hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Xuefu Zhang (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HIVE-15543) Don't try to get memory/cores to decide parallelism when Spark dynamic allocation is enabled
Date Thu, 05 Jan 2017 03:29:58 GMT

     [ https://issues.apache.org/jira/browse/HIVE-15543?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Xuefu Zhang updated HIVE-15543:
-------------------------------
    Attachment: HIVE-15543.patch

cc: [~csun], [~ruili]

> Don't try to get memory/cores to decide parallelism when Spark dynamic allocation is
enabled
> --------------------------------------------------------------------------------------------
>
>                 Key: HIVE-15543
>                 URL: https://issues.apache.org/jira/browse/HIVE-15543
>             Project: Hive
>          Issue Type: Improvement
>          Components: Spark
>    Affects Versions: 2.2.0
>            Reporter: Xuefu Zhang
>            Assignee: Xuefu Zhang
>         Attachments: HIVE-15543.patch
>
>
> Presently Hive tries to get numbers for memory and cores from the Spark application and
use them to determine RS parallelism. However, this doesn't make sense when Spark dynamic
allocation is enabled because the current numbers doesn't represent available computing resources,
especially when SparkContext is initially launched.
> Thus, it makes send not to do that when dynamic allocation is enabled.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message