hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "punwinger (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-13437) ThriftServer leaks ZooKeeper connections
Date Fri, 10 Apr 2015 11:57:14 GMT

    [ https://issues.apache.org/jira/browse/HBASE-13437?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14489486#comment-14489486
] 

punwinger commented on HBASE-13437:
-----------------------------------

I test master and branch-1.0 in my test cluster and the problem still exist. So I read their
codebase and as [~ndimiduk] said master and branch-1 use the new interface ClusterConnection
to getTable directly. But the problem is ConnectionCache doesn't close the connection as it
manager ClusterConnection to do the cache thing. And I also test branch-1.0 which also leaks
connection. So I suggest this patch should apply to all these branches.

> ThriftServer leaks ZooKeeper connections
> ----------------------------------------
>
>                 Key: HBASE-13437
>                 URL: https://issues.apache.org/jira/browse/HBASE-13437
>             Project: HBase
>          Issue Type: Bug
>          Components: Thrift
>    Affects Versions: 0.98.8
>            Reporter: punwinger
>         Attachments: hbase-13437-fix.patch
>
>
> HBase ThriftServer will cache Zookeeper connection in memory using org.apache.hadoop.hbase.util.ConnectionCache.
This class has a mechanism called chore to clean up connections idle for too long(default
is 10 min). But method timedOut for testing whether idle exceed for maxIdleTime always return
false which leads to never release the Zookeeper connection. If we send request to ThriftServer
every maxIdleTime then ThriftServer will keep thousands of Zookeeper Connection soon.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message