I believe that the mapred.child.env only concerns the spawned child process where the map is run. You probably need to put this path into the system property where the map task demon runs and possibly where the job controller runs.
Zhu, Guojun Modeling Sr Graduate 571-3824370 [email protected] Financial Engineering Freddie Mac Marek Miglinski <[email protected]> 06/11/2012 09:54 AM Please respond to [email protected] To "[email protected]" <[email protected]> cc Subject snappy codec Hi, I have a Clouderas CDH3u3 installed on my cluster and mapred.child.env set to "LD_LIBRARY_PATH=/usr/lib/hadoop-0.20/lib/native/Linux-amd64-64" (with libsnappy.so in the folder) in mapred-site.xml. Cloudera says that Snappy is included in their hadoop-0.20-native package and it is also installed on each of the nodes. But when I run a mapreduce task with "mapred.map.output.compression.codec" set to "org.apache.hadoop.io.compress.SnappyCodec" I get an exception: java.lang.RuntimeException: native snappy library not available Any idea why? Thanks, Marek M.
