -> Failed to connect to system scope bus via local transport: No such file or directory

How to get rid of that? Is that new issue - I never noticed before? Is UI/API supposed to use timedatectl or does that break LXC support?

TOPIC 2) sys-kernel-config.mount failed
while searching for info: finding some other issues (systemctl degraded: sys-kernel-config.mount) in journalctl, made me migrate (backup/restore) to a privileged LXC, which solved that issue, but still has the timedatectl issue.

I would prefer to have a non-privileged one. Doc says LXC is supported. Is 'unprivileged' LXC for PMG supported? ... then how to properly solve the sys-kernel-config.mount issue in an unprivileged container?

proxmox-mailgateway-container: 9.0.0 (API: 9.0.7/d11aa4b692ae, running kernel: 7.0.0-3-pve)
pmg-api: 9.0.7
pmg-gui: 5.0.5
clamav-daemon: 1.4.3+dfsg-1
ifupdown: residual config
ifupdown2: 3.3.0-1+pmx12
libarchive-perl: 3.7.4.1
libjs-extjs: 7.0.0-5
libproxmox-acme-perl: 1.7.1
libproxmox-acme-plugins: 1.7.1
libpve-apiclient-perl: 3.4.2
libpve-common-perl: 9.1.12
libpve-http-server-perl: 6.0.5
libxdgmime-perl: 1.2.0
lvm2: not correctly installed
pmg-docs: 9.0.6
pmg-i18n: 3.7.3
pmg-log-tracker: 3.0.1
proxmox-mini-journalreader: 1.6
proxmox-spamassassin: 4.0.2-1
proxmox-widget-toolkit: 5.2.2
pve-xtermjs: 6.0.0-1

This looks to me like a container topic. Any hint/help how to solve or track down appreciated.

 

[fabian]

please post the full config of the container..

 

[sthielo]

full config

 

[sthielo]

config of container in PVE

[~]$ pct config 204   
arch: amd64
cores: 2
features: nesting=1
hostname: mailgw
memory: 4096
nameserver: 192.168.101.203
net0: name=eth0,bridge=v101,gw=192.168.101.1,hwaddr=BC:24:11:91:72:90,ip=192.168.101.204/24,type=veth
onboot: 1
ostype: debian
rootfs: thvmstore:vm-204-disk-0,mountoptions=discard,size=10G
searchdomain: thiel.internal thiel.team
startup: order=10
swap: 1024

 

[fabian]

okay, that looks correct. anything visible in the logs? does the problem persist if you reboot the container?

 

[sthielo]

- persist on reboots. ... updated to newest packages before posting and rebooted. ... same error
- nothing suspicious in /var/log/**/* or journalctl . systemctl status is happy.

only that error mentioned in UI and when executing on cmdline.
note: actually everything seems to work fine, but that UI error and timedatectl in shell (is that supposed to work within a LXC?)

 

[fabian]

yes, it is.. anything visible in the host journal?

note that privileged containers are not recommended for security reasons, the second issue from your original post is benign and can be ignored.

 

[sthielo]

topic 1)
note: ... meanwhile I upgraded [and rebooted] the pve hosts to see if that changes anything (last upgrade was ~20d ago ... but things change fast ;-) )
=> no change. BTW: same behavior on an other host, which I tested while the CT was migrated (stopped/started) to another host by ha-manager during the upgrade procedure.

journalctl on host-PVE does not show anything suspicious, too

I would expect something on PMG-LXC for
a) systemctl | grep dbus (isn't that the thing that timedatectl's err msg is referring to) and
b) systemctl | grep time (other then *.timers, proc-uptime.mount, timers.target) ... so something like *timedated* or *timed*
systemctl list-unit-files at least shows systemd-timedated.service static -
but for dbus it only hows some entries for freedesktop with state 'alias'

note: I have a PBS as container/LXC in my homelab which shows dbus.service and dbus.socket and systemd-timedated.service ... and timedatectl works there!
the only "similar" container. as everything else in containers is using Alpine.

-------
topic 2)
... I converted back to "unprivileged" - thanks
sys-kernel-config.mount loaded failed failed is back in systemctl ... I'm ignoring it.

 

[fabian]

yes, it seems the dbus-system-bus is only recommended by systemd, and hence missing in this minimal container. could you try installing the "dbus" package?

 

[sthielo]

YES - it was as simple as that. apt install dbus && systemctl start dbus
So, I do NOT understand how this package can be missing. ok, my container is quite old (2021) made some upgrades ... even some major ones from 8 to 9.
Thanks a lot.