Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Not A Problem
-
0.8.1
-
None
-
None
-
CentOS
12 boxes
30 map/reduce tasks max per box @ 2GB memory each
Description
When running Pig scripts through the grunt shell I kept getting Java Heap Space error messages. The cause was that my Hadoop cluster was running 467 map tasks which were being passed down to only 199 reducers. The notice I was getting through the Grunt shell was:
"Neither PARALLEL nor default parallelism is set for this job. Setting number of reducers to 199"
I tried the following commands, none of which worked:
SET DEFAULT_PARALLELISM 2;
SET mapred.reduce.tasks 450;
SET mapred.running.reduce.limit 30;
When I ran the script directly from a file (pig filename.pig) it ran without problems.