You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "stack (JIRA)" <ji...@apache.org> on 2011/06/11 01:58:59 UTC

[jira] [Resolved] (HBASE-1787) New TableServers instance if new HBaseConfiguration

     [ https://issues.apache.org/jira/browse/HBASE-1787?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

stack resolved HBASE-1787.
--------------------------

    Resolution: Won't Fix

This issue is made redundant by Karthick work over in hbase-3777

> New TableServers instance if new HBaseConfiguration
> ---------------------------------------------------
>
>                 Key: HBASE-1787
>                 URL: https://issues.apache.org/jira/browse/HBASE-1787
>             Project: HBase
>          Issue Type: Bug
>            Reporter: stack
>
> Down in guts of TableServers is a static map keyed by HBaseConfiguration instance.  If in such as HTable constructor, you keep passing the same HBaseConfiguration instance, then you will get back a previously made TableServers with cached region locations, zk connection setup, etc.   If you pass a new HBaseConfiguration each time, then on each new HTable, a new TableServers instance will be built with attendant zk setup and resource costs etc.
> Users are surprised by this behavior: "What seems like a bug to me is that the configuration object is caching state.  I understand if HTable hangs on to connections until it is closed.  And I understand how calling new HTABLE twice might open a new connection.  That all seems okay.  What I wouldn't expect is that using an old configuration object will give me different behavior from using a new one.  A configuration seems like a static thing to me.  It should only change if I change it." -- Our Jim Firby

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira