Details
-
Bug
-
Status: Closed
-
Minor
-
Resolution: Fixed
-
1.0.0
-
None
-
None
-
Patch Available
Description
After upgrade to Hadoop 0.19.1 Reducer is initialized in a different order than before (see http://svn.apache.org/viewvc?view=rev&revision=736239). IndexingFilters populate current JobConf with field options that are required for IndexerOutputFormat to function properly. However, the filters are instantiated in Reducer.configure(), which is now called after the OutputFormat is initialized, and not before as previously.
The workaround for now is to instantiate IndexinigFilters once again inside IndexerOutputFormat. This issue should be revisited before 1.1 in order to find a better solution.
See this thread for more information: http://www.lucidimagination.com/search/document/7c62c625c7ea17fe/problem_with_crawling_using_the_latest_1_0_trunk