Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-16349

IsolatedClientLoader ignores needed Hadoop classes not present in Spark's loader

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Resolved
    • Major
    • Resolution: Fixed
    • 2.0.0
    • 2.1.0
    • SQL
    • None

    Description

      While trying to use a custom classpath for metastore jars (spark.sql.hive.metastore.jars pointing at some filesystem path), I ran into the following issue:

      java.lang.ClassNotFoundException: java.lang.NoClassDefFoundError: org/apache/hadoop/mapred/MRVersion when creating Hive client using classpath
      

      The issue here is that MRVersion is not packaged anywhere with Spark, and the code in IsolatedClientLoader only ever tries the parent class loader when loading hadoop classes in this configuration. So even though I had the class in the list of files in spark.sql.hive.metastore.jars, Spark never tries to load it.

      Attachments

        Activity

          People

            vanzin Marcelo Masiero Vanzin
            vanzin Marcelo Masiero Vanzin
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: