10-09-2013 12:23 PM - edited 10-09-2013 12:29 PM
other non-partitioned and smaller number partition tables still work.
This table has 6,000 partitions.
hive> select * from table_name limit 10;
FAILED: SemanticException org.apache.thrift.transport.TTransportException: java.net.SocketTimeoutException: Read timed out
Query: select count(1) , to_date(dt) from table_name group by to_date(dt) order by to_date(dt) limit 10000
ERROR: AnalysisException: Failed to load metadata for table: table_name
CAUSED BY: TableLoadingException: Failed to load metadata for table: table_name
CAUSED BY: TTransportException: java.net.SocketTimeoutException: Read timed out
CAUSED BY: SocketTimeoutException: Read timed out
10-09-2013 03:59 PM - edited 10-09-2013 04:00 PM
With that many partitions, it takes time to load them from the metastore. Set hive.metastore.client.socket.timeout to a large number, like 300 (units are seconds).
You may also want/need to increase the heap size for the metastore.