IP Address duplicates/conflicts when cloning from a template

F

fsociety3765

Member
Proxmox Subscriber
Jan 23, 2021
51
19
13
Hi all,

One of the first things I did after setting up my PVE was creating a basic installation of Ubuntu Server, get it set up with all the basic bits, and then convert it to a template.

I have cloned (full clone) this template a number of times and I did notice that each time I cloned it, the new VM would boot up on the same IP as the previous one I set up. I obviously checked mac addresses and PVE is assigning a new mac address to each clone, so I was a bit confused as my router shouldn't have been assigning an in-use IP to a different client. Most of the servers are still set to DHCP while I am going through setting each one of them up. I tried renewing the DHCP lease from the console but that didn't make a difference. What I had to do, was a sudo dhclient -r ens18, and then remove the network interface from the hardware of the VM and then add a new one back. Then it would pick up a new DHCP address. I thought it was a bit weird, and a bit of pain but I went through each VM instance and did this one by one, while at the same time setting a unique hostname.

Now, something even weirder has happened. I had to restart my PVE a couple of times today, and when I went to access one of my VMs in a web browser I could not connect. What found was that a whole bunch of my VMs were up on the same IP address. Some even had the conflicted address and then another unique address. It was really weird. I have currently got all the conflicting ones shut down while I investigate what the problem might be. I've probably done something wrong but I'm hoping it's fixable.

What is the current guidance on cloning from a template and networking configuration of the new machine? Is there something I should have done before converting to a template or is there something I need to do now?

Any help and advice would be greatly appreciated.

Thanks,

FS
 
Last edited:
In the end it is just the MAC address of the virtual NIC that is the same, change the mac address in the VM and the problem is solved. machine-id is a deep rabbit hole and the tools to change it are lousy and really convoluted.

Machine-ID also is not used by a lot of things, the MAC address of a NIC is a fundamental ID of a network interface, this is a much easier way:

Supposing your vm has only one virtual NIC, you can

  • stop the vm
  • from pve web interface, remove the virtual nic (eg: eth0)
  • create a new one, this will get a new MAC, and your vm will always see it as eth0.
  • restart the vm and check if everything works fine
This is done at the host level, so is guest OS independent.
 
Last edited:
This is so strange.
Had the same problem.
The strange part is that i created a new MAC on the same interface on VM1, and still the dhcp server said it delegated the same IP to both different MAC adresses on VM1 and VM2. I had to remove the adapter, and add a new one on one of the machines to pull a new IP.

Incredible

Thanks for the Post and answers :)
 
  • Like
Reactions: stefa
If you want to read more regarding this, this article explains how the DUID, which is used as a identifier for dhcp-clients, is generated: https://unix.stackexchange.com/ques...network-duid-iaid-and-dhcpv4-clientidentifier

In essence, all the values used are the same for common vms - a static prefix, the machine id and the interface name (in cloud init oftentimes eth0).
To generate a new one, you can do the following:

rm -f /etc/machine-id && dbus-uuidgen --ensure=/etc/machine-id && rm /var/lib/dbus/machine-id && dbus-uuidgen --ensure

For reference, i took that from here: https://unix.stackexchange.com/questions/402999/is-it-ok-to-change-etc-machine-id

If you want to change the machine id in an offline image, you need to make sure the file exists - otherwise systemd-networkd will not be able to create a DUID for dhcp-requests.
 
Last edited:
You must log in or register to reply here.
Top Bottom