No servers available
Tip
If you see server location disabled (resource_unavailable, *) in your cluster events or machine conditions, Hetzner Cloud has run out of capacity in that location.
Hetzner Cloud occasionally runs out of server capacity in specific locations. When that happens, Syself cannot create new cloud servers there. This blocks upgrades and self-healing: no replacement servers can come up to replace old or failed nodes.
Your cluster can end up in a degraded state, for example with only two control planes active instead of three, until Hetzner capacity returns. If capacity does not return quickly, you need a permanent fix instead of waiting. The two options below move nodes off cloud capacity.
Option 1: Move control planes to bare metal#
The permanent fix is to run control planes on Hetzner bare-metal (Robot) servers. Bare-metal servers are pre-registered by you and are always available, so cloud capacity limits do not affect them.
To migrate control planes to bare metal, see Bare-metal control planes .
Note
If you are on an older cluster-stack version that does not support bare-metal control planes, do the stack upgrade and the migration to bare metal in a single step. Running the upgrade first with cloud control planes triggers a rollout that cannot complete if cloud capacity is still exhausted.
Option 2: Move worker nodes to bare metal#
You can also move worker nodes to bare-metal servers. For workloads that run continuously, bare-metal servers are often a better fit: more CPU and memory per euro, and no cloud capacity limits. See Bare metal and cloud for the trade-offs.
See Add bare-metal servers for the steps.