The following diagram illustrates a clustered Endeca Server
deployment serving multiple data domain clusters.
In this diagram,
starting from the top, the following components of the Endeca Server cluster
deployment are included:
- The external load balancer
serves as the single point of entry to the Endeca Server cluster. All requests
to all hosted data domain clusters go through this load balancer.
- The Endeca Server cluster
comprises a set of four Endeca Server nodes. All Endeca Server nodes in the
cluster are aware of one another.
- Three out of four Endeca
Server nodes are hosting the Cluster Coordinator instances. The Cluster
Coordinator maintains a cluster state for all participating members of the
Endeca Server cluster.
- Three data domain clusters
are hosted on the Endeca Server nodes. (Depending on the size of the Endeca
Server cluster, multiple data domain clusters can be hosted.)
- Starting from the left:
- The first data domain
cluster consists of two Dgraph nodes, one of which is the leader node. This
data domain cluster uses "Index 1" stored on a shared file system.
- The second data domain
cluster consists of one Dgraph node that serves both as the leader and follower
node; this data domain cluster uses "Index 2" stored on a shared file system.
- Finally, the third data
domain cluster has two Dgraph nodes, with one leader node and one follower
node; this data domain cluster uses "Index 3", which is also stored on a shared
file system.
- The following
characteristics are shared by all data domain clusters in the Endeca Server
cluster and are also reflected in this diagram:
- Each Dgraph node is
hosted on a different Endeca Server node — the allocation of Dgraph nodes to
the Endeca Server nodes occurs automatically when you add data domains.
- In each data domain
cluster that is not read-only, exactly one node serves as the leader node, and
the remaining nodes are follower nodes. (In a read-only data domain cluster,
all Dgraph nodes serve as follower nodes.)
- All Endeca Server nodes
hosting all data domain nodes have access to a shared file system on which the
indices for each data domain are stored. The cluster state is also maintained
in the shared file system by the Cluster Coordinator.