elasticsearch api conventions


Elasticsearch is composed of a number of modules, which are responsible for its functionality. These modules have the following two types of settings −

  • Static Settings − These settings need to be configured in config (elasticsearch.yml) file before starting Elasticsearch. You need to update all the concern nodes in the cluster to reflect the changes by these settings.

  • Dynamic Settings − These settings can be set on live Elasticsearch.

We will discuss the different modules of Elasticsearch in the following sections of this chapter.

Cluster-Level Routing and Shard Allocation

Cluster level settings decide the allocation of shards to different nodes and reallocation of shards to rebalance cluster. These are the following settings to control shard allocation −

Cluster-Level Shard Allocation

Setting Possible value Description
cluster.routing.allocation.enable
all This default value allows shard allocation for all kinds of shards.
primaries This allows shard allocation only for primary shards.
new_primaries This allows shard allocation only for primary shards for new indices.
none This does not allow any shard allocations.
cluster.routing.allocation .node_concurrent_recoveries Numeric value (by default 2) This restricts the number of concurrent shard recovery.
cluster.routing.allocation .node_initial_primaries_recoveries Numeric value (by default 4) This restricts the number of parallel initial primary recoveries.
cluster.routing.allocation .same_shard.host Boolean value (by default false) This restricts the allocation of more than one replica of the same shard in the same physical node.
indices.recovery.concurrent _streams Numeric value (by default 3) This controls the number of open network streams per node at the time of shard recovery from peer shards.
indices.recovery.concurrent _small_file_streams Numeric value (by default 2) This controls the number of open streams per node for small files having size less than 5mb at the time of shard recovery.
cluster.routing.rebalance.enable
all This default value allows balancing for all kinds of shards.
primaries This allows shard balancing only for primary shards.
replicas This allows shard balancing only for replica shards.
none This does not allow any kind of shard balancing.
cluster.routing.allocation .allow_rebalance
always This default value always allows rebalancing.
indices_primaries _active This allows rebalancing when all primary shards in cluster are allocated.
Indices_all_active This allows rebalancing when all the primary and replica shards are allocated.
cluster.routing.allocation.cluster _concurrent_rebalance Numeric value (by default 2) This restricts the number of concurrent shard balancing in cluster.
cluster.routing.allocation .balance.shard Float value (by default 0.45f) This defines the weight factor for shards allocated on every node.
cluster.routing.allocation .balance.index Float value (by default 0.55f) This defines the ratio of the number of shards per index allocated on a specific node.
cluster.routing.allocation .balance.threshold Non negative float value (by default 1.0f) This is the minimum optimization value of operations that should be performed.

Disk-based Shard Allocation

Setting Possible value Description
cluster.routing.allocation .disk.threshold_enabled Boolean value (by default true) This enables and disables disk allocation decider.
cluster.routing.allocation .disk.watermark.low String value (by default 85%) This denotes maximum usage of disk; after this point, no other shard can be allocated to that disk.
cluster.routing.allocation .disk.watermark.high String value (by default 90%) This denotes the maximum usage at the time of allocation; if this point is reached at the time of allocation, then Elasticsearch will allocate that shard to another disk.
cluster.info.update.interval String value (by default 30s) This is the interval between disk usages checkups.
cluster.routing.allocation .disk.include_relocations Boolean value (by default true) This decides whether to consider the shards currently being allocated, while calculating disk usage.

Discovery

This module helps a cluster to discover and maintain the state of all the nodes in it. The state of cluster changes when a node is added or deleted from a cluster. The cluster name setting is used to create logical difference between different clusters. There are some modules which help you to use the APIs provided by cloud vendors and those are −

  • Azure discovery
  • EC2 discovery
  • Google compute engine discovery
  • Zen discovery

Gateway

This module maintains the cluster state and the shard data across full cluster restarts. Following are the static settings of this module −

Setting Possible value Description
gateway.expected_ nodes numeric value (by default 0) The number of nodes that are expected to be in the cluster for the recovery of local shards.
gateway.expected_ master_nodes numeric value (by default 0) The number of master nodes that are expected to be in the cluster before start recovery.
gateway.expected_ data_nodes numeric value (by default 0) The number of data nodes expected in the cluster before start recovery.
gateway.recover_ after_time String value (by default 5m) This specifies the time for which the recovery process will wait to start regardless of the number of nodes joined in the cluster.

  • gateway.recover_ after_nodes
  • gateway.recover_after_ master_nodes
  • gateway.recover_after_ data_nodes

HTTP

This module manages the communication between HTTP client and Elasticsearch APIs. This module can be disabled by changing the value of http.enabled to false. The following are the settings (configured in elasticsearch.yml) to control this module −

Sr.No Setting & Description
1

http.port

This is a port to access Elasticsearch and it ranges from 9200-9300.

2

http.publish_port

This port is for http clients and is also useful in case of firewall.

3

http.bind_host

This is a host address for http service.

4

http.publish_host

This is a host address for http client.

5

http.max_content_length

This is the maximum size of content in an http request. Its default value is 100mb.

6

http.max_initial_line_length

This is the maximum size of URL and its default value is 4kb.

7

http.max_header_size

This is the maximum http header size and its default value is 8kb.

8

http.compression

This enables or disables support for compression and its default value is false.

9

http.pipelinig

This enables or disables HTTP pipelining.

10

http.pipelining.max_events

This restricts the number of events to be queued before closing an HTTP request.

Indices

This module maintains the settings, which are set globally for every index. The following settings are mainly related to memory usage −

Circuit Breaker

This is used for preventing operation from causing an OutOfMemroyError. The setting mainly restricts the JVM heap size. For example, indices.breaker.total.limit setting, which defaults to 70% of JVM heap.

Fielddata Cache

This is used mainly when aggregating on a field. It is recommended to have enough memory to allocate it. The amount of memory used for the field data cache can be controlled using indices.fielddata.cache.size setting.

Node Query Cache

This memory is used for caching the query results. This cache uses Least Recently Used (LRU) eviction policy. Indices.queries.cahce.size setting controls the memory size of this cache.

Indexing Buffer

This buffer stores the newly created documents in the index and flushes them when the buffer is full. Setting like indices.memory.index_buffer_size control the amount of heap allocated for this buffer.

Shard Request Cache

This cache is used to store local search data for every shard. Cache can be enabled during the creation of index or can be disabled by sending URL parameter.

Disable cache - ?request_cache = true
Enable cache "index.requests.cache.enable": true

Indices Recovery

It controls the resources during recovery process. The following are the settings −

Setting Default value
indices.recovery.concurrent_streams 3
indices.recovery.concurrent_small_file_streams 2
indices.recovery.file_chunk_size 512kb
indices.recovery.translog_ops 1000
indices.recovery.translog_size 512kb
indices.recovery.compress true
indices.recovery.max_bytes_per_sec 40mb

TTL Interval

Time to Live (TTL) interval defines the time of a document, after which the document gets deleted. The following are the dynamic settings for controlling this process −

Setting Default value
indices.ttl.interval 60s
indices.ttl.bulk_size 1000

Node

Each node has an option to be data node or not. You can change this property by changing node.data setting. Setting the value as false defines that the node is not a data node.


Table of Contents
1.elasticsearch tutorial

2.elasticsearch basic concepts

3.elasticsearch installation

4.elasticsearch populate

5.elasticsearch migration between versions

6.elasticsearch api conventions

7.elasticsearch document apis

8.elasticsearch search apis

9.elasticsearch aggregations

10.elasticsearch index apis

11.elasticsearch cluster apis

12.elasticsearch query dsl

13.elasticsearch mapping

14.elasticsearch analysis

15.elasticsearch modules

16.elasticsearch testing

17.elasticsearch quick guide

18.elasticsearch useful resources

19.discuss elasticsearch


Leave a Reply

Your email address will not be published. Required fields are marked *