Hibernation
Hibernation puts a whole cluster to sleep without deleting it. Configuration, workload definitions, and cluster state survive; compute spend stops. It’s a first-class action on every cluster — press Hibernate, later press Wake.
What hibernating does
Section titled “What hibernating does”- Managed (on-demand) nodes are deleted. They’re cattle; Hetzner stops billing them the moment they’re gone.
- Pooled servers are released back to your team’s fleet — never deleted, price intact. While the cluster sleeps they’re free for other uses (including other clusters).
- The control plane scales to zero. API server, scheduler, controllers stop; your cluster’s etcd state is retained on our platform, so everything you deployed is still defined, just not running.
Hibernating typically takes a couple of minutes; waking, a few minutes more. Both are visible live in the console.
What it costs while asleep
Section titled “What it costs while asleep”- Hourly (PAYG) plan: €0. The meter counts awake hours only — a hibernated cluster meters nothing. This is the whole point of the hourly plan; see pricing.
- Monthly and annual plans: the flat fee keeps running — the plan reserves your control plane’s slot. If your cluster sleeps most of the time, the hourly plan is the cheaper fit.
- Your Hetzner bill drops with the managed nodes deleted. Two honest exceptions:
- Volumes persist and keep billing at Hetzner until you delete them — hibernation never touches your data.
- Pooled servers keep billing at their (grandfathered) rate — you own them; hibernation just returns them to your fleet.
Waking up
Section titled “Waking up”Waking reverses the process: the control plane starts from its retained state, managed nodes are re-ordered, pooled servers are re-claimed from your fleet, and workloads reschedule. Two things to know:
- Managed nodes come back at current Hetzner prices — they’re new orders. Pooled servers are unaffected; their rate is protected. (This is one more reason to adopt the servers whose price matters.)
- If something else claimed your pooled servers while the cluster slept, the pool refills up to whatever is Ready in the fleet.
On the hourly plan, each wake starts a new active period with a 4-hour minimum — five ten-minute wake-ups bill twenty hours, one afternoon of work bills the afternoon. Steady on/off patterns (nights and weekends off) are where PAYG plus hibernation shines: a dev cluster awake 50 hours a month costs a few euros of PaaSbox fee plus a fraction of the Hetzner compute.
What survives, exactly
Section titled “What survives, exactly”| Survives hibernation? | |
|---|---|
| Cluster configuration, node pool definitions | ✓ |
| Deployed workloads, Services, ConfigMaps, Secrets (etcd state) | ✓ |
| PersistentVolumes on Hetzner volumes | ✓ — and keep billing at Hetzner |
| Adopted (pooled) servers | ✓ — released to your fleet, re-claimed on wake |
| Running pods / in-memory state | ✗ — pods reschedule on wake |
| Managed nodes (the machines themselves) | ✗ — recreated on wake at current prices |
When to use it
Section titled “When to use it”- Dev and staging clusters outside working hours — the classic case.
- Spiky or seasonal workloads on the hourly plan — wake for the burst, sleep after.
- Keeping a configured cluster on standby without paying for its compute.
For billing edge cases — spend caps hibernating clusters at the limit, dunning behavior — see billing & spend controls.