Hi all,
When I run a native MR job with the MAPREDUCE keyword and store the
intermediate data in HBase with:
stored = MAPREDUCE 'my.jar'
STORE x INTO 'hbase://temp_table'
USING
org.apache.pig.backend.hadoop.hbase.HBaseStorage('hbase_schema')
.... and the rest ....;
Pig tries to delete the temp files, which in this case has an HBase path, and
fails with the exception:
Caused by: java.lang.IllegalArgumentException: java.net.URISyntaxException:
Relative path in absolute URI: file:hbase:/temp_table
at org.apache.hadoop.fs.Path.initialize(Path.java:148)
at org.apache.hadoop.fs.Path.<init>(Path.java:126)
at
org.apache.pig.backend.hadoop.datastorage.HDataStorage.isContainer(HDataStorage.java:197)
at
org.apache.pig.backend.hadoop.datastorage.HDataStorage.asElement(HDataStorage.java:128)
at org.apache.pig.impl.io.FileLocalizer.delete(FileLocalizer.java:415)
at
org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.MapReduceLauncher.launchPig(MapReduceLauncher.java:419)
at org.apache.pig.PigServer.launchPlan(PigServer.java:1322)
The reason is that in the FileLocalizer class I see that hbase tables are not
handled properly; in fact the given file spec is only checked against
LOCAL_PREFIX ("file:") and if it's not a local file then HDFS is used as the
file system. Anyway, I created an issue for this
(PIG-3617<https://issues.apache.org/jira/browse/PIG-3617>) and will be happy to
help you guys implement the fix.
Thanks,
Nezih
