Spark on Livy

Configuring Apache Livy

Apache Livy can be used by Cloudera Hadoop Distribution, Amazon EMR or custom-made Spark cluster.

1. Set spark_engine to livy to use Apache Livy as a way to submit spark jobs.

[spark]
spark_engine = livy

2.Set Apache Livy Url in DBND config

[livy]
url = http://<livy_server_url>:8998

[livy] Configuration Section Parameter Reference

• root - Data outputs location override
• disable_task_band - Disable task_band file creation
• url - Determine livy's connection url, e.g. http://livy:8998
• auth - Set livy auth, e.g. None, Kerberos, or Basic_Access
• user - Set livy auth user.
• password - Set livy auth password.
• ignore_ssl_errors - Enable ignoring ssl errors.
• job_submitted_hook - Set the user code to be run after livy batch submit. This is a reference to a function. The expected interface is (LivySparkCtrl, Dict[str, Any]) -> None
• job_status_hook - Set the user code to be run at each livy batch status update. This is a reference to a function. The expected interface is (LivySparkCtrl, Dict[str, Any]) -> None
• retry_on_status_error - Set the number of retries for http requests if the status code is not accepted.
• retry_on_status_error_delay - Determien the amount of time, in seconds, in between retries for http requests.

Q: How can I debug HTTP requests related to the Livy server?

A: You can always run specific modules at debug state:
dbnd run ….. --set log.at_debug=module.you.want.to.debug
In our case, you want to log both livy_batch and reliable_http_client:
dbnd run ….. --set log.at_debug=dbnd._core.utils.http.reliable_http_client --set log.at_debug=dbnd_spark.livy.livy_batch