Luiz Svoboda created AIRFLOW-4289: ------------------------------------- Summary: spark_binary argument in SparkSubmitHook is ignored when building the connection_cmd Key: AIRFLOW-4289 URL: https://issues.apache.org/jira/browse/AIRFLOW-4289 Project: Apache Airflow Issue Type: Bug Components: contrib, hooks Affects Versions: 1.10.3 Reporter: Luiz Svoboda
When using the SparkSubmitOperator, although it is possible to specify the parameter _spark_binary_, its value is ignored during the creation of the _connection_cmd_. Instead, the value used for this property is extracted from the connection parameters, or it defaults to _spark-submit_ as can be seen in [spark_submit_hook line:190|https://github.com/apache/airflow/blob/1.10.3/airflow/contrib/hooks/spark_submit_hook.py#L190] Actually, this configuration is a bit confusing as the user can configure it via _connection_ or directly when creating the operator instance. I suggest keeping only one option, and in this case, [IMHO] the connection approach seems to be better as it is already used to configure some other options. -- This message was sent by Atlassian JIRA (v7.6.3#76005)