Use Databricks as a destination - Kameleoon Documentation

Once you have enabled the Databricks integration for your project, you can activate Use Databricks as a destination to send events to your Databricks account whenever one of your Kameleoon experiments exposes visitors.

Two tables. The first shows the data flow from Kameleoon to Databricks, which sends exposure events. The second marks web experiments, web personalizations and feature experiments as supported.

Activate Databricks as a destination

To activate Databricks as a destination:

Click Use Databricks as a destination.
Databricks catalog: name of the Databricks “catalog” that contains the Kameleoon schema to write to (the top level folder).
Schema: Schema where Kameleoon writes its event table, for example kameleoon_events.

Databricks integration settings with Use Databricks as a destination disabled, showing empty Databricks catalog and Schema fields.

What “Use Databricks as a destination” does

After you enable Use Databricks as a destination, Kameleoon writes every experiment exposure to Databricks. Kameleoon stores events in the kameleoon_events schema of your Databricks catalog. This catalog is the one you set up during setup, with write access granted to Kameleoon’s service principal. Kameleoon saves the data in a table named kameleoon_experiment_event and writes each exposure as one row. Your own Databricks settings control how long Databricks retains these rows. An exposure reaches your Databricks table in two steps: a web, mobile or server SDK sends the exposure payload to Kameleoon’s Ingestion API, and Kameleoon then writes it to Databricks with an authenticated insert.

Kameleoon doesn’t keep a copy of exposure events. It flushes each event to Databricks, so a failed write loses that event. Kameleoon doesn’t retry failed writes, and it doesn’t send notifications when a write fails. The SQL schema of this table is the following:

CREATE TABLE  kameleoon_experiment_event (
     nonce  BIGINT  PRIMARY KEY, -- unique identifier of the event
     timestamp  BIGINT, -- timestamp in millis of the event
     visitor_code  VARCHAR(255), -- Kameleoon visitor identifier
     custom_visitor_id  VARCHAR(255), -- Visitor identifier used by your company
     experiment_id  BIGINT, -- Kameleoon id of the experiment
     variation_id  BIGINT -- Kameleoon id of the variation
 );

Kameleoon reads the custom_visitor_id from the Cross Device Reconciliation custom data, if you set it.

Integrations

​Activate Databricks as a destination

​What “Use Databricks as a destination” does

Activate Databricks as a destination

What “Use Databricks as a destination” does