Talos install error - couldn't get current server API group list: - tls: internal error

[DavidIlie]

Following the pathway to install Talos, I have an issue where the cluster does seem to setup but the workers do not join, and the masters have these errors:

error refreshing pod status and the error is related to TLS (tls: internal error) and also controller failed errors too

this is what I see in my terminal while setting it up

[onedr0p]

I think this is expected due to etcd not being bootstrapped?

https://github.com/onedr0p/cluster-template/blob/b6234fcbad3f9b3be13f94a563b6b73cbe3741f5/.taskfiles/Talos/Taskfile.yaml#L55-L63

It doesn't seem like the above was ran. At what point in the task commands did you get to and was there any errors on the client side?

[bojanraic]

In addition to @onedr0p's comments, when I tried to install Talos manually, it took some time for the master to be ready. In both of the screenshots, uptime is only a few minutes so maybe it hasn't finished bootstrapping yet. I went back to k3s in the meantime, but please update us on your Talos install progress via this template and I may take another stab at it when time permits. Good luck!

[DavidIlie]

I left it running the whole night yesterday and the same thing happened. I am also sure that I think all scripts are running and before the node first reboots/loads there are errors regarding something like a "admin" certificate

Any ideas?

[onedr0p]

Maybe give it another shot when you have a moment? Not sure what happened here to be honest could be a ton of different issues from misconfig to network issues to anything else really :/

The important bits of the config that can really go wrong if not set right are the network and disk selectors.

[DavidIlie]

Disk selectors work I believe, data is being written to the disk and network is working o I believe on all the nodes.

The error is just the "tls: internal error" every time the masters try to fetch something from their own localhost IP

[DavidIlie]

The bootstrap first begins with these errors in the console

But I believe that's when the nodes get rebooted as then it boots and continues til kubelet is healthy but the error is back:

And then my terminal tries to connect to the VIP and nothing happens

[onedr0p]

I saw this in your previous config (sorry this is all I have to go on from https://github.com/onedr0p/cluster-template/issues/1398#issuecomment-2029711711)

    networkInterfaces:
      - deviceSelector:
          hardwareAddr: ""

That should be the nodes mac address, are you sure this is populated? It should be in xx:xx:xx:xx:xx:xx format and be unique per-node.

https://github.com/onedr0p/cluster-template/blob/28ae26d3a8d1b3ea79f8552ae420faf93366dda5/config.sample.yaml#L57

[onedr0p]

I added validation on talos_nic here to hopefully catch this for other people in the future.

[DavidIlie]

I already populated those, I just redacted them when I sent it here. Every single value is present

[DavidIlie]

https://github.com/onedr0p/cluster-template/assets/47594764/302633b5-723c-431d-af21-d0d041ca1c81

This is a recording of what happens

[onedr0p]

I wonder if you need to use a different type of network selector in the Talos/talhelper config or change something in the NIC settings on the VM in Proxmox?

I just hand-held someone thru the whole repo who is using bare-metal nodes and we had success after figuring out they were not setting the correct value for talos_nic which lead me to commit validation on that.

[DavidIlie]

Do you have an example of what I would need to do?

[onedr0p]

I am probably not the best person to ask about that as I do not use any hypervisors in my life right now 😄

Maybe a good start is to review the talos proxmox docs and see if everything lines up there and with the rendered config here.

[onedr0p]

Keep in mind there are a bunch of different network selectors you can use so maybe mac address is not the best with PVE? I dunno.