Resizing an OpenSearch Cluster

Resize an OpenSearch cluster to change the node configuration to meet performance and other requirements.

The two ways you can resize an OpenSearch cluster are vertical or horizontal.

Horizontal resizing, which is when you adjust the number of data nodes, leader nodes, and OpenSearch Dashboard nodes, only supports increasing these settings for a cluster, you can't decrease them.

Vertical resizing, which is when you adjust the memory and OCPU resources for a cluster's leader nodes, data nodes, and OpenSearch Dashboard nodes, supports increasing or decreasing these settings.


When you resize a cluster, the cluster's search functionality is in read-only mode until the resize operation completes, see Cluster Patching and Resize Operations. This applies to all vertical and horizontal resize operations, except the operation to increase the data node storage. When you increase the storage for a data node during the resize operation, the search functionality stays in read/write mode. You can increase the data node storage setting only during a vertical resize operation in which no other settings are being adjusted.

Limitations and Considerations

Depending on whether you're scaling up (horizontal or vertical resizing) or scaling down (vertical resizing only), keep the following items in mind when resizing OpenSearch clusters.

Considerations with Scaling Up Cluster Settings

  • When resizing to improve availability and resiliency, use Horizontally Resizing an OpenSearch Cluster.

  • When resizing to improve performance, use Vertically Resizing an OpenSearch Cluster.

  • When you horizontally resize an OpenSearch cluster, you might also want to vertically resize the cluster.

  • We recommend that you vertically resize an OpenSearch cluster before you horizontally resize the cluster.

  • The more data nodes that you configure for an OpenSearch cluster, the more of an impact it has on the cluster's performance.

  • During the resize process, an OpenSearch cluster's performance might be impacted.

Considerations with Scaling Down Cluster Settings

  • Decreasing the memory or CPU for a cluster's nodes can impact a cluster's performance. Before you vertically downscale a cluster, use the available Search with OpenSearch metrics to monitor a cluster's performance to help you decide about the settings you can adjust.

  • You can't decrease data node storage, you can only increase data node storage for a cluster.

  • You can't decrease node configuration settings below the minimum configuration values.

    The minimum memory configuration required for cluster nodes are as follows:
    • Data node: 20 GB
    • Leader node: 20 GB
    • OpenSearch Dashboard node: 8 GB
    The minimum OCPU count required for cluster nodes are as follows:
    • Data node: 4
    • Leader node: 1
    • OpenSearch Dashboard node: 1
  • When you downscale a cluster, before Search with OpenSearch resizes the cluster, the service performs a check. Using the new configuration values, Search with OpenSearch calculates whether the updated configuration impacts performance outside of target CPU utilization or memory utilization, based on the cluster's recent metrics history. If this check doesn't pass, the vertical resize operation fails with an Invalidparameter error code and a cannot downscale error message.

  • When you decrease any node OCPU or memory settings, ensure that you have used enough metrics data to decide on the appropriate values to decrease to. Use the metrics data to come up with specific values for downscaling, we don't recommend testing out various configuration settings without understanding the potential performance impact of those settings. If you're uncertain about this process, you can contact support for recommendations.