Configure Spark SQL task parameters

更新时间:
复制 MD 格式

Spark SQL tasks support different parameters depending on the service type.

Background information

Spark SQL tasks support various service types across different compute engines. For more information, see .

Spark SQL task parameters

Available task parameters vary by service type.

Livy

Parameter name

Description

livy.session.driverMemory

Memory allocated to the driver process.

livy.session.driverCores

Number of cores for the driver process.

livy.session.executorMemory

Memory allocated to each executor process.

livy.session.executorCores

Number of cores for each executor.

livy.session.numExecutors

Number of executors to launch for the session.

spark.yarn.queue

The execution queue. Defaults to Default.

Kyuubi

Parameter name

Default value

Description

spark.executor.instances

1

Number of statically allocated executors.

spark.executor.cores

1

Number of cores for each executor.

spark.executor.memory

1G

Memory allocated to each executor process.

spark.yarn.am.memory

512M

Memory allocated to the YARN Application Master in client mode.

spark.yarn.am.memoryOverhead

-

Non-heap memory allocated to the Application Master in client mode.

spark.driver.memory

1G

Memory allocated to the driver process.

spark.driver.memoryOverhead

-

Non-heap memory allocated to the driver process in cluster mode.

spark.executor.memoryOverhead

-

Additional memory for each executor process, covering VM overhead, internal strings, and other native overheads.

spark.yarn.queue

Default

The execution queue. Defaults to Default.

Thrift server, MapReduce (MRS)

Parameter name

Default value

Description

spark.executor.instances

1

Number of statically allocated executors.

spark.executor.cores

1

Number of cores for each executor.

spark.executor.memory

1G

Memory allocated to each executor process.

spark.yarn.am.memory

512M

Memory allocated to the YARN Application Master in client mode.

spark.yarn.am.memoryOverhead

-

Non-heap memory allocated to the Application Master in client mode.

spark.driver.memory

1G

Memory allocated to the driver process.

spark.driver.memoryOverhead

-

Non-heap memory allocated to the driver process in cluster mode.

spark.executor.memoryOverhead

-

Additional memory for each executor process, covering VM overhead, internal strings, and other native overheads.