I'm attempting to copy 30 billion rows from one hive table into another hive table. The tables are both created the same and are partitioned on date (DT). Currently there are 1173 partitions. I'm using the following query: insert into accesslog_new PARTITION (DT) select * from accesslog;
This query has been running for almost 3 days straight on a cluster with 18 data nodes.
My issue is that the Map-Reduce job only creates one reducer step. Btw, we are using MR2. I'm guessing this is drastically slowing things down. Is there a way to force the number of reducers to be much larger? How do you also figure out what an appropriate number of reducers would be for that volume of data?