[SPARK-29474] CLI support for Spark-on-Docker-on-Yarn

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.1.0
• Component/s: Spark Core, Spark Shell, YARN
• Assignee: Unassigned
• Reporter: Adam Antal
• Created: 15/Oct/19 10:15
• Updated: 17/May/20 18:13

Description

The Docker-on-Yarn feature is stable for a while now in Hadoop.
One can run Spark on Docker using the Docker-on-Yarn feature by providing runtime environments to the Spark AM and Executor containers similar to this:

--conf spark.yarn.appMasterEnv.YARN_CONTAINER_RUNTIME_TYPE=docker
--conf spark.yarn.appMasterEnv.YARN_CONTAINER_RUNTIME_DOCKER_IMAGE=repo/image:tag
--conf spark.yarn.appMasterEnv.YARN_CONTAINER_RUNTIME_DOCKER_MOUNTS="/etc/passwd:/etc/passwd:ro,/etc/hadoop:/etc/hadoop:ro"
--conf spark.executorEnv.YARN_CONTAINER_RUNTIME_TYPE=docker
--conf spark.executorEnv.YARN_CONTAINER_RUNTIME_DOCKER_IMAGE=repo/image:tag
--conf spark.executorEnv.YARN_CONTAINER_RUNTIME_DOCKER_MOUNTS="/etc/passwd:/etc/passwd:ro,/etc/hadoop:/etc/hadoop:ro"

This is not very user friendly. I suggest to add CLI options to specify:

• whether docker image should be used (--docker)
• which docker image should be used (--docker-image)
• what docker mounts should be used (--docker-mounts)
  for the AM and executor containers separately.

Let's discuss!