Please help!! Reserving IP in isolated network

[SviridoffA]

Hello everyone! Maybe someone can suggest some kind of workaround for me. I have an isolated network where a Kubernetes cluster is running, which I can't stop. I need to add a physical server to this isolated network (setting up a VLAN on the switch port). I'm trying to reserve IP addresses for the static address of this server via CIDR. My subnet is 10.1.1.0/24, and if I try to set, for example, 10.1.1.0/25, I get the network split in half, but right now in the half that I'm trying to free up from the DHCP pool, there are already working Kubernetes nodes. Unfortunale? I don't manage this k8s cluster, and I can't do any changes in this cluster. Is there any workaround to solve this situation? For example, can I somehow expand the network live? Maybe someone has done this through the database (I understand it's not the best practice), or is there some more proper way?

[StepBee]

A potential workaround i could imagine if it's only about one physical server:
Create a dummy VM within the isolated network.
At the time of deployment, you can already assign the IP of your choice to the dummy VM.
Keep the VM stopped and the IP will not be assigned to any other VM.
To make sure the IP is not used by an OS within the VM, you could keep the VM without an installed OS for example by deploying the VM via ISO instead of template and never install the OS.

I am not sure if this is the clean solution you are searching for, but i think it's a very easy workaround without any harmful changes.

[SviridoffA]

Hi @StepBee ! and thanks for your reply, I was thinking in this way and I'm not discarding this option. But I would like a cleaner solution. I'm not sure this solution exist, but maybe someone has done something similar. Also, as a plan B, I'm thinking about creating a separate L2 network for physical servers. But I don't want to pass traffic through L3 for now. In general, I think it would be a nice feature if in isolated networks it was possible to regulate the DHCP range as it can be done, for example, in shared networks.

[SviridoffA]

I talked to my client and in the end we decided to use the option with a fake virtual machine, because for the client now unacceptable stops in the work of virtual machines for further reconfiguration, and also does not acceptable option with L3. I created a virtual machine in the off state and deleted its disk. Also took this virtual machine to a separate project which has access to this network so that it does not appear among the client's virtual machines. For now, we decided to solve problem in this way.

Also I noticed that at the moment there is already a feature request for this #12661. So I am closing this discussion.