Github user chtyim commented on the pull request:
https://github.com/apache/spark/pull/11337#issuecomment-188707421
The reason why failing is bad in this case is because if either the
`HADOOP_CONF_DIR` or the `YARN_CONF_DIR` is not readable, exception is raised,
even if one of those directories already provides all the hadoop conf files
that are needed by Spark.
E.g. in one of our use case, we actually launch Spark job from a YARN
container, which the `HADOOP_CONF_DIR` is set to be a private directory that is
owned by the NM (that's how Cloudera CM works anyway), which the container
fails to read. Even we try to workaround it by setting the `YARN_CONF_DIR` to a
readable directory, however, due to the exception being raised, we can't submit
the job even those hadoop files are available.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
