Single-Node Clusters
Harvester supports single-node clusters for implementations that can tolerate lower resilience or require minimal initial deployment resources. You can create single-node clusters using the standard installation methods (ISO, USB, and PXE boot).
Single-node clusters support most Harvester features, including the creation of RKE2 clusters and node upgrades (with some limitations). However, this deployment type has the following key disadvantages:
- No high availability: Errors and updates that require rebooting of the node cause downtime to running VMs.
- No multi-replica support: Only one replica is created for each volume in Longhorn.
- No live migration and zero-downtime support during upgrades.
Prerequisites
Before you begin deploying your single-node cluster, ensure that the following requirements are met.
- Hardware: Use server-class hardware with sufficient resources to run Harvester and a production workload. Laptops and nested virtualization are not supported.
- Network: Configure ports based on the type of traffic to be transmitted among VMs.
Replica Count of the Default StorageClass
Harvester uses StorageClasses to describe how Longhorn must provision volumes. Each StorageClass has a parameter that defines the number of replicas to be created for each volume.
The default StorageClass harvester-longhorn
has a replica count value of 3 for high availability. If you use harvester-longhorn
in your single-node cluster, Longhorn is unable to create the default number of replicas, and volumes are marked as Degraded on the Volumes screen of the Harvester UI.
To avoid this issue, you can perform either of the following actions:
Change the replica count of
harvester-longhorn
to 1 using a Harvester configuration file.Create a new StorageClass with the Number of Replicas parameter set to 1. Once created, locate the new StorageClass in the list and then select ⋮ > Set as Default.
Upgrades and Maintenance
Single-node clusters do not support Live Migration, so VMs become unavailable during cluster upgrades. Harvester forcibly shuts down all VMs before starting the upgrade process.
Enabling Maintenance Mode is also not possible because that operation relies on Live Migration functionality, and Harvester cannot place the only control plane in Maintenance Mode.