Then if we setup our new TableMapReduceUtil TableMapReduceUtil.initTableReduceJob(table, reducer, job) to set the number of reduces to the same number of regions we could make sure no more then one reduce task was writing to more then one region at a time.
Does anyone out there have any experience writing a hadoop partitioner we
need on for hbase to split the records from Map outputs
So that all records will for a region will fall in one partition would need
something that is fast as each output would have to be ran by it
