[
https://issues.apache.org/jira/browse/HIVE-10551?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14571430#comment-14571430
]
Vikram Dixit K commented on HIVE-10551:
---------------------------------------
[~rajesh.balamohan] This OOM is happening in the mapper? Is this when re-using
a container that was used in the mapper phase? What is the Xmx value used for
the container? Can you provide the heap dump for this please. From the looks of
it, the group by operator probably decided to flush its results and that could
have resulted in the ReduceSinkOperator requesting for the sort buffers which
could have pushed this over the edge. Can you also post the stack trace here
for reference please.
> OOM when running query_89 with vectorization on & hybridgrace=false
> -------------------------------------------------------------------
>
> Key: HIVE-10551
> URL: https://issues.apache.org/jira/browse/HIVE-10551
> Project: Hive
> Issue Type: Bug
> Reporter: Rajesh Balamohan
> Attachments: HIVE-10551-explain-plan.log, hive_10551.png
>
>
> - TPC-DS Query_89 @ 10 TB scale
> - Trunk version of Hive + Tez 0.7.0-SNAPSHOT
> - Additional settings ( "hive.vectorized.groupby.maxentries=1024 ,
> tez.runtime.io.sort.factor=200 tez.runtime.io.sort.mb=1800
> hive.tez.container.size=4096 ,hive.mapjoin.hybridgrace.hashtable=false" )
> Will attach the profiler snapshot asap.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
