Andrey Smirnov

It looks like upstreaming changes didn't quite work yet? Anything we can do to help?

As we're going to branch off v1.2.0 stable branch today, I moved it to the next iteration - v1.3.0

> Update: [insomniacslk/dhcp#469](https://github.com/insomniacslk/dhcp/pull/469) has been merged and I've now refactored [insomniacslk/dhcp#470](https://github.com/insomniacslk/dhcp/pull/470) which might take one more review round. During that work I located and fixed the final long-standing bug that...

Extensions are activated during install, it could be that your node was booted from disk image skipping install?

Image can be built with system extensions pre-installed via https://www.talos.dev/v1.1/talos-guides/configuration/system-extensions/#building-a-talos-image-with-system-extensions

Yes, probably it's not clear, but when booting from a disk image install phase is skipped, as Talos is already "installed" in the disk image. That's why I referred to...

This was an issue with 1.1.0 release only, it should work for < 1.1.0 and >= 1.1.1. This is a race condition and it depends on the bootstrap order, but...

During join sequence, other cp nodes discover addresses of other etcd members via k8s control plane, which I assume works, as you see in the logs errors about the connection....

There should be also a message in the log of the joining node(s) with something like `retrying: ... `, this might give a hint on why joining node can't connect...

It's hard to say what the issue is without having full logs, but I believe on Hetzner there's an issue with some IPs being assigned relatively late into the boot...