[SPARK-18660] Parquet complains "Can not initialize counter due to context is not a instance of TaskInputOutputContext, but is org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl "

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Incomplete
• Component/s: SQL
• Labels: bulk-closed
• Assignee: Unassigned
• Reporter: Yin Huai
• Created: 30/Nov/16 22:14
• Updated: 12/Dec/22 18:11
• Resolved: 25/May/21 01:39

Description

Parquet record reader always complain "Can not initialize counter due to context is not a instance of TaskInputOutputContext, but is org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl". Looks like we always create TaskAttemptContextImpl (https://github.com/apache/spark/blob/2f7461f31331cfc37f6cfa3586b7bbefb3af5547/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/parquet/ParquetFileFormat.scala#L368). But, Parquet wants to use TaskInputOutputContext, which is a subclass of TaskAttemptContextImpl.