SparkJDBCOperator

Spark

This operator extends the SparkSubmitOperator specifically for performing data transfers to/from JDBC-based databases with Apache Spark. As with the SparkSubmitOperator, it assumes that the “spark-submit” binary is available on the PATH.

View Source

Last Updated: Apr. 27, 2021

Access Instructions

Install the Spark provider package into your Airflow environment.

Import the module into your DAG file and instantiate it with your desired params.

Parameters

spark_app_namestrName of the job (default airflow-spark-jdbc)
spark_conn_idstrConnection id as configured in Airflow administration
spark_confdictAny additional Spark configuration properties
spark_py_filesstrAdditional python files used (.zip, .egg, or .py)
spark_filesstrAdditional files to upload to the container running the job
spark_jarsstrAdditional jars to upload and add to the driver and executor classpath
num_executorsintnumber of executor to run. This should be set so as to manage the number of connections made with the JDBC database
executor_coresintNumber of cores per executor
executor_memorystrMemory per executor (e.g. 1000M, 2G)
driver_memorystrMemory allocated to the driver (e.g. 1000M, 2G)
verboseboolWhether to pass the verbose flag to spark-submit for debugging
keytabstrFull path to the file that contains the keytab
principalstrThe name of the kerberos principal used for keytab
cmd_typestrWhich way the data should flow. 2 possible values: spark_to_jdbc: data written by spark from metastore to jdbc jdbc_to_spark: data written by spark from jdbc to metastore
jdbc_tablestrThe name of the JDBC table
jdbc_conn_idstrConnection id used for connection to JDBC database
jdbc_driverstrName of the JDBC driver to use for the JDBC connection. This driver (usually a jar) should be passed in the 'jars' parameter
metastore_tablestrThe name of the metastore table,
jdbc_truncatebool(spark_to_jdbc only) Whether or not Spark should truncate or drop and recreate the JDBC table. This only takes effect if 'save_mode' is set to Overwrite. Also, if the schema is different, Spark cannot truncate, and will drop and recreate
save_modestrThe Spark save-mode to use (e.g. overwrite, append, etc.)
save_formatstr(jdbc_to_spark-only) The Spark save-format to use (e.g. parquet)
batch_sizeint(spark_to_jdbc only) The size of the batch to insert per round trip to the JDBC database. Defaults to 1000
fetch_sizeint(jdbc_to_spark only) The size of the batch to fetch per round trip from the JDBC database. Default depends on the JDBC driver
num_partitionsintThe maximum number of partitions that can be used by Spark simultaneously, both for spark_to_jdbc and jdbc_to_spark operations. This will also cap the number of JDBC connections that can be opened
partition_columnstr(jdbc_to_spark-only) A numeric column to be used to partition the metastore table by. If specified, you must also specify: num_partitions, lower_bound, upper_bound
lower_boundint(jdbc_to_spark-only) Lower bound of the range of the numeric partition column to fetch. If specified, you must also specify: num_partitions, partition_column, upper_bound
upper_boundint(jdbc_to_spark-only) Upper bound of the range of the numeric partition column to fetch. If specified, you must also specify: num_partitions, partition_column, lower_bound
create_table_column_types(spark_to_jdbc-only) The database column data types to use instead of the defaults, when creating the table. Data type information should be specified in the same format as CREATE TABLE columns syntax (e.g: "name CHAR(64), comments VARCHAR(1024)"). The specified types should be valid spark sql data types.

Documentation

This operator extends the SparkSubmitOperator specifically for performing data transfers to/from JDBC-based databases with Apache Spark. As with the SparkSubmitOperator, it assumes that the “spark-submit” binary is available on the PATH.

See also

For more information on how to use this operator, take a look at the guide: SparkJDBCOperator

Was this page helpful?