Distinct Counts and Percentiles in Elasticsearch

This applies to: Visual Data Discovery

Distinct count and percentiles metrics return approximate values in Elasticsearch. The precision of the result returned by distinct count metric depends on the precision threshold setting (default value is 1000).

You can change the value of the precision threshold by setting the elasticsearch.query.cardinality.precision.threshold property in the zoomdata.properties file.

See Elasticsearch's documentation on the following for more information:

• For Elasticsearch version 7, see the following for percentiles and distinct count.
• For Elasticsearch version 8, see the following for percentiles and distinct count.

The table below lists all available properties that you can modify to work with Elasticsearch.

Property	Default	Use	Notes
elasticsearch.query.cardinality.precision.threshold	1000	control the level of accuracy of the distinct counts	The maximum supported value is 40000. However, Symphony does not recommend to set such value as it may result in performance issues and the data source itself may return errors. For more info, refer to the Precision Control section by Elasticsearch.
elasticsearch.query.limit.nongrouped	10000	set the limit for the number of non-grouped records (per shard) to execute on.
elasticsearch.query.limit.grouped	10000	set the limit for the number of grouped records (per shard) to execute on.

If you need to change the default settings, you can add the corresponding properties (listed above) to the zoomdata.properties file and assign the required values. For more details about working with the zoomdata.properties file, refer to the topic Managing Configurations in Symphony .