Skip to main content

🗓️ 06112024 2344

ELASTICSEARCH CLUSTER

Terminology​

TermDescription
NodeRunning instance of EES
Cluster>= 1 node
Master nodeSingle elected node in the cluster responsible for cluster wide changes (+/- index, +/- node)
Shard
  • Containers of data (stores documents)
  • Allocated to node
Primary ShardDocument store
Replica ShardReplica of primary shard

chapter2_cluster.png

Adding failover​

  • Avoid single point of failure by running multiple nodes

Horizontal scaling​

chapter2_hscale.png

  • Adding one node

chapter2_hscale_replica.png

  • Increasing number of replicas

Coping with failure​

WARNING

Scenario: One node killed

chapter2_failure.png

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

References​

  • Chapter 2 of ElasticSearch Oreilly book