Clusters
If your workload has grown beyond what a single machine can reliably handle, or you need cross-node scheduling, failover, and distributed storage, what you usually need is not a single server — it’s a cluster.
A cluster is a runtime environment composed of multiple Kubernetes nodes, designed for service architectures that require high availability, horizontal scaling, and distributed storage.
Why You Need a Cluster
A server is a good fit when you run all services on a single node: the architecture is simple, costs are predictable, and operational overhead is low. But when your system starts requiring the following capabilities, a cluster is the better choice:
- Multi-node scheduling: Spread different services and replicas across multiple machines, avoiding all workloads being concentrated on a single host.
- Node failure tolerance: When a single node goes offline, workloads can be rescheduled to other healthy nodes, reducing the risk of a full outage.
- Distributed storage: When you need to mount and migrate persistent data across nodes, distributed volumes are a better fit than local disks on a single machine.
- Greater scaling flexibility: You can expand capacity by adding nodes, rather than being limited to vertically upgrading a single host.
In short, a server solves “exclusive resources on one machine”, while a cluster solves “multi-machine coordination and high availability”.
Server vs. Cluster
| Feature | Server | Cluster |
|---|---|---|
| Number of nodes | Single node | Multiple Kubernetes nodes |
| Compute model | Single-machine deployment | Cross-node scheduling |
| Storage | Primarily single-machine local volumes | Distributed volumes that can follow workloads across nodes |
| Failure impact | Services on the affected machine are impacted | Workloads can be rescheduled to other nodes, reducing single-point-of-failure impact |
| Scaling | Primarily vertical (upgrading the single host) | Horizontal (adding nodes) |
| Best for | Small-to-medium services, monolithic apps, stable workloads | High-availability requirements, multi-replica services, distributed systems |
If you only need fixed resources and a single data node, a server is usually sufficient. If you want services to automatically failover when a node goes down, or you need a true multi-node architecture, choose a cluster.
Getting Started
- To create a new cluster directly through Zeabur, see Purchase a Cluster.
- If you already have your own Kubernetes cluster, see Connect an Existing Cluster.
- To learn about the single-machine option, see Servers.