Access a partitioned table created in Impala using Hive

[#1]

I've recently created the following table in Impala....

create table hpchistperf.oozie_job_stats
(
name string,
start_time timestamp,
end_time timestamp,
rows_inserted integer,
last_id_loaded integer,
primary key(name,start_time))
PARTITION BY HASH PARTITIONS 16
stored as kudu;

I read that there was no problem creating the table in Impala and it could be accessed in Hive since they share the same metadata.

The problem is that when I run the query on Hive it gives me the following error:

FAILED: RuntimeException java.lang.ClassNotFoundException: org.apache.kudu.mapreduce.KuduTableInputFormat

[#2 · Accepted solution]

@Hari 

When you create a table in Impala with data in kudu, the metadata resides in the Hive metastore and you can see the table in hive as well as both Hive and Impala share the same metadata. But if you try to access the data by running a query in hive over kudu table, you are bound to get the below error.

FAILED: RuntimeException java.lang.ClassNotFoundException: org.apache.kudu.mapreduce.KuduTableInputFormat

This is because in CDH, Hive doesn't support accessing Kudu. It is added in Hive 4.0 in HIVE-12971 and is designed to work with Kudu 1.2+

You can try out in CDP where in this feature has been added.

https://cwiki.apache.org/confluence/display/Hive/Kudu+Integration

https://issues.apache.org/jira/browse/HIVE-12971