Tag Archives: [Hive on Tez] Input path does not exists error

[Hive on Tez] Input path does not exists error

Why are there a series of technical challenges behind “OMG buy it”>>>

Issue

When hive metadata has partition information partition = x and HDFS path does not have partition directory partition = X. Executing some hive SQL will report an error: org. Apache. Hadoop. Mapred. Invalidinputexception: input path does not exist

This is tez’s usual directory inconsistency problem

See issue: https://issues.apache.org/jira/browse/HIVE-13781

Impact: hive on tez. Hive3

Solutions

When executing hive SQL, the consistency between HDFS directory and hive metadata is guaranteed

Retreat on MR

Drop all partitions, execute msck repair table to repair the whole table (or use mask repair table sync partitions to synchronize the partition folder information on HDFS)