Target hadoop version

[hsn]

What hadoop version is targeted? I looked into sources and if i link it against CDH4, it discovers lot of deprecated API used.

CDH4 is hadoop 0.23.X based (subbranch of hadoop-2) and its available now.

[Costin Leau]

We're building against vanilla, Apache Hadoop 1.0.x (currently 1.0.3). From time to time we test against other distro as well. The API deprecation has been longed discussed in the Hadoop community (see the old vs new mapreduce api (mapred vs mapreduce) and the undeprecation of the old API right before 1.0 was released).
As hadoop-2 will stabilize we'll migrate the to its APIs as well but considering it is currently in alpha phase and there are no functional bugs, there is nothing to worry about.

[hsn]

Plain Hadoop is not much used, most users are using CDH distribution.

http://www.cloudera.com/hadoop/

[Costin Leau]

That might be the case, but there are other distros out there beside CDH, such as Greenplum, MapR, HortonWorks' etc...
By testing against vanilla Hadoop, SHDP guarantees compatibility with all of them. The latest CDH might include some deprecation but we need to be compatible with the version before that as well (CDH3).

My point is we're using a safe approach in increasing requirements so we can cater to as many Hadoop users out there.