[
https://issues.apache.org/jira/browse/AIRFLOW-1409?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Luis Angel Vicente Sanchez updated AIRFLOW-1409:
------------------------------------------------
Description:
If you configure the connection using an ENV variable (and possibly through the
UI), SparkSubmitHook is ignoring the scheme and doesn't produce valid master
URL. Example, this ENV variable,
AIRFLOW_CONN_SPARK_CLUSTER=spark://spark-master:7077
will produce the following master URL:
spark-master:7077
This URL is not valid and the job submission fails. The issue is how the url is
parsed to create a connection, I have a partial fix working and I will submit a
PR to github soon.
was:
If you configure the connection using an ENV variable (and possibly through the
UI), SparkSubmitHook is ignoring the scheme and doesn't produce valid master
URL. Example, this ENV variable,
AIRFLOW_CONN_SPARK_CLUSTER=spark://spark-master:7077
will produce the following master URL:
spark-master:7077
This URL is not valid and the job submission fails (the same happens with a
mesos master).
I will create a PR on github that fixes this.
> SparkSubmitHook doesn't generate a valid master URL
> ---------------------------------------------------
>
> Key: AIRFLOW-1409
> URL: https://issues.apache.org/jira/browse/AIRFLOW-1409
> Project: Apache Airflow
> Issue Type: Bug
> Reporter: Luis Angel Vicente Sanchez
>
> If you configure the connection using an ENV variable (and possibly through
> the UI), SparkSubmitHook is ignoring the scheme and doesn't produce valid
> master URL. Example, this ENV variable,
> AIRFLOW_CONN_SPARK_CLUSTER=spark://spark-master:7077
> will produce the following master URL:
> spark-master:7077
> This URL is not valid and the job submission fails. The issue is how the url
> is parsed to create a connection, I have a partial fix working and I will
> submit a PR to github soon.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
