[jira] [Commented] (SPARK-1879) Default PermGen size too small when using Hadoop2 and Hive

["Matei Zaharia (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-1879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14001451#comment-14001451 ] 

Matei Zaharia commented on SPARK-1879:
--------------------------------------

BTW the warning on Java 8 is the following:

{code}
Java HotSpot(TM) 64-Bit Server VM warning: ignoring option MaxPermSize=128m; support was removed in 8.0
{code}

Not *that* scary. I think we can live with this for a bit and possibly test java -version in a later release.

> Default PermGen size too small when using Hadoop2 and Hive
> ----------------------------------------------------------
>
>                 Key: SPARK-1879
>                 URL: https://issues.apache.org/jira/browse/SPARK-1879
>             Project: Spark
>          Issue Type: Bug
>          Components: Spark Core
>            Reporter: Matei Zaharia
>            Assignee: Matei Zaharia
>            Priority: Critical
>             Fix For: 1.0.0
>
>
> If you launch a spark-shell with Hadoop 2 and Hive on the classpath, and try to use Hive therein, the PermGen quickly reaches 85 MB after a few commands, at which point Java gives up and freezes. We should pass a MaxPermSize to prevent this. Unfortunately passing this results in a warning on Java 8, but that's still better than not passing it.
> I don't think this affects stuff other than the shell; it's just the combination of Scala compiler + Hive + Hadoop 2 that pushes things over the edge.



--
This message was sent by Atlassian JIRA
(v6.2#6252)