[ 
https://issues.apache.org/jira/browse/HBASE-11158?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13998293#comment-13998293
 ] 

Jean-Marc Spaggiari edited comment on HBASE-11158 at 5/15/14 11:28 AM:
-----------------------------------------------------------------------

{quote}
It can't check the whole cluster though JMS, right?
{quote}

I agree. Just trying to find a way to bring the alert back to the user without 
having to check the whole cluster.  I have not found any good idea to achieve 
that for now. But I got caught already twice because of that, so trying to 
figure if there is anything doable to alert user about it. Maybe with another 
command line? Still thinking about it.


was (Author: jmspaggi):
[quote]
It can't check the whole cluster though JMS, right?
[quote]

I agree. Just trying to find a way to bring the alert back to the user without 
having to check the whole cluster.  I have not found any good idea to achieve 
that for now. But I got caught already twice because of that, so trying to 
figure if there is anything doable to alert user about it. Maybe with another 
command line? Still thinking about it.

> bin/hbase upgrade -check should also check compression codecs
> -------------------------------------------------------------
>
>                 Key: HBASE-11158
>                 URL: https://issues.apache.org/jira/browse/HBASE-11158
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 0.99.0
>            Reporter: Jean-Marc Spaggiari
>
> When upgrading insitu from 0.94 to 0.96 or 0.98 codecs are usually already 
> there in the servers so it's all fine.
> But when doing an upgrade by moving the data from one cluster to a brand new 
> one, compression codecs might be missing.
> bin/hbase upgrade -check will not report any missing codec, but HBase will 
> not be able to start after the upgrade because the codes are missing.
> I think bin/hbase upgrade -check should check the compression codecs 
> configured on the tables and make sure they are available on the new cluster. 
> f not, it should be reported.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to