[
https://issues.apache.org/jira/browse/HIVE-790?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12749757#action_12749757
]
Ning Zhang commented on HIVE-790:
---------------------------------
Tested performance on a larger data set: the base table is 47 million rows of
ad_imps_2 with compressed storage of 89MB. The query is
insert overwrite table Tmp select * from (select * from Base union all select *
from Base) s;
With synchronized UnionOperator.processOp(), mapper 1 got average time of
476.05 secs, mapper2 average time = 648.5 secs
With non-sync UnionOperator.processOp(), mapper 1 average = 476.34 secs,
mapper2 average time = 655.117 sec.
The union operator is executed in mapper 1 and there is almost no difference by
introducing the synchronized keyword. So no new JIRA is needed for performance
regression.
> race condition related to ScriptOperator + UnionOperator
> --------------------------------------------------------
>
> Key: HIVE-790
> URL: https://issues.apache.org/jira/browse/HIVE-790
> Project: Hadoop Hive
> Issue Type: Bug
> Reporter: Zheng Shao
> Assignee: Ning Zhang
> Attachments: Hive-790.patch, Hive-790_2.patch, Hive-790_3.patch,
> Hive-790_4.patch
>
>
> ScriptOperator uses a second thread to output the rows to the children
> operators. In a corner case which contains a union, 2 threads might be
> outputting data into the same operator hierarchy and caused race conditions.
> {code}
> CREATE TABLE tablea (cola STRING);
> SELECT *
> FROM (
> SELECT TRANSFORM(cola)
> USING 'cat'
> AS cola
> FROM tablea
> UNION ALL
> SELECT cola as cola
> FROM tablea
> ) a;
> {code}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
