Cloud-Init overwriting entire Proxmox VE node (7.1)

[exounx]

I'm trying to use cloud-init so I can setup k8s nodes much more quickly but the node I set it up on gets overwritten by cloud-init on reboot.

I followed this exactly - https://pve.proxmox.com/wiki/Cloud-Init_Support

I've done this many times before on PVE 6.x with no issues like this.

This has happened twice now. I thought I had done something wrong initially so I just deleted the node and created a new one.

However I tried again while being extra careful and it happened again.

It's overwriting the proxmox node with the cloud-init of the VM that I create on this.

I also have ceph setup. The storage drive gets placed on ceph but it mounts cloud-init from local-lvm.

I'm not quite sure what I could be doing wrong here...

Help appreciated.

 

[fabian]

you are not supposed to install cloud-init on the PVE host - you should install it inside the VM (either via a cloud-init enabled disk image that you import, or manually)

 

[AppState95]

I ran into the same problem as the OP. After rebooting the node from the next round of updates after following the Cloud-init doc, the node booted up with the hostname of the Cloud-init VM that I had created and was no longer showing as active in the cluster. It did, however, still use the root password for the pve node, so I was able to log in.

The steps that I used to recover (without having to rebuild the node) were:
1. Deleted the cloud-init OS .img file that I had downloaded, which in my case was in the /root directory.
2. apt remove cloud-init
3. apt autoremove (to remove the additional python modules and other leftover stuff related to cloud-init)
4. hostnamectl set-hostname <original node hostname>
5. reboot

After a reboot, the node returned to active in the cluster, however, its host keys had changed, so I had to remove the entries from the other nodes' /.ssh/known_hosts files and reconnect (via ssh as root from the other nodes to the rejoined host).

 

[ganon]

@AppState95 Thank you very much, you saved my homelab.
I wouldn't lost all since I had backups but they were a bit old...
Thank you again for coming posting this answer here even 7 months later I'm sure it's useful for a lot of users !

If anyone else come here, the tricky part will be maygbe to have the original node hostname.
For me I though it was pve but after reboot, I had access to the proxmox GUI but I had one node pve running empty and one pve1 offline.
So I changed my hostnamectl to pve1, rebooted and all is working again