Skip to main content

OpenSearch Node Groups

An OpenSearch Managed Database cluster consists of node groups into which nodes are organized based on their role:

Node group with the Manager role

The node group with the Manager role is responsible for:

  • cluster state management;
  • data and task distribution;
  • fault tolerance;
  • load balancing;
  • management operation processing;
  • master node election;
  • monitoring and metrics collection.

This is a required node group that includes three nodes. This node group cannot be modified or deleted.

Node group with the Data role

A node group with the Data role is a cluster component that performs tasks related to indexing, sharding, replication, and request processing. For example, a node group with the Data role is responsible for:

  • data storage;
  • request processing;
  • data replication;
  • load balancing;
  • ensuring fault tolerance;
  • index optimization;
  • horizontal scaling support;
  • interaction with master nodes.

An OpenSearch cluster must have at least one node group with the Data role. You can add one more node group, but their total number in the cluster cannot exceed two. Each node group can contain from one to ten nodes. Node groups with the Data role cannot be deleted.

To avoid overload and performance degradation, we recommend monitoring CPU, RAM, and disk space usage.

Node group with the Dashboard role

The node group with the Dashboard role hosts a data visualization tool — OpenSearch Dashboards. Nodes in this group are optimized for processing user requests and performing tasks related to analytics and visualization. The node group with the Dashboard role is responsible for:

  • data visualization;
  • request processing;
  • dashboard management;
  • cluster health monitoring visualization;
  • analytical request processing;
  • backup operations;
  • user management and access definition.

This is an optional node group that includes only one node. A cluster can have only one node group with the Dashboard role. This node group can be deleted.

Dashboards are not involved in data processing or sharding management, so a group of nodes with the Dashboard role can be separated from the main cluster load — for example, you can configure a dashboard yourself and connect it to an OpenSearch cluster.