You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Szehon Ho (JIRA)" <ji...@apache.org> on 2016/03/17 21:36:33 UTC
[jira] [Commented] (HIVE-13217) Replication for HoS mapjoin small
file needs to respect dfs.replication.max
[ https://issues.apache.org/jira/browse/HIVE-13217?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15200322#comment-15200322 ]
Szehon Ho commented on HIVE-13217:
----------------------------------
Thanks Chinna for working on the patch.
For this line:
if (minReplication > dfsMaxRep)
minReplication = dfsMaxRep;
Is it possible to change it to minReplication = Math.max(minReplication, dfsMaxRep) (for clarity), and also put a comment explaining why? Thanks
> Replication for HoS mapjoin small file needs to respect dfs.replication.max
> ---------------------------------------------------------------------------
>
> Key: HIVE-13217
> URL: https://issues.apache.org/jira/browse/HIVE-13217
> Project: Hive
> Issue Type: Bug
> Components: Spark
> Affects Versions: 1.2.1, 2.0.0
> Reporter: Szehon Ho
> Assignee: Chinna Rao Lalam
> Priority: Minor
> Attachments: HIVE-13217.1.patch
>
>
> Currently Hive on Spark Mapjoin replicates small table file to a hard-coded value of 10. See SparkHashTableSinkOperator.MIN_REPLICATION.
> When dfs.replication.max is less than 10, HoS query fails. This constant should cap at dfs.replication.max.
> Normally dfs.replication.max seems set at 512.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)