Nodeedit

Any time that you start an instance of Elasticsearch, you are starting a node. A collection of connected nodes is called a cluster. If you are running a single node of Elasticsearch, then you have a cluster of one node.

Every node in the cluster can handle HTTP and transport traffic by default. The transport layer is used exclusively for communication between nodes; the HTTP layer is used by REST clients.

All nodes know about all the other nodes in the cluster and can forward client requests to the appropriate node.

Node rolesedit

You define a node’s roles by setting node.roles in elasticsearch.yml. If you set node.roles, the node is only assigned the roles you specify. If you don’t set node.roles, the node is assigned the following roles:

  • master
  • data
  • data_content
  • data_hot
  • data_warm
  • data_cold
  • data_frozen
  • ingest
  • ml
  • remote_cluster_client
  • transform

If you set node.roles, ensure you specify every node role your cluster needs. Every cluster requires the following node roles:

  • master
  • data_content and data_hot
    OR
    data

Some Elastic Stack features also require specific node roles:

  • Cross-cluster search and cross-cluster replication require the remote_cluster_client role.
  • Stack Monitoring and ingest pipelines require the ingest role.
  • Fleet, the Elastic Security app, and transforms require the transform role. The remote_cluster_client role is also required to use cross-cluster search with these features.
  • Machine learning features, such as anomaly detection, require the ml role.

As the cluster grows and in particular if you have large machine learning jobs or continuous transforms, consider separating dedicated master-eligible nodes from dedicated data nodes, machine learning nodes, and transform nodes.

Master-eligible node
A node that has the master role, which makes it eligible to be elected as the master node, which controls the cluster.
Data node
A node that has the data role. Data nodes hold data and perform data related operations such as CRUD, search, and aggregations. A node with the data role can fill any of the specialised data node roles.
Ingest node
A node that has the ingest role. Ingest nodes are able to apply an ingest pipeline to a document in order to transform and enrich the document before indexing. With a heavy ingest load, it makes sense to use dedicated ingest nodes and to not include the ingest role from nodes that have the master or data roles.
Remote-eligible node
A node that has the remote_cluster_client role, which makes it eligible to act as a remote client.
Machine learning node
A node that has the ml role. If you want to use machine learning features, there must be at least one machine learning node in your cluster. For more information, see Machine learning settings and Machine learning in the Elastic Stack.
Transform node
A node that has the transform role. If you want to use transforms, there must be at least one transform node in your cluster. For more information, see Transforms settings and Transforming data.

Coordinating node

Requests like search requests or bulk-indexing requests may involve data held on different data nodes. A search request, for example, is executed in two phases which are coordinated by the node which receives the client request — the coordinating node.

In the scatter phase, the coordinating node forwards the request to the data nodes which hold the data. Each data node executes the request locally and returns its results to the coordinating node. In the gather phase, the coordinating node reduces each data node’s results into a single global result set.

Every node is implicitly a coordinating node. This means that a node that has an explicit empty list of roles via node.roles will only act as a coordinating node, which cannot be disabled. As a result, such a node needs to have enough memory and CPU in order to deal with the gather phase.

Master-eligible nodeedit

The master node is responsible for lightweight cluster-wide actions such as creating or deleting an index, tracking which nodes are part of the cluster, and deciding which shards to allocate to which nodes. It is important for cluster health to have a stable master node.

Any master-eligible node that is not a voting-only node may be elected to become the master node by the master election process.

Master nodes must have a path.data directory whose contents persist across restarts, just like data nodes, because this is where the cluster metadata is stored. The cluster metadata describes how to read the data stored on the data nodes, so if it is lost then the data stored on the data nodes cannot be read.

Dedicated master-eligible nodeedit

It is important for the health of the cluster that the elected master node has the resources it needs to fulfill its responsibilities. If the elected master node is overloaded with other tasks then the cluster will not operate well. The most reliable way to avoid overloading the master with other tasks is to configure all the master-eligible nodes to be dedicated master-eligible nodes which only have the master role, allowing them to focus on managing the cluster. Master-eligible nodes will still also behave as coordinating nodes that route requests from clients to the other nodes in the cluster, but you should not use dedicated master nodes for this purpose.

A small or lightly-loaded cluster may operate well if its master-eligible nodes have other roles and responsibilities, but once your cluster comprises more than a handful of nodes it usually makes sense to use dedicated master-eligible nodes.

To create a dedicated master-eligible node, set:

node.roles: [ master ]

Voting-only master-eligible nodeedit

A voting-only master-eligible node is a node that participates in master elections but which will not act as the cluster’s elected master node. In particular, a voting-only node can serve as a tiebreaker in elections.

It may seem confusing to use the term "master-eligible" to describe a voting-only node since such a node is not actually eligible to become the master at all. This terminology is an unfortunate consequence of history: master-eligible nodes are those nodes that participate in elections and perform certain tasks during cluster state publications, and voting-only nodes have the same responsibilities even if they can never become the elected master.

To configure a master-eligible node as a voting-only node, include master and voting_only in the list of roles. For example to create a voting-only data node:

node.roles: [ data, master, voting_only ]

Only nodes with the master role can be marked as having the voting_only role.

High availability (HA) clusters require at least three master-eligible nodes, at least two of which are not voting-only nodes. Such a cluster will be able to elect a master node even if one of the nodes fails.

Voting-only master-eligible nodes may also fill other roles in your cluster. For instance, a node may be both a data node and a voting-only master-eligible node. A dedicated voting-only master-eligible nodes is a voting-only master-eligible node that fills no other roles in the cluster. To create a dedicated voting-only master-eligible node, set:

node.roles: [ master, voting_only ]

Since dedicated voting-only nodes never act as the cluster’s elected master, they may require less heap and a less powerful CPU than the true master nodes. However all master-eligible nodes, including voting-only nodes, are on the critical path for publishing cluster state updates. Cluster state updates are usually independent of performance-critical workloads such as indexing or searches, but they are involved in management activities such as index creation and rollover, mapping updates, and recovery after a failure. The performance characteristics of these activities are a function of the speed of the storage on each master-eligible node, as well as the reliability and latency of the network interconnections between the elected master node and the other nodes in the cluster. You must therefore ensure that the storage and networking available to the nodes in your cluster are good enough to meet your performance goals.

Data nodeedit

Data nodes hold the shards that contain the documents you have indexed. Data nodes handle data related operations like CRUD, search, and aggregations. These operations are I/O-, memory-, and CPU-intensive. It is important to monitor these resources and to add more data nodes if they are overloaded.

The main benefit of having dedicated data nodes is the separation of the master and data roles.

To create a dedicated data node, set:

node.roles: [ data ]

In a multi-tier deployment architecture, you use specialized data roles to assign data nodes to specific tiers: data_content,data_hot, data_warm, data_cold, or data_frozen. A node can belong to multiple tiers, but a node that has one of the specialized data roles cannot have the generic data role.

Content data nodeedit

Content data nodes are part of the content tier. Data stored in the content tier is generally a collection of items such as a product catalog or article archive. Unlike time series data, the value of the content remains relatively constant over time, so it doesn’t make sense to move it to a tier with different performance characteristics as it ages. Content data typically has long data retention requirements, and you want to be able to retrieve items quickly regardless of how old they are.

Content tier nodes are usually optimized for query performance—​they prioritize processing power over IO throughput so they can process complex searches and aggregations and return results quickly. While they are also responsible for indexing, content data is generally not ingested at as high a rate as time series data such as logs and metrics. From a resiliency perspective the indices in this tier should be configured to use one or more replicas.

The content tier is required. System indices and other indices that aren’t part of a data stream are automatically allocated to the content tier.

To create a dedicated content node, set:

node.roles: [ data_content ]

Hot data nodeedit

Hot data nodes are part of the hot tier. The hot tier is the Elasticsearch entry point for time series data and holds your most-recent, most-frequently-searched time series data. Nodes in the hot tier need to be fast for both reads and writes, which requires more hardware resources and faster storage (SSDs). For resiliency, indices in the hot tier should be configured to use one or more replicas.

The hot tier is required. New indices that are part of a data stream are automatically allocated to the hot tier.

To create a dedicated hot node, set:

node.roles: [ data_hot ]

Warm data nodeedit

Warm data nodes are part of the warm tier. Time series data can move to the warm tier once it is being queried less frequently than the recently-indexed data in the hot tier. The warm tier typically holds data from recent weeks. Updates are still allowed, but likely infrequent. Nodes in the warm tier generally don’t need to be as fast as those in the hot tier. For resiliency, indices in the warm tier should be configured to use one or more replicas.

To create a dedicated warm node, set:

node.roles: [ data_warm ]

Cold data nodeedit

Cold data nodes are part of the cold tier. When you no longer need to search time series data regularly, it can move from the warm tier to the cold tier. While still searchable, this tier is typically optimized for lower storage costs rather than search speed.

For better storage savings, you can keep fully mounted indices of searchable snapshots on the cold tier. Unlike regular indices, these fully mounted indices don’t require replicas for reliability. In the event of a failure, they can recover data from the underlying snapshot instead. This potentially halves the local storage needed for the data. A snapshot repository is required to use fully mounted indices in the cold tier. Fully mounted indices are read-only.

Alternatively, you can use the cold tier to store regular indices with replicas instead of using searchable snapshots. This lets you store older data on less expensive hardware but doesn’t reduce required disk space compared to the warm tier.

To create a dedicated cold node, set:

node.roles: [ data_cold ]

Frozen data nodeedit

Frozen data nodes are part of the frozen tier. Once data is no longer being queried, or being queried rarely, it may move from the cold tier to the frozen tier where it stays for the rest of its life.

The frozen tier requires a snapshot repository. The frozen tier uses partially mounted indices to store and load data from a snapshot repository. This reduces local storage and operating costs while still letting you search frozen data. Because Elasticsearch must sometimes fetch frozen data from the snapshot repository, searches on the frozen tier are typically slower than on the cold tier.

To create a dedicated frozen node, set:

node.roles: [ data_frozen ]

Ingest nodeedit

Ingest nodes can execute pre-processing pipelines, composed of one or more ingest processors. Depending on the type of operations performed by the ingest processors and the required resources, it may make sense to have dedicated ingest nodes, that will only perform this specific task.

To create a dedicated ingest node, set:

node.roles: [ ingest ]

Coordinating only nodeedit

If you take away the ability to be able to handle master duties, to hold data, and pre-process documents, then you are left with a coordinating node that can only route requests, handle the search reduce phase, and distribute bulk indexing. Essentially, coordinating only nodes behave as smart load balancers.

Coordinating only nodes can benefit large clusters by offloading the coordinating node role from data and master-eligible nodes. They join the cluster and receive the full cluster state, like every other node, and they use the cluster state to route requests directly to the appropriate place(s).

Adding too many coordinating only nodes to a cluster can increase the burden on the entire cluster because the elected master node must await acknowledgement of cluster state updates from every node! The benefit of coordinating only nodes should not be overstated — data nodes can happily serve the same purpose.

To create a dedicated coordinating node, set:

node.roles: [ ]

Remote-eligible nodeedit

A remote-eligible node acts as a cross-cluster client and connects to remote clusters. Once connected, you can search remote clusters using cross-cluster search. You can also sync data between clusters using cross-cluster replication.

node.roles: [ remote_cluster_client ]

Machine learning nodeedit

Machine learning nodes run jobs and handle machine learning API requests. For more information, see Machine learning settings.

To create a dedicated machine learning node, set:

node.roles: [ ml, remote_cluster_client]

The remote_cluster_client role is optional but strongly recommended. Otherwise, cross-cluster search fails when used in machine learning jobs or datafeeds. If you use cross-cluster search in your anomaly detection jobs, the remote_cluster_client role is also required on all master-eligible nodes. Otherwise, the datafeed cannot start. See Remote-eligible node.

Transform nodeedit

Transform nodes run transforms and handle transform API requests. For more information, see Transforms settings.

To create a dedicated transform node, set:

node.roles: [ transform, remote_cluster_client ]

The remote_cluster_client role is optional but strongly recommended. Otherwise, cross-cluster search fails when used in transforms. See Remote-eligible node.

Changing the role of a nodeedit

Each data node maintains the following data on disk:

  • the shard data for every shard allocated to that node,
  • the index metadata corresponding with every shard allocated to that node, and
  • the cluster-wide metadata, such as settings and index templates.

Similarly, each master-eligible node maintains the following data on disk:

  • the index metadata for every index in the cluster, and
  • the cluster-wide metadata, such as settings and index templates.

Each node checks the contents of its data path at startup. If it discovers unexpected data then it will refuse to start. This is to avoid importing unwanted dangling indices which can lead to a red cluster health. To be more precise, nodes without the data role will refuse to start if they find any shard data on disk at startup, and nodes without both the master and data roles will refuse to start if they have any index metadata on disk at startup.

It is possible to change the roles of a node by adjusting its elasticsearch.yml file and restarting it. This is known as repurposing a node. In order to satisfy the checks for unexpected data described above, you must perform some extra steps to prepare a node for repurposing when starting the node without the data or master roles.

  • If you want to repurpose a data node by removing the data role then you should first use an allocation filter to safely migrate all the shard data onto other nodes in the cluster.
  • If you want to repurpose a node to have neither the data nor master roles then it is simplest to start a brand-new node with an empty data path and the desired roles. You may find it safest to use an allocation filter to migrate the shard data elsewhere in the cluster first.

If it is not possible to follow these extra steps then you may be able to use the elasticsearch-node repurpose tool to delete any excess data that prevents a node from starting.

Node data path settingsedit

path.dataedit

Every data and master-eligible node requires access to a data directory where shards and index and cluster metadata will be stored. The path.data defaults to $ES_HOME/data but can be configured in the elasticsearch.yml config file an absolute path or a path relative to $ES_HOME as follows:

path.data:  /var/elasticsearch/data

Like all node settings, it can also be specified on the command line as:

./bin/elasticsearch -Epath.data=/var/elasticsearch/data

The contents of the path.data directory must persist across restarts, because this is where your data is stored. Elasticsearch requires the filesystem to act as if it were backed by a local disk, but this means that it will work correctly on properly-configured remote block devices (e.g. a SAN) and remote filesystems (e.g. NFS) as long as the remote storage behaves no differently from local storage. You can run multiple Elasticsearch nodes on the same filesystem, but each Elasticsearch node must have its own data path.

The performance of an Elasticsearch cluster is often limited by the performance of the underlying storage, so you must ensure that your storage supports acceptable performance. Some remote storage performs very poorly, especially under the kind of load that Elasticsearch imposes, so make sure to benchmark your system carefully before committing to a particular storage architecture.

When using the .zip or .tar.gz distributions, the path.data setting should be configured to locate the data directory outside the Elasticsearch home directory, so that the home directory can be deleted without deleting your data! The RPM and Debian distributions do this for you already.

Don’t modify anything within the data directory or run processes that might interfere with its contents. If something other than Elasticsearch modifies the contents of the data directory, then Elasticsearch may fail, reporting corruption or other data inconsistencies, or may appear to work correctly having silently lost some of your data. Don’t attempt to take filesystem backups of the data directory; there is no supported way to restore such a backup. Instead, use Snapshot and restore to take backups safely. Don’t run virus scanners on the data directory. A virus scanner can prevent Elasticsearch from working correctly and may modify the contents of the data directory. The data directory contains no executables so a virus scan will only find false positives.

Other node settingsedit

More node settings can be found in Configuring Elasticsearch and Important Elasticsearch configuration, including: