Cluster service considerations
A cluster requires at least one master, called the primary master, which provides the cluster services. For failover purposes in a cluster with multiple nodes, we can configure up to three more masters in the environment. The required number of masters depends on which services you use and your failover requirements. The following table depicts the valid master configurations.
| Number of masters | Combination of masters | Considerations |
|---|---|---|
| 1 | Primary master only. | No failover for cluster services. |
| 2 | Primary master and secondary master. | This configuration includes a secondary master to provide failover for the cluster services, which include the distributed session cache (DSC), configuration database, geolocation database, and runtime database. |
| 3 | Primary master, secondary master, and tertiary master. | We can optionally designate a tertiary master to provide
extra failover for the distributed session cache.
Only the distributed session cache recognizes the tertiary master node. The configuration, geolocation, and runtime databases consider the tertiary node as a non-master node. |
| 4 | Primary master, secondary master, tertiary master, and quaternary master. | We can optionally designate tertiary and quaternary
masters to provide extra failover for the distributed session cache.
Only the distributed session cache recognizes the tertiary and quaternary master nodes. The configuration, geolocation, and runtime databases consider these nodes as non-master nodes. |
For high availability in a cross data center environment, we can
consider separating the master appliances between the data centers
as depicted in Figure 1.
- Distributed session cache
- The primary master maintains the master copy of the distributed session cache and the other master nodes keep slave copies for failover purposes.
- Runtime database
- If you are using the internal runtime database, the primary master
maintains the master copy of this data, while the secondary master
keeps a slave copy for failover purposes.
If you are using an external runtime database, the cluster does not provide failover. In this case, the external database server is responsible for ensuring high availability.
- Configuration and geolocation databases
- The primary master is the only master on which we can update the configuration and geolocation databases. The other nodes in the cluster, including secondary, tertiary, and quaternary masters, maintain a read-only copy of the information from these databases.
Parent topic: High availability of cluster services