[
https://issues.apache.org/jira/browse/PIG-5203?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15947180#comment-15947180
]
Adam Szita commented on PIG-5203:
---------------------------------
It seems like that in case CombinerOptimizer usage is triggered, the value of
requested parallelism is not propagated correctly to the ReduceBy operator.
It is overwritten by -1 coming from the foreach operator.
Attached patch to fix this: [^PIG-5203.0.patch]
[~kellyzly] can you take a look please?
> Partitioner E2E test fails on spark
> -----------------------------------
>
> Key: PIG-5203
> URL: https://issues.apache.org/jira/browse/PIG-5203
> Project: Pig
> Issue Type: Sub-task
> Components: spark
> Reporter: Adam Szita
> Assignee: Adam Szita
> Fix For: spark-branch
>
> Attachments: PIG-5203.0.patch
>
>
> Cause of failure is the difference between expected and actual result:
> {color:red}Actual:
> 18 160
> 19 164
> 20 166
> ...
> {color}{color:green}
> Expected:
> 18 80
> 18 80
> 19 82
> 19 82
> ....
> {color}
> Seems like although we specify {{PARTITION BY
> org.apache.pig.test.utils.SimpleCustomPartitioner2 parallel 2}} in the query
> it has no effect in Spark mode.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
