Okay first of all my problem is exactly the same. What had somehow escaped me is that the insert I was running was actually inserting records selected from another parquet table. The error of course was from hive being unable to read the source parquet table.   Tried the workaround and I can confirm it is working. It took me a while to realize that you have to recreate the partitions in the new table otherwise you get no output. It does introduce some warnings, though, as below:   hive> select * from tbl_ptr limit 1; OK SLF4J: Failed to load class "org.slf4j.impl.StaticLoggerBinder". SLF4J: Defaulting to no-operation (NOP) logger implementation SLF4J: See http://www.slf4j.org/codes.html#StaticLoggerBinder for further details. <RECORD OUTPUT OKAY> Time taken: 0.387 seconds, Fetched: 1 row(s) hive> quit; Jul 3, 2015 10:36:20 AM WARNING: parquet.hadoop.ParquetRecordReader: Can not initialize counter due to context is not a instance of TaskInputOutputContext, but is org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl Jul 3, 2015 10:36:20 AM INFO: parquet.hadoop.InternalParquetRecordReader: RecordReader initialized will read a total of 17636531 records. Jul 3, 2015 10:36:20 AM INFO: parquet.hadoop.InternalParquetRecordReader: at row 0. reading next block Jul 3, 2015 10:36:21 AM INFO: parquet.hadoop.InternalParquetRecordReader: block read in memory in 755 ms. row count = 17636531 ... View more

Hi,   I am having this same issue using Cloudera 5.4.2. The error appears even without running the invalidate metadata or compute stats from impala. I have dropped and recreated the table even with a different name to just be sure there wasn't some residual metadata that was causing this. I am using partitions too.   I ended up having to create the table and partitions and insert data using impala. Surprisingly, hive cannot select data from the table either. Same error. ... View more