Important cluster concepts

This topic introduces the leader and follower nodes and the Cluster Coordinator.

In a cluster composed of nodes, each of which runs an Oracle Endeca Server that is hosting a specific data store powered by the Dgraph process, the following definitions are used:

Leader node

The leader node is a single Dgraph process in the cluster responsible for processing queries and for receiving updates to the data files and to the configuration. This node is responsible for obtaining information about the latest versions of the data files and propagating this information to the follower nodes through the Cluster Coordinator.

When you create a new cluster, you start the leader node first and then add follower nodes. The leader node has the following characteristics:
  • Each cluster must have one and only one leader node.
  • The leader node must have write access to the same shared file system on which the data files are stored and to which all follower nodes also have access.
  • The Cluster Coordinator service must be running on the leader node.
  • The entities outside the cluster (such as connectors in Integrator and components of Studio) must be able to access the Oracle Endeca Server that is hosting the Dgraph process for the leader node.

The leader node periodically receives full or incremental data files updates from Integrator. It also receives administration or configuration updates. It is the only node in the cluster that has access with write permissions to the on-disk representation of the index.

Once the leader node acquires access to the new version of the data files, it updates the data files, adding new information and deleting information that has become obsolete. The Cluster Coordinator notifies all follower nodes, alerting them to start using the updated version of the data files. The follower nodes acquire read-only access to an updated version of the data files.

Follower node

A follower node is a node in the cluster responsible for processing queries. The follower node does not update the data files, although it has read-only access to their latest copy.

You can start a follower node after you have started the leader node and the Cluster Coordinator. The follower node has the following characteristics:
  • Each cluster can have more than one follower node.
  • Each follower node must have a unique name across the cluster. The name also must be a valid directory name (characters such as slashes (/) are not allowed).
  • All follower nodes must reference the host name and port of the Cluster Coordinator service.
  • All follower nodes must have read-only access to the same shared file system on which the data files are stored. (The leader node must have write access to the file system.)

During the process of acquiring access to the recently updated data files, both the follower and the leader nodes continue to serve queries. Each query is processed against a specific version of the data files that is available to a cluster node at any given time. Query processing performance may slow down as the follower nodes acquire read-only access to the updated data files.

Cluster Coordinator

An entity that provides a mechanism for the nodes to communicate with each other.

Related information
Cluster behavior