Early node discovery and talos API access before cluster is bootstrapped

[mj-sakellaropoulos]

We are working on a CLI tool which discovers nodes and bootstraps the cluster automatically. Previously, we discovered nodes with talos.config and assumed the first node to hit our endpoint was a control plane and that all install disks were /dev/sda

We wanted to offer a choice of node role and install disk for each machine discovered, but hit a roadblock since the Nodes remain in Booting stage while trying to contact talos.config endpoint - which prevents access to most of the API.

Right now the workaround we are using is to discover nodes via talos.events.sink, that way Nodes can enter Maitenance mode instead of waiting for the machineconfig and being stuck in Booting.

We are wondering if there is a better way to do early node discovery while still being able to get install disk info.
From what I understand, we cannot use the discovery service features, since that is only available after the cluster has initially been created.

[smira]

You can look into how Omni works - basically it solves the problem you are describing.