[jira] Commented: (PIG-1106) FR join should not spill

["Ankit Modi (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/PIG-1106?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12789294#action_12789294 ] 

Ankit Modi commented on PIG-1106:
---------------------------------

Tests I ran were using two files

file format
f1: random chararray(100)
f2: random int

leftside file contained 100 tuples and right side file contain 3million tuples.

Code
{noformat}
A = load 'leftsidefrjoin.txt' as ( key, value);
B = load 'rightsidefrjoin.txt' as (key, value);
C = join A by key left, B by key using "repl";
--- Fragmented input and replicated input
store C into 'output';
{noformat}

This generated following error
{noformat}
FATAL org.apache.hadoop.mapred.TaskTracker: Error running child : java.lang.OutOfMemoryError: GC overhead limit exceeded
	at java.util.ArrayList.<init>(ArrayList.java:112)
	at org.apache.pig.data.DefaultTuple.<init>(DefaultTuple.java:63)
	at org.apache.pig.data.DefaultTupleFactory.newTuple(DefaultTupleFactory.java:35)
	at org.apache.pig.backend.hadoop.executionengine.physicalLayer.relationalOperators.POLocalRearrange.constructLROutput(POLocalRearrange.java:369)
	at org.apache.pig.backend.hadoop.executionengine.physicalLayer.relationalOperators.POLocalRearrange.getNext(POLocalRearrange.java:288)
	at org.apache.pig.backend.hadoop.executionengine.physicalLayer.relationalOperators.POFRJoin.setUpHashMap(POFRJoin.java:351)
	at org.apache.pig.backend.hadoop.executionengine.physicalLayer.relationalOperators.POFRJoin.getNext(POFRJoin.java:211)
	at org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.PigMapBase.runPipeline(PigMapBase.java:250)
	at org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.PigMapBase.map(PigMapBase.java:241)
	at org.apache.pig.backend.hadoop.executionengine.mapReduceLayer.PigMapOnly$Map.map(PigMapOnly.java:65)
	at org.apache.hadoop.mapred.MapRunner.run(MapRunner.java:50)
	at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:358)
	at org.apache.hadoop.mapred.MapTask.run(MapTask.java:307)
	at org.apache.hadoop.mapred.Child.main(Child.java:170)
{noformat}

I ran the same job with same records on left hand side and 100K records on right hand side. The job completed successfully.

> FR join should not spill
> ------------------------
>
>                 Key: PIG-1106
>                 URL: https://issues.apache.org/jira/browse/PIG-1106
>             Project: Pig
>          Issue Type: Bug
>            Reporter: Olga Natkovich
>            Assignee: Ankit Modi
>             Fix For: 0.7.0
>
>         Attachments: frjoin-nonspill.patch
>
>
> Currently, the values for the replicated side of the data are placed in a spillable bag (POFRJoin near line 275). This does not make sense because the whole point of the optimization is that the data on one side fits into memory. We already have a non-spillable bag implemented (NonSpillableDataBag.java) and we need to change FRJoin code to use it. And of course need to do lots of testing to make sure that we don't spill but die instead when we run out of memory

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.