You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by "Jean-Daniel Cryans (JIRA)" <ji...@apache.org> on 2009/07/07 16:31:15 UTC
[jira] Commented: (HBASE-1614) single zk node buckling under small
node? connections never timing out?
[ https://issues.apache.org/jira/browse/HBASE-1614?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12728122#action_12728122 ]
Jean-Daniel Cryans commented on HBASE-1614:
-------------------------------------------
It is important to remind that the ZK doc is very clear that a ZK server should have its own disk and even its own machine. I bet that people put the ZK servers on the same machines as the RS servers which is a cheap way to make it work but not really a good way. So a single ZK instance on the same node as a RS/TT/DN is probably the worst setup ever.
> single zk node buckling under small node? connections never timing out?
> ------------------------------------------------------------------------
>
> Key: HBASE-1614
> URL: https://issues.apache.org/jira/browse/HBASE-1614
> Project: Hadoop HBase
> Issue Type: Bug
> Reporter: stack
> Fix For: 0.20.0
>
>
> A few (Irfan on the list and jgray) report that a single node zk seems insufficient supporting small cluster < 20 nodes. This seems odd (though, yeah, we're recommending serious deploys must have a quorum of >= 3 zk nodes). Also, reports that all connections fill up and don't time out though job is done (on IRC). Take a look.
> (Check zk stat to see if it'll dump connections)
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.