[SOLVED] proxmox8.1 centos7 and rocky8 lxc templates?

[EmilioMoreno]

Hi
I am going to migrate from proxmox6 to proxmox8.1 by installing a new cluster and re-provision all my lxc servers and Kvms in a kind-of blue/green upgrade. I am using ansible to create lxc and KVM so provision is simple and fast (well, not so fast, but it is automatic) , but I am not finding centos7 or rocky8 lxc templates in the proxmox8.1 system.

Is there any way to download them or use previous proxmox templates in 8.1?

and yes, I know I should be using modern versions of centos/rocky, but some of our production systems have requirements can only be met with "old" OS like centos7/rocky8, so upgrading all is not an option.

Thanks
Emilio

 

[VictorSTS]

You can update your current cluster to PVE7 and then to PVE8. If for whatever reason you can't update your current nodes, a simple backup/restore can move your VM and LXC to the new cluster. In any of these two methods, check systemd version in the LXCs, you may need to disable cgroupv2 [1].

I haven't tested, but you can download the LXC templates from PVE6 host. Its a file in /var/lib/vz/template/cache and upload them to your updated host. Then use it to deploy new LXCs. Just checked and PVE7 also has templates for rocky8/centos7.

[1] https://pve.proxmox.com/wiki/Upgrade_from_6.x_to_7.0#CGroupV2

 

[Stoiko Ivanov]

Keep in mind that running outdated OS-versions in containers is nothing that is supported - which means we don't test it too extensively, and there might be problems when running an old userspace on a new kernel (LXC shares the kernel from the host-node).

If you need to run legacy OS versions I would consider using full VMs because the provide a better isolation.

I hope this helps!

 

[EmilioMoreno]

You can update your current cluster to PVE7 and then to PVE8. If for whatever reason you can't update your current nodes, a simple backup/restore can move your VM and LXC to the new cluster. In any of these two methods, check systemd version in the LXCs, you may need to disable cgroupv2 [1].

I haven't tested, but you can download the LXC templates from PVE6 host. Its a file in /var/lib/vz/template/cache and upload them to your updated host. Then use it to deploy new LXCs. Just checked and PVE7 also has templates for rocky8/centos7.

[1] https://pve.proxmox.com/wiki/Upgrade_from_6.x_to_7.0#CGroupV2

Hi Victor, thanks for your suggestions.

I know i can upgrade from 6,to 7 and then 8, but in my tests the upgrade process from 6 to 7 has proven to be unreliable in a small percentage of the test, but not so small to discard the risk for the production environment. I am running a 8 server cluster in development environment and a bigger one for production and while i have no issue with the devel env, definitely the blue/green upgrade with provision seems to be a safer way to go for the production environment.

As for the process itself, as you suggested, i have directly gone for the tar.xz templates in both my pve6 and directly download.proxmox.com, downloaded into the cache directory and been able to run both containers (disabling cgroups2 for centos7) and everything seems to be working fine, i need to do some heavy testing though, thanks for the help!

 

[EmilioMoreno]

Keep in mind that running outdated OS-versions in containers is nothing that is supported - which means we don't test it too extensively, and there might be problems when running an old userspace on a new kernel (LXC shares the kernel from the host-node).

If you need to run legacy OS versions I would consider using full VMs because the provide a better isolation.

I hope this helps!

Thanks Stoiko. While I have not gone in detail yet through the pve8 documentation, I see that both rocky8 and centos7 were supported in pve7 and pve9 will not support legacy cgroups controller, but I am a bit lost so far with pve8, can you point me to any specific/already known issues or areas I should check?


I would like to avoid the vm route, provisioning is quite easy and automated, but performance difference, specially in heavy networking environments. is really impressive with containers and migrating to rocky9 or similar newer systems is a no-go for us by now.