why does the random sampler get some identical values?

[forbbs forbbs]

I use:
InputSampler.Sampler<LongWritable, NullWritable> sampler = new
InputSampler.RandomSampler<LongWritable, NullWritable>(1, 1280, 5);

My input data doesn't contain identical values, but the sampler writes
some identical values into the partition file, which causes:
'java.lang.IllegalArgumentException: Can't read partitions file'
'java.io.IOException: Split points are out of order'.

The hadoop version is 0.19.0.