You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by "Kannan Muthukkaruppan (JIRA)" <ji...@apache.org> on 2010/01/29 23:51:34 UTC
[jira] Commented: (HBASE-1679) Flapping DNS does us more harm than
it need to
[ https://issues.apache.org/jira/browse/HBASE-1679?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12806543#action_12806543 ]
Kannan Muthukkaruppan commented on HBASE-1679:
----------------------------------------------
Yup, this is pretty much the issue we ran into.
And when the regionserver was asked to restarted, it registered itself again under /hbase/rs in zookeeper using a new startcode as the element (znode) name.
> Flapping DNS does us more harm than it need to
> ----------------------------------------------
>
> Key: HBASE-1679
> URL: https://issues.apache.org/jira/browse/HBASE-1679
> Project: Hadoop HBase
> Issue Type: Bug
> Reporter: stack
> Fix For: 0.21.0
>
>
> Over in HBASE-1675, JSharp has posted logs where a temporary DNS outage does his cluster a death blow.
> When cluster members report in, the master composes the regionserver name by doing a hostname lookup and appending it to port and startcode passed over by the regionserver. The host lookup during a DNS outage when from name to IP. Master then thought this regionserver an unknown host and told it restart.... and so on.
> If the regionserver composed its name once, it could pass this the master and avoid a DNS lookup per regionserver report.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.