You may be affected by unavailability of Hetzner cloud servers, which is an increasingly common issue. This can halt upgrade and self-healing processes, as no new servers can be provisioned to replace the old nodes. As a result, you may end with clusters in a degraded state, such as having only two control planes, until there is availability again.

The only way to completely avoid this is moving to Hetzner bare metal servers. It is advised to do so at least for your control planes, following the process described here . This will also avoid getting your cluster in a degraded state, for example with only two control planes, until Hetzner capacity returns.

You can also swap your worker nodes to bare metal servers, as described here . Bare metal servers offer increased performance, a better network stack, and more cost-efficiency for your clusters.