And as stack mentioned in another email, you may not have actually
lost data but only suffered a benign region assignment collision
(lucky), leaving a stray mapfile directory in place which can be
removed so the region servers will stop getting confused. Before
you remove the directory, be sure there really is no /data and 
/index files in there.

  - Andy

> From: Andrew Purtell <[email protected]>
> Subject: Re: corrupt hbase meta
> To: [email protected]
> Date: Thursday, January 8, 2009, 5:30 PM
> Hi Dru,
> 
> You might try 
> 
>   % hadoop fs -rmr
> hdfs://dev-hadoop0.brandaffinity.net:9000/hbase/talentmaps/2022550034/base/mapfiles/7007481172739780749/
> 
> to at least partially recover the region. Don't try
> this until someone else weighs in on this approach.
> 
> Since you've already suffered data loss, possibly as a
> result of HBASE-1104, I do not think there is any harm to
> try it. 
> 
>    - Andy



      

Reply via email to