0
votes

I'm wondering is there any way for spark-submit to temporarily change yarn job's configuration?

The question is because our Spark over yarn cluster's history server only allows admin to access, which is not convenience for user to retrieve their own job's log. I learned "mapreduce.job.acl-view-job" in mapreduce-default.xml can changes specific job's ACL. As I'm using spark-submit to launch job, and "--conf" is reserved for spark itself's, but how can I set yarn's configuration from command line along with application?

1

1 Answers

2
votes

You can modify Spark's Hadoop Configuration (obtained via SparkContaxt.hadoopConfiguration) by adding a --conf with a spark.hadoop. prefixed.

In you example it would be

spark-submit --conf spark.hadoop.mapreduce.job.acl-view-job=YOUR_ACL_STATEMENT ...