Skip to main content
Version: Next

Usage

Configuring the OpenLineage Spark integration is straightforward. It uses built-in Spark configuration mechanisms. However, for Databricks users, special considerations are required to ensure compatibility and avoid breaking the Spark UI after a cluster shutdown.

Your options are:

  1. Setting the properties directly in your application.
  2. Using --conf options with the CLI.
  3. Adding properties to the spark-defaults.conf file in the ${SPARK_HOME}/conf directory.

Setting the properties directly in your application​

The below example demonstrates how to set the properties directly in your application when constructing a SparkSession.

warning

The setting config("spark.extraListeners", "io.openlineage.spark.agent.OpenLineageSparkListener") is extremely important. Without it, the OpenLineage Spark integration will not be invoked, rendering the integration ineffective.

note

Databricks For Databricks users, you must include com.databricks.backend.daemon.driver.DBCEventLoggingListener in addition to io.openlineage.spark.agent.OpenLineageSparkListener in the spark.extraListeners setting. Failure to do so will make the Spark UI inaccessible after a cluster shutdown.

import org.apache.spark.sql.SparkSession

object OpenLineageExample extends App {
  val spark = SparkSession.builder()
    .appName("OpenLineageExample")
    // This line is EXTREMELY important
    .config("spark.extraListeners", "io.openlineage.spark.agent.OpenLineageSparkListener")
    .config("spark.openlineage.transport.type", "http")
    .config("spark.openlineage.transport.url", "http://localhost:5000")
    .config("spark.openlineage.namespace", "spark_namespace")
    .config("spark.openlineage.parentJobNamespace", "airflow_namespace")
    .config("spark.openlineage.parentJobName", "airflow_dag.airflow_task")
    .config("spark.openlineage.parentRunId", "xxxx-xxxx-xxxx-xxxx")
    .getOrCreate()

  // ... your code

  spark.stop()
}

// For Databricks
import org.apache.spark.sql.SparkSession

object OpenLineageExample extends App {
  val spark = SparkSession.builder()
    .appName("OpenLineageExample")
    // This line is EXTREMELY important
    .config("spark.extraListeners", "io.openlineage.spark.agent.OpenLineageSparkListener,com.databricks.backend.daemon.driver.DBCEventLoggingListener")
    .config("spark.openlineage.transport.type", "http")
    .config("spark.openlineage.transport.url", "http://localhost:5000")
    .config("spark.openlineage.namespace", "spark_namespace")
    .config("spark.openlineage.parentJobNamespace", "airflow_namespace")
    .config("spark.openlineage.parentJobName", "airflow_dag.airflow_task")
    .config("spark.openlineage.parentRunId", "xxxx-xxxx-xxxx-xxxx")
    .getOrCreate()

  // ... your code

  spark.stop()
}

Using --conf options with the CLI​

The below example demonstrates how to use the --conf option with spark-submit.

note

Databricks Remember to include com.databricks.backend.daemon.driver.DBCEventLoggingListener along with the OpenLineage listener.

spark-submit \
  --conf "spark.extraListeners=io.openlineage.spark.agent.OpenLineageSparkListener" \
  --conf "spark.openlineage.transport.type=http" \
  --conf "spark.openlineage.transport.url=http://localhost:5000" \
  --conf "spark.openlineage.namespace=spark_namespace" \
  --conf "spark.openlineage.parentJobNamespace=airflow_namespace" \
  --conf "spark.openlineage.parentJobName=airflow_dag.airflow_task" \
  --conf "spark.openlineage.parentRunId=xxxx-xxxx-xxxx-xxxx" \
  # ... other options

# For Databricks
spark-submit \
  --conf "spark.extraListeners=io.openlineage.spark.agent.OpenLineageSparkListener,com.databricks.backend.daemon.driver.DBCEventLoggingListener" \
  --conf "spark.openlineage.transport.type=http" \
  --conf "spark.openlineage.transport.url=http://localhost:5000" \
  --conf "spark.openlineage.namespace=spark_namespace" \
  --conf "spark.openlineage.parentJobNamespace=airflow_namespace" \
  --conf "spark.openlineage.parentJobName=airflow_dag.airflow_task" \
  --conf "spark.openlineage.parentRunId=xxxx-xxxx-xxxx-xxxx" \
  # ... other options

Adding properties to the spark-defaults.conf file in the ${SPARK_HOME}/conf directory​

warning

You may need to create this file if it does not exist. If it does exist, we strongly suggest that you back it up before making any changes, particularly if you are not the only user of the Spark installation. A misconfiguration here can have devastating effects on the operation of your Spark installation, particularly in a shared environment.

The below example demonstrates how to add properties to the spark-defaults.conf file.

note

Databricks For Databricks users, include com.databricks.backend.daemon.driver.DBCEventLoggingListener in the spark.extraListeners property.

spark.extraListeners=io.openlineage.spark.agent.OpenLineageSparkListener
spark.openlineage.transport.type=http
spark.openlineage.transport.url=http://localhost:5000
spark.openlineage.namespace=MyNamespace

For Databricks,

spark.extraListeners=io.openlineage.spark.agent.OpenLineageSparkListener,com.databricks.backend.daemon.driver.DBCEventLoggingListener
spark.openlineage.transport.type=http
spark.openlineage.transport.url=http://localhost:5000
spark.openlineage.namespace=MyNamespace
info

The spark.extraListeners configuration parameter is non-additive. This means that if you set spark.extraListeners via the CLI or via SparkSession#config, it will replace the value in spark-defaults.conf. This is important to remember if you are using spark-defaults.conf to set a default value for spark.extraListeners and then want to override it for a specific job.

info

When it comes to configuration parameters like spark.openlineage.namespace, a default value can be supplied in the spark-defaults.conf file. This default value can be overridden by the application at runtime, via the previously detailed methods. However, it is strongly recommended that more dynamic or quickly changing parameters like spark.openlineage.parentRunId or spark.openlineage.parentJobName be set at runtime via the CLI or SparkSession#config methods.