To integrate Spark with DataHub, we provide a lightweight Java agent that listens for Spark application and job events and pushes metadata out to DataHub in real-time. The agent listens to events such application start/end, and SQLExecution start/end to create pipelines (i.e. DataJob) and tasks (i.e. DataFlow) in Datahub along with lineage to datasets that are being read from and written to. Read on to learn how to configure this for different Spark scenarios.

Configuring Spark agent

The Spark agent can be configured using a config file or while creating a spark Session.

Before you begin: Versions and Release Notes

Versioning of the jar artifact will follow the semantic versioning of the main DataHub repo and release notes will be available here. Always check the Maven central repository for the latest released version.

Configuration Instructions: spark-submit

When running jobs using spark-submit, the agent needs to be configured in the config file.

#Configuring datahub spark agent jar
spark.jars.packages                          io.acryl:datahub-spark-lineage:0.8.23
spark.extraListeners                         datahub.spark.DatahubSparkListener
spark.datahub.rest.server                    http://localhost:8080

Configuration for Amazon EMR

Set the following spark-defaults configuration properties as it stated here

spark.jars.packages        io.acryl:datahub-spark-lineage:0.8.23
spark.extraListeners                         datahub.spark.DatahubSparkListener
spark.datahub.rest.server                    https://your_datahub_host/gms
#If you have authentication set up then you also need to specify the Datahub access token
spark.datahub.rest.token                     yourtoken

Configuration Instructions: Notebooks

When running interactive jobs from a notebook, the listener can be configured while building the Spark Session.

spark = SparkSession.builder \
          .master("spark://spark-master:7077") \
          .appName("test-application") \
          .config("spark.jars.packages","io.acryl:datahub-spark-lineage:0.8.23") \
          .config("spark.extraListeners","datahub.spark.DatahubSparkListener") \
          .config("spark.datahub.rest.server", "http://localhost:8080") \
          .enableHiveSupport() \
          .getOrCreate()

Configuration Instructions: Standalone Java Applications

The configuration for standalone Java apps is very similar.

spark = SparkSession.builder()
        .appName("test-application")
        .config("spark.master", "spark://spark-master:7077")
        .config("spark.jars.packages","io.acryl:datahub-spark-lineage:0.8.23")
        .config("spark.extraListeners", "datahub.spark.DatahubSparkListener")
        .config("spark.datahub.rest.server", "http://localhost:8080")
        .enableHiveSupport()
        .getOrCreate();

Configuration details

Field	Required	Default	Description
spark.jars.packages	✅		Set with latest/required version io.acryl:datahub-spark-lineage:0.8.23
spark.extraListeners	✅		datahub.spark.DatahubSparkListener
spark.datahub.rest.server	✅		Datahub server url eg:http://localhost:8080
spark.datahub.rest.token			Authentication token.
spark.datahub.rest.disable_ssl_verification		false	Disable SSL certificate validation. Caution: Only use this if you know what you are doing!
spark.datahub.metadata.pipeline.platformInstance			Pipeline level platform instance
spark.datahub.metadata.dataset.platformInstance			dataset level platform instance
spark.datahub.metadata.dataset.env		PROD	Supported values. In all other cases, will fallback to PROD
spark.datahub.coalesce_jobs		false	Only one datajob(taask) will be emitted containing all input and output datasets for the spark application
spark.datahub.parent.datajob_urn			Specified dataset will be set as upstream dataset for datajob created. Effective only when spark.datahub.coalesce_jobs is set to true

What to Expect: The Metadata Model

As of current writing, the Spark agent produces metadata related to the Spark job, tasks and lineage edges to datasets.

A pipeline is created per Spark <master, appName>.
A task is created per unique Spark query execution within an app.

Custom properties & relating to Spark UI

The following custom properties in pipelines and tasks relate to the Spark UI:

appName and appId in a pipeline can be used to determine the Spark application
description and SQLQueryId in a task can be used to determine the Query Execution within the application on the SQL tab of Spark UI

Other custom properties of pipelines and tasks capture the start and end times of execution etc. The query plan is captured in the queryPlan property of a task.

Spark versions supported

The primary version tested is Spark/Scala version 2.4.8/2_11. This library has also been tested to work with Spark versions(2.2.0 - 2.4.8) and Scala versions(2.10 - 2.12). For the Spark 3.x series, this has been tested to work with Spark 3.1.2 and 3.2.0 with Scala 2.12. Other combinations are not guaranteed to work currently. Support for other Spark versions is planned in the very near future.

Environments tested with

This initial release has been tested with the following environments:

spark-submit of Python/Java applications to local and remote servers
Jupyter notebooks with pyspark code
Standalone Java applications

Note that testing for other environments such as Databricks is planned in near future.

Spark commands supported

Below is a list of Spark commands that are parsed currently:

InsertIntoHadoopFsRelationCommand
SaveIntoDataSourceCommand (jdbc)
CreateHiveTableAsSelectCommand
InsertIntoHiveTable

Effectively, these support data sources/sinks corresponding to Hive, HDFS and JDBC.

DataFrame.persist command is supported for below LeafExecNodes:

FileSourceScanExec
HiveTableScanExec
RowDataSourceScanExec
InMemoryTableScanExec

Spark commands not yet supported

View related commands
Cache commands and implications on lineage
RDD jobs

Important notes on usage

It is advisable to ensure appName is used appropriately to ensure you can trace lineage from a pipeline back to your source code.
If multiple apps with the same appName run concurrently, dataset-lineage will be captured correctly but the custom-properties e.g. app-id, SQLQueryId would be unreliable. We expect this to be quite rare.
If spark execution fails, then an empty pipeline would still get created, but it may not have any tasks.
For HDFS sources, the folder (name) is regarded as the dataset (name) to align with typical storage of parquet/csv formats.

Debugging

Following info logs are generated

On Spark context startup

YY/MM/DD HH:mm:ss INFO DatahubSparkListener: DatahubSparkListener initialised.
YY/MM/DD HH:mm:ss INFO SparkContext: Registered listener datahub.spark.DatahubSparkListener

On application start

YY/MM/DD HH:mm:ss INFO DatahubSparkListener: Application started: SparkListenerApplicationStart(AppName,Some(local-1644489736794),1644489735772,user,None,None)
YY/MM/DD HH:mm:ss INFO McpEmitter: REST Emitter Configuration: GMS url <rest.server>
YY/MM/DD HH:mm:ss INFO McpEmitter: REST Emitter Configuration: Token XXXXX

On pushing data to server

YY/MM/DD HH:mm:ss INFO McpEmitter: MetadataWriteResponse(success=true, responseContent={"value":"<URN>"}, underlyingResponse=HTTP/1.1 200 OK [Date: day, DD month year HH:mm:ss GMT, Content-Type: application/json, X-RestLi-Protocol-Version: 2.0.0, Content-Length: 97, Server: Jetty(9.4.46.v20220331)] [Content-Length: 97,Chunked: false])

On application end

YY/MM/DD HH:mm:ss INFO DatahubSparkListener: Application ended : AppName AppID

To enable debugging logs, add below configuration in log4j.properties file

log4j.logger.datahub.spark=DEBUG
log4j.logger.datahub.client.rest=DEBUG

Known limitations

Only postgres supported for JDBC sources in this initial release. Support for other driver URL formats will be added in future.
Behavior with cached datasets is not fully specified/defined in context of lineage.
There is a possibility that very short-lived jobs that run within a few milliseconds may not be captured by the listener. This should not cause an issue for realistic Spark applications.