Help with getting a VirtFusion based VPS working with Talos (QEMU)

[benosman]

Hi all,

I'd like to add a simple VPS to my newly created OCI Talos cluster, but I'm having some issues.

Background: OCI based cluster is 3 Arm based nodes in control plane, and 1 worker node. They all communicated over private subnet, kubespan is enabled, and there is a load balancer pointing at the talos and kubernetes APIs. That part is all good.

I want to add an external VPS node to the cluster, it is provided by an independent web host, and they use the VirtFusion QEMU Virtualization platform. It is pretty rudimentary, I can pick from a choice of linux os, boot from a preselected or custom ISO.

There is no DHCP server, there is however a RO virtual disk at /dev/sda that contains cloud init files - meta-data, user-data and network-config

Approach 1. Install nocloud Talos variant

Given that there is a disk with the label cidata, i figured I could use the nocloud variant and it could load the network config from that disk and it would work. So using the rescue image I flashed the primary disk with talos no cloud.

However, it doesn't seem to pick up the drive, and so when it boots it just fails with network unreachable.

So with this, I reached a dead end - is there a way to get Talos to load config from this drive?

Approach 2. Install metal ISO Talos variant

For this, I need to boot from the ISO and manually set the networking, then run machine config to install on to the drive.

I was having some issues where when it rebooted after configuration, or if it booted at any time without the ISO present it would fail to become ready. As I was going through it systematically to document it, it now seems to work, so for now I will go with this approach unless I start hitting that issue again.

I am curious though, why approach 1 doesn't work, and if it could be made to.

[smira]

If you have problem like that (network not configured), please submit full console logs. It is called 'serial console logs' usually in the UI.

[benosman]

Unfortunately, I can't see anything like that in my hosts UI. The view I have screenshotted is from the VNC plugin they offer, but that is it.

Are these logs persistent between reboots? I could try boot from the rescue image and retrieve them if so.

[smira]

No, but QEMU does support them for sure. If you have VNC, you can do Alt+F1, switch to the first console and scroll back the logs to see all of them on a bigger screen. But either way, if Talos fails to pick up nocloud, it would say so and why.