I found out, I had to change the config in
/etc/pve/lxc/<vmid>.conf instead of
/var/lib/lxc/<vmid>/config.
But while doing so, I seem to have broken something. First I couldn't start my test container as LXC couldn't acquire access to my chosen idmap
Code:
lxc.idmap = u 0 200000 65536
lxc.idmap = g 0 200000 65536
.
After adding my idmap range in
/etc/subuid and
/etc/subgid, LXC accepted the config and started the container but I cannot start it using
pct. Instead I get these errors:
Code:
root@appserver3:~# pct start <vmid> --debug=1
problem with monitor socket, but continuing anyway: got timeout
get_rundir: 261 HOME isn't set in the environment
Failed to create lock for <vmid>
main: 242 Failed to create lxc_container
_rundir:261 - HOME isn't set in the environment
ERROR lxc_start - tools/lxc_start.c:main:242 - Failed to create lxc_container
Actually, I cannot start any container anymore, with this same error. And I cannot find any information on this. The error message
is generated in this file at line 263 shortly after the check for XDG_RUNTIME_DIR which it obviously passes but which also is the only context with which I can find results in the web.
The only action using the regular Proxmox toolkit I can remember which resulted in an error was my attempt to clone a container snapshot while I've been in the directory of the destination <vmid>. Rsync failed, but after leaving and removing the directory, the repeated attempt completed successfully. Interestingly, when starting the container via
lxc-start -n <vmid> (after having mounted it's rootfs partition), it works without error and with every container I tried but failed using
pct.
Oh, there was another massive action I had to make: I had to
kill an
lxc-start -F while debugging the container launch. Couldn't figure out how to stop the container until after I
kill'ed it. Yeah, don't laugh at me, please. I didn't think of
lxc-stop :-D.
Finally, even undoing all I could undo didn't help me with starting containers again using
pct (or the web UI). Any hints?
