Life inside a cluster

Terminology

Term	Description
Node	Running instance of EES
Cluster	>= 1 node
Master node	Single elected node in the cluster responsible for cluster wide changes (+/- index, +/- node)
Shard	Containers of data (stores documents) Allocated to node
Primary Shard	Document store
Replica Shard	Replica of primary shard

Adding failover

• Avoid single point of failure by running multiple nodes

Horizontal scaling

• Adding one node

• Increasing numebr of replicas

Coping with failure

Scenario: One node killed

• Actions:
  • One node promtoed to master (node 2)
  • Replica shards promoted to primary shard
  • Health check yellow (insufficient replicas)