Proxmox node turned into vm after reboot

[nsky.link]

Please help me understand how this can happen and why. How to return the Proxmox node to its original state?!
The Proxmox node has a hostname for example "node7" and is in the cluster for 7 nodes. Has zfs-RAID10. 2 virtual machines (vm3 and vm4) installed with the help of terraform worked on the node.
What was done before the problem was identified:
- network reboot on the node
- turn off all (2 pcs.) vm on the node
- command in the console "reboot"

After that, I could not enter via ssh to the node. Already in ipmi I saw an invitation to the console but already under the hostname vm4.

I noticed that the file "authorized keys" was overwritten and the proxmox and haproxy services were down. All other services such as Zabbix, Nagios, etc. are working fine.

i do pvesr.service restarting and give with error "vm4 pvesr: ipcc_send_rec[1] failed: Connection refused

Network is working normal.

From web interface "node6" i see what a "node7" is unavailable (grey)


I also attach a screenshot from ipmi after the first reboot
P.S. Thanks in advance!

 

[mira]

You installed cloud-init on the host. This has to be installed in the VM!

First I'd suggest purging cloud-init from your host: apt purge cloud-init
Afterwards check /etc/network/interfaces and make sure it is set up correctly again (as it was before).
Then edit /etc/hostname and /etc/hosts to match the ones on your other nodes, but with the right hostname and IP for this node.

This should be enough to get it up and running again.

 

[nsky.link]

Thanks a lot ! you are my savior. I did everything you advised and the node really returned to the cluster. Now everything is working normally!

You installed cloud-init on the host. This has to be installed in the VM!

First I'd suggest purging cloud-init from your host: apt purge cloud-init
Afterwards check /etc/network/interfaces and make sure it is set up correctly again (as it was before).
Then edit /etc/hostname and /etc/hosts to match the ones on your other nodes, but with the right hostname and IP for this node.

This should be enough to get it up and running again.

 

[mira]

That's great!
Made the same mistake a few years ago.