Wrong number of partitions in keyset

[David Poisson <Da...@ca.fujitsu.com>]

I seem to have a problem with writing to an HFile: I always get an error saying I'm using the wrong number of partitions in keyset.

I'm using configureIncrementalLoad() on my CDH4 vmware: hbase-0.94.2+202-1.cdh4.2.0.p0.11.el6.x86_64 and hadoop-2.0.0+922-1.cdh4.2.0.p0.12.el6.x86_64.



I've modified my configuration to be in pseudo-distributed mode, however I find it strange that mapred.job.tracker returns "local". Is this an issue?



My table contains 2 regions according to http://localhost:60010/master-status/myTable.

After I call configureIncrementalLoad(), job.getNumReduceTasks() returns "2", yet when I watch job.getNumReduceTasks() in job.waitForCompletion(), it returns "1".

Finally, splitPoints.length is set to 1 (since I have 2 regions, I would think it's ok to have a single split point, right?).



K[] splitPoints = readPartitions(fs, partFile, keyClass, conf);

if (splitPoints.length != job.getNumReduceTasks() - 1) {

throw new IOException("Wrong number of partitions in keyset");

}



I've tried to look into the partitions.lst file to see if I could spot something wrong, however it doesn't seem to be very human readable.



I'd appreciate any pointers!



Cheers,



David