Not sure what you are trying, or if MR is the right tool...
Anyway, the basic idea is to write a map-only job with a custom RecordReader
to stop emitting inputs to your map when the mapper/recordreader finds the
needle in the haystack.
Arun
Sent from my iPhone
On Sep 30, 2011, at 8:39 PM,
Arun,
Let's say there is 1PB of data in HDFS and the requirement is to look for a
particular sequence in the whole data. If the sequence is found then the MR
job could be aborted. There is no meaning in keeping the Map/Reduce tasks
running.
There was a query in one of the forums on how a MR job
