OS Inmutable as core of Proxmox.

[HJBG]

How to implement Proxmox in its various variants on an immutable or atomic operating system, such as Fedora SilverBlue, Fedora Core, VanillaOS, or others. Also, how to improve the update system and restore in case of an error similar to XCP-NG.

For me, these are key suggestions that would put Proxmox in a superior state with the implementation of immutable operating systems and fail-safe updates. In my opinion, this last factor is Proxmox's Achilles' heel, where it has not yet been achieved that each node in a cluster can be updated without fear of it not starting later and that it does so independently, not all nodes at the same time. The philosophy inherited by XCP-NG in this regard has been very pragmatic and saves many headaches when there are impacts on the affected hypervisors and the consequent loss of time, money, prestige, etc.

I think it's a trend, and perhaps we should consider other rock-solid paradigms. Implementations like Podman, Docker, Kubernetes, OSTree, ABRoot, and related technologies should be considered and included in future releases.

 

[Johannes S]

How to implement Proxmox in its various variants on an immutable or atomic operating system, such as Fedora SilverBlue, Fedora Core, VanillaOS, or others. Also, how to improve the update system and restore in case of an error similar to XCP-NG.

It would also make host backups a lot easier because then you would just have to backup the modifiable files (configuration etc).
So I see the benefits, but I also see a huge issue: A big benefit of ProxmoxVE is it's flexiblity since it's basically a regular Debian with an Ubuntu kernel. So different than e.G. TrueNAS (which you acutally can't modify during runtime by default and where the official docs clearly stay that only their official tools, API or WebUI should be used) you can do a lot of stuff which isn't supported but still possible (like using a different kernel for otherwise non-supported hardware, installing third-party software e.g. for monitoring or debugging purposes, using non-integrated file systems like gluster, ocfs or gfs...). This flexibility would be lost with an immutable system.

For me, these are key suggestions that would put Proxmox in a superior state with the implementation of immutable operating systems and fail-safe updates. In my opinion, this last factor is Proxmox's Achilles' heel, where it has not yet been achieved that each node in a cluster can be updated without fear of it not starting later and that it does so independently, not all nodes at the same time.

I see your point but shouldn't this be mitigated by the sheer existence of a cluster? I mean you could always remove a node from the cluster, reinstall it and add it again.

I think it's a trend, and perhaps we should consider other rock-solid paradigms. Implementations like Podman, Docker, Kubernetes, OSTree, ABRoot, and related technologies should be considered and included in future releases.

Since this is the community forum this suggestion might get lost (I'm also not a staff member) so my suggestion is to open a request at https://bugzilla.proxmox.com I'm definitively interested what the ProxmoxVE developers think on this.