On Sat, Nov 5, 2011 at 7:16 PM, Mikhail Bautin <[email protected]> wrote:
> Hi Ted,
>
> I actually just run every test class (Test*.java) in its own mapper using 
> "mvn -Dtest=Test… test" from a script running over Hadoop streaming, and 
> upload test results into HDFS.
>
> The effort to categorize tests into small, medium, and large is very useful. 
> We probably won't do this in 89-fb for now, though. I guess HBASE-4746 would 
> be good enough for parallelizing 89-fb tests at the moment.
>

Running the tests in a mapreduce job made me lol -- especially the bit
about a 100 node cluster.  I like it.
St.Ack

Reply via email to