[
https://issues.apache.org/jira/browse/ACCUMULO-2831?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14005496#comment-14005496
]
ASF subversion and git services commented on ACCUMULO-2831:
-----------------------------------------------------------
Commit db1b63c4cb9f2f1bfdb45765b9c79f84105d601a in accumulo's branch
refs/heads/1.6.1-SNAPSHOT from [~dlmarion]
[ https://git-wip-us.apache.org/repos/asf?p=accumulo.git;h=db1b63c ]
ACCUMULO-2831: Reset the classloader on the Hadoop Configuration objects that
were created before the VFS ClassLoader was created.
> HDFS Configuration classloader not correct when using VFSClassLoader
> --------------------------------------------------------------------
>
> Key: ACCUMULO-2831
> URL: https://issues.apache.org/jira/browse/ACCUMULO-2831
> Project: Accumulo
> Issue Type: Bug
> Components: start
> Affects Versions: 1.6.0
> Reporter: Dave Marion
> Assignee: Dave Marion
> Labels: classloader
> Fix For: 1.6.1
>
>
> When using hdfs to load jars with the VFS classloader, the Hadoop FileSystem
> object is created before the VFS classloader. The FileSystem object and
> associated Configuration object are cached by the DFS Client and use the
> thread context classloader at the time of creation, which is not the VFS
> class loader. This leads to Accumulo classes not being found when they are
> loaded via the Configuration object, such as when WAL recovery is performed.
> The WAL is read using a SequenceFile$Reader object, which tries to load the
> Key and Value classes for the SequenceFile using
> WritableName.getClass(String, Configuration)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
