[
https://issues.apache.org/jira/browse/HIVE-2316?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Patrick Angeles updated HIVE-2316:
----------------------------------
Description:
With hive.auto.convert.join = true, it appears that when the small tables are
compressed, the optimizer cannot properly guess the size of the in-memory
tables. It proceeds with a MapJoin based on the compressed size, instead of
falling back to a CommonJoin.
This can sometimes result in a OOME from the map tasks.
was:With hive.auto.convert.join = true, it appears that when the small tables
are compressed, the optimizer cannot properly guess the size of the in-memory
tables. It proceeds with a MapJoin based on the compressed size, instead of
falling back to a CommonJoin.
> MapJoin optimizer can make the wrong decision on compressed tables
> ------------------------------------------------------------------
>
> Key: HIVE-2316
> URL: https://issues.apache.org/jira/browse/HIVE-2316
> Project: Hive
> Issue Type: Bug
> Reporter: Patrick Angeles
>
> With hive.auto.convert.join = true, it appears that when the small tables are
> compressed, the optimizer cannot properly guess the size of the in-memory
> tables. It proceeds with a MapJoin based on the compressed size, instead of
> falling back to a CommonJoin.
> This can sometimes result in a OOME from the map tasks.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
