I don't use Tailscale myself, but it could potentially be as simple as using Tailscale to connect to the LXC guest, then accessing the PVE web UI via its usual IP and port (assuming you don't have any firewall restrictions in place).
Known issue that's being addressed currently. You can try an experimental kernel build if you're brave:
https://forum.proxmox.com/threads/slow-memory-leak-in-6-8-12-13-pve.168961/post-803023
To close the loop on this, I ended up restoring from a backup made before the lxc.idmap: ... lines were added to the 103.conf file. After re-adding the lines and rebooting the container, the backups now run without issue.
Thanks Fabian.
Here is the output of ls -ln on the paths in question:
root@jellyfin:~# ls -ln /var/cache /var/lib /var/log
/var/cache:
total 19
drwxr-xr-x 3 0 0 5 Jan 25 06:38 apt
drwxr-xr-x 2 0 0 7 Jan 8 23:45 debconf
drwxr-x--- 6 107 4 6 Jan 16 01:54 jellyfin...
Bumping this thread as I appear to be running into the same issue.
Error logs:
INFO: filesystem type on dumpdir is 'zfs' -using /var/tmp/vzdumptmp991849_103 for temporary files
INFO: Starting Backup of VM 103 (lxc)
INFO: Backup started at 2022-01-24 04:00:55
INFO: status = running
INFO: CT...