Installing "Workstation" on CentOS 8 and rebooting "firewalls" my LXC container

[koalillo]

I just submitted the following bug, spamming here to get thoughts:

https://bugzilla.proxmox.com/show_bug.cgi?id=2674

Apparently, installing the "Workstation" group on an LXC CentOS 8 container and restarting the container prevents further connections to non-SSH daemons on the container?

Any idea why this might happen?

 

[tom]

I assume the group "Workstation" is not really the best selection for a LXC container.

If you need this, better go for a qemu VM.

And/Or check logs inside inside your CentOS to debug your issue further, I doubt that is a "bug" on Proxmox VE.

 

[koalillo]

I assume the group "Workstation" is not really the best selection for a LXC container.

Why? The purpose of this host is to provide me a desktop I can access from anywhere. It's actually working very well (using xrdp + xorgxrdp) and it's very fast and lightweight because LXC containers are faster and lighter than VMs.

I do suspect that Workstation is pulling some stuff that doesn't make sense on a container (it does pull a kernel which is going to be unused), and I'm happy to find which package is causing trouble and pulling it out.

And/Or check logs inside inside your CentOS to debug your issue further, I doubt that is a "bug" on Proxmox VE.

Doubt that this is going to work because I can connect to daemons from *within* the host (using it's IP, not using loopback), so it seems that the host is "unaware" of the problem.

I suspect Proxmox has some "magic" for tcp:22 and it's probably interacting with my issue in some way.

The other odd thing is:

$ telnet 10.42.42.75 22
SSH-2.0-OpenSSH_8.0
$ telnet localhost 3389
Trying ::1...
telnet: connect to address ::1: Connection refused
Trying 127.0.0.1...
Connected to localhost.
[alex@ws ~]$ telnet 10.42.42.75 3389
Trying 10.42.42.75...
Connected to 10.42.42.75.

; tcp:3389 prompts an IPv6 message on loopback, but that doesn't happen if not using loopback or if doing tcp:22...

 

[tom]

Why? The purpose of this host is to provide me a desktop I can access from anywhere. It's actually working very well (using xrdp + xorgxrdp) and it's very fast and lightweight because LXC containers are faster and lighter than VMs.

Because the metapackage install countless packages which are totally useless and not working inside a container. Just think of the kernel ...

BTW, I just tried the same and the logs/journal is full with warnings and errors (due to the issue from above).

But the httpd is still working. There is no Proxmox VE "magic" here.

 

[koalillo]

Well, I've narrowed it down to "Core", which is also not great, but should help. I'll try to narrow it down...

 

[koalillo]

Installing the group "Gnome" seems to pull in everything I want, so for the moment I'll do that and skip the problematic "Core".