1. Adding Nodes for Development

1-1. Duplicate ElasticSearch

1. duplicate elasticsearch in new directory
2. change and uncomment node.name to give different name than the default

// elasticsearch.yml
node.name: node-2

3. run another elasticsearch instance

• ElasticSearch automatically puts new node into the cluster

ex) running 2 nodes. now the status is green as replica shard can be allocated

{
  "cluster_name" : "elasticsearch",
  "status" : "green",
  "timed_out" : false,
  "number_of_nodes" : 2,
  "number_of_data_nodes" : 2,
  "active_primary_shards" : 7,
  "active_shards" : 7,
  "relocating_shards" : 0,
  "initializing_shards" : 0,
  "unassigned_shards" : 0,
  "delayed_unassigned_shards" : 0,
  "number_of_pending_tasks" : 0,
  "number_of_in_flight_fetch" : 0,
  "task_max_waiting_in_queue_millis" : 0,
  "active_shards_percent_as_number" : 100.0
}

1-2. Run ElasticSearch with option

• rather than duplicating elasticsearch, new node can be run by running

$ bin/elasticsearch -Enode.name={node name} -Epath.data={data dir} -Epath.logs={log dir}

-E option overwrites configuration

path.data and path.logs must be specified differently, else error will occur

• because it is better to manage configs in elasticsearch directory, 1-1 method is a better method

2. Node roles

• Node can have more roles than just storing shards
• GET /_cat/nodes?v to see node.role and master
• default is dim(data, ingest, master)

2-1. master node

• perform cluster wide actions ex) create/delete indexes, allocate shard to nodes, etc.

• elected by voting system

• stable master node is import for stable cluster

• dedicated master node will be needed if elected master node is busy

2-2. data node

• store part of cluster data
• performs queries
• purpose of having dedicated data node is to separate it from master node

2-3. ingest node

• enable node to run ingest pipelines
• ingest pipelines : series of steps that are performed when indexing documents into elasticsearch
• a simplified version of Logstash within Elasticsearch

2-4. machine learning node

• node.ml and xpack.ml.enabled can set a node for machine learning

2-5. coordination node

• distribute queries and aggregation of results
• can set a node to coordination node by disabling all other roles
• only useful for large cluster as load balancer

2-6. When to change node roles

• typically when optimizing cluster to scale the number of requests
• for better understanding what hardware resources are used for
• depends on the situation