[
https://issues.apache.org/jira/browse/HBASE-6820?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13861317#comment-13861317
]
chendihao commented on HBASE-6820:
----------------------------------
[~enis] [~stack] I'm working with MiniZooKeeperCluster and glad to see the
patch. But I don't think it's done completely.
KillCurrentActiveZooKeeperServer() and killOneBackupZooKeeperServer() will
shutdown the ZooKeeperServer and need to close ZKDatabase as well. I open a new
jira https://issues.apache.org/jira/browse/HBASE-10274 and it would be nice to
fix it.
> [WINDOWS] MiniZookeeperCluster should ensure that ZKDatabase is closed upon
> shutdown()
> --------------------------------------------------------------------------------------
>
> Key: HBASE-6820
> URL: https://issues.apache.org/jira/browse/HBASE-6820
> Project: HBase
> Issue Type: Bug
> Affects Versions: 0.94.3, 0.95.2
> Reporter: Enis Soztutar
> Assignee: Enis Soztutar
> Labels: windows
> Fix For: 0.95.0
>
> Attachments: hbase-6820_v1-0.94.patch, hbase-6820_v1-trunk.patch
>
>
> MiniZookeeperCluster.shutdown() shuts down the ZookeeperServer and
> NIOServerCnxnFactory. However, MiniZookeeperCluster uses a deprecated
> ZookeeperServer constructor, which in turn constructs its own FileTxnSnapLog,
> and ZKDatabase. Since ZookeeperServer.shutdown() does not close() the
> ZKDatabase, we have to explicitly close it in MiniZookeeperCluster.shutdown().
> Tests effected by this are
> {code}
> TestSplitLogManager
> TestSplitLogWorker
> TestOfflineMetaRebuildBase
> TestOfflineMetaRebuildHole
> TestOfflineMetaRebuildOverlap
> {code}
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
