[
https://issues.apache.org/jira/browse/AMBARI-5628?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Srimanth Gunturi updated AMBARI-5628:
-------------------------------------
Attachment: AMBARI-5628.patch
Review at https://reviews.apache.org/r/20871/
> Explicitly disabling datanucleus l2 cache for hive
> --------------------------------------------------
>
> Key: AMBARI-5628
> URL: https://issues.apache.org/jira/browse/AMBARI-5628
> Project: Ambari
> Issue Type: Bug
> Components: client
> Affects Versions: 1.5.1
> Reporter: Srimanth Gunturi
> Assignee: Srimanth Gunturi
> Fix For: 1.6.1
>
> Attachments: AMBARI-5628.patch
>
>
> Ambari installations of hive currently do not set any datanucleus related
> properties. There is such a thing as a datanucleus l2 cache, that is pretty
> bad for hive in a distributed environment if it is set. (If there is a lone
> embedded hive instance, with no other codepaths to the db, then it's fine,
> but that never happens in a distributed environment.)
> By default, if no setting is present, datanucleus defaults the l2 cache to
> being on, so hive ups the ante by defaulting to turning it off by default if
> no other setting is configured.
> Now, in a war of "defaults", the hive default should win, but this is an area
> where we have had recurring support issues from clients that turn it on
> expecting improved performance. Thus, I'd like ambari installed hive-site.xml
> to explicitly have this config parameter turned off, with a comment asking
> users to not switch it on as it impacts hive negatively.
> The parameter in question is "datanucleus.cache.level2.type" , and it's value
> should be "none". (Note that I've seen some older configs that seem to do
> things like turning datanucleus.cache.level2 = false and stuff like that,
> that is bogus config and does nothing and should not be assumed to be a
> catch-all enabler.)
> As a comment, I'd like the following comment "Disables datanucleus l2 cache.
> This must be set to 'none' for hive to work properly" or something to that
> effect.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
