[
https://issues.apache.org/jira/browse/HIVE-12951?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15607154#comment-15607154
]
Richard Xin commented on HIVE-12951:
------------------------------------
Thanks Rui and Xuefu for your quick replies.
>> The right comparison is to measure the performance with and without
>> container prewarming to see if there is any benefit for your first run
we didn't feel any performance benefit with prewarming for the first run. I
assume that "first run" here refers to first run per hive session, instead of
first run after hive restarts.
We are trying to build a web service to run ad-hoc queries against hive, so
each call could be the "first call" for the session. do you have any
recommendations on how we should implement the solution? Based to our tests,
some sessions' first query under hive on spark is slower than hive on mr.
Thanks in advance for your insight.
> Reduce Spark executor prewarm timeout to 5s
> -------------------------------------------
>
> Key: HIVE-12951
> URL: https://issues.apache.org/jira/browse/HIVE-12951
> Project: Hive
> Issue Type: Bug
> Components: Spark
> Affects Versions: 1.2.0
> Reporter: Xuefu Zhang
> Assignee: Xuefu Zhang
> Fix For: 2.1.0
>
> Attachments: HIVE-12951.patch, HIVE-12951.patch
>
>
> Currently it's set to 30s, which tends to be longer than needed. Reduce it to
> 5s, only considering jvm startup time. (Eventually, we may want to make this
> configurable.)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
