This page describes your high availability (HA) options in GKE On-Prem.
HA for admin clusters
HA for the admin cluster control plane is currently not supported. To prevent a single point of failure in the admin cluster, set up your vSphere cluster to enable vSphere High Availability.
vSphere High Availability protects GKE On-Prem admin clusters from going down in the event of an underlying host failure.
To learn more, see Create a vSphere HA Cluster.
HA for user clusters
Figure: GKE On-Prem architecture with highly-available user clusters.
Multi-control plane deployments for user control planes are supported and can be
configured through the configuration file's
usercluster.masternode.replicas field. The field supports values 1 or 3.
To enable HA, choose 3 control plane replicas. With HA enabled, three
control plane VMs are deployed within one vSphere cluster. Each control plane
component is replicated with leader election enabled where necessary. Control
plane replicas run with Pod-level anti-affinity, causing them to spread across
the three control plane VMs.
During this feature's beta phase, anti-affinity for the VMs themselves to run across hosts is not enabled. You can expect this feature in a future release. Until that time, you should enable vSphere High Availability.