Sounds good. I'll wipe it out and start over after changing the xceiver
parameters.

Thanks a lot for the help (you too Stack)!

Larry

On Thu, Jan 29, 2009 at 4:22 PM, Andrew Purtell <[email protected]> wrote:

> Hi Larry,
>
> A DFS error has caused permanent corruption. HBase is currently
> at the mercy of DFS limitations.
>
> Given that the mapfile for your root region has been corrupted,
> the easiest course for "recovery" is to delete all HBase files
> in DFS and start over. It is on the roadmap to provide a "hbase
> fsck" to deal with these issues in the future. If you have data
> that you would like to try harder to preserve, perhaps it would
> be possible to copy the 'data' and 'index' files for the ROOT
> region of a pristine installation on top of the corrupt ones in
> your installation, restart, and have it work out. But I suspect
> there is file level corruption beyond your ROOT region, or,
> perhaps you will be lucky.
>
> Before restarting, you should try increasing the number of
> configured datanode xceivers. I run with 2048. Also, given you
> have only a 4 node cluster, you may find it is just too small
> to spread the DFS load and more nodes will need to be added.
>
>   - Andy
>
> > From: Larry Compton
> > NativeException:
> > org.apache.hadoop.hbase.client.NoServerForRegionException:
> > Timed out trying to locate root region
>
>
>
>
>

Reply via email to