[jira] [Updated] (HDFS-3664) BlockManager race when stopping active services

["Colin Patrick McCabe (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HDFS-3664?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Colin Patrick McCabe updated HDFS-3664:
---------------------------------------

    Attachment: HDFS-3664.001.patch

* Previously, we were setting blocksMap to {{null}} before calling {{datanodeManager#close}}.  However, {{DatanodeManager#decommissionThread}} accesses blocksMap at various times while the thread runs.  Instead, wait until the decommissionThread has been stopped before closing blocksMap.

* We also previously were not joining {{DatanodeManager#decommissionThread}} or {{DatanodeManager#heartbeatManager}}, so there was no guarantee those threads would have exited before blocksMap was nulled.  Let's join these threads.
                
> BlockManager race when stopping active services
> -----------------------------------------------
>
>                 Key: HDFS-3664
>                 URL: https://issues.apache.org/jira/browse/HDFS-3664
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: test
>    Affects Versions: 2.0.0-alpha
>            Reporter: Eli Collins
>         Attachments: HDFS-3664.001.patch
>
>
> Looks like there's a race hit by the test where we try to do block management while shutting down (eg allocate a block or compute replication work after BlocksMap#close, resulting in an NPE).

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira