Installing and Configuring Elasticsearch¶

Installing from a download package¶

Links to various formats and also older versions of Elasticsearch can be found on the download page [4]. Once downloaded and extracted, you can start Elasticsearch with:

bin/elasticsearch

For more details see the installation instructions [5]

Quick command line example with 5.6.11:

Download the ES package:

cd ~
wget https://artifacts.elastic.co/downloads/elasticsearch/elasticsearch-5.6.11.deb
sudo dpkg -i elasticsearch-5.6.11.deb
sudo update-rc.d elasticsearch defaults 95 10
sudo /etc/init.d/elasticsearch start

Indices¶

Elasticsearch (and Lucene) store information in indices. Within an index can be one or more document types. Searchlight’s indexing service uses an index per service that has a plugin available, and each plugin generally will have its own document type. For instance, the glance plugin has glance.image and glance.metadef. Since the volume of data is lower than a typical use case for Elasticsearch it may make sense to change the default sharing and replication mechanism. We also recommend disabling implicit index creation, though if you are sharing an Elasticsearch installation this may be inadvisable. The following options control indexing behavior:

# Number of shards for each index (performance)
index.number_of_shards: 5
# Number of replicas per shard (redundancy and recovering)
index.number_of_replicas: 1

# Disable automatic index creation so that index creation
# is an explicit action
action.auto_create_index: false

[elasticsearch]
index_settings = refresh_interval:2s,number_of_replicas:1

Scripts¶

The scripting module allows to use scripts in order to evaluate custom expressions. Scripting is turned off by default in elasticsearch latest versions. Searchlight doesn’t allow scripts in the search api but requires scripts to sync Index updates from notifications. For security purpose index updates are allowed only for admin role.

script.engine.groovy.inline.update: on

See also [8].

Development¶

For development, Elasticsearch’s default configuration is overkill. It’s possible to run Elasticsearch with a much lower memory footprint than by default, and you may wish to disable clustering behavior.

# Configures elasticsearch as a single node (no discovery)
node.local: true

# Disable sharding and replication
index.number_of_shards: 1
index.number_of_replicas: 0

export ES_HEAP_SIZE=40m

Production¶

Some settings you may wish to change for production:

# Cluster name is used by cluster discovery; it's important to ensure
# this is set across all nodes you wish to be in the cluster
cluster.name: searchlight

# By default elasticsearch picks a random name from a list of Marvel
# comic characters. If you specify this, make sure it's different on
# each node in the cluster
node.name: This Node Name

# Bind to a non-standard address
network.host: 0.0.0.0

# Bind to a non-standard port
http.port: 9200

# Configure the default data and log directories. By default, these
# directories will be created in $ES_HOME.
path:
  logs: /var/log/elasticsearch
  data: /var/data/elasticsearch

# This setting locks the Elasticsearch process address space into RAM
# (preventing locking). If you set this, ensure that you've configured
# ES_HEAP_SIZE appropriately (see below). Linux only.
bootstrap.mlockall: true

For more details see Elasticsearch’s configuration information [6].

discovery.zen.ping.multicast.enabled: false
discovery.zen.ping.unicast.hosts: ['w.x.y.z', 'w.x.y.z']

export ES_HEAP_SIZE=2g