Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HBase, mail # user - Duplicates region after upgrade to 0.89


Copy link to this message
-
Re: Duplicates region after upgrade to 0.89
Andrey Stepachev 2010-09-27, 19:24
this is restored data (before any upates
and all region looks good)

http://paste.ubuntu.com/501716/

2010/9/27 Andrey Stepachev <[EMAIL PROTECTED]>:
> Thanks for reply Stack.
>
> How data was imported and chaged (full story).
> 1. stop 0.20.6.
> 2. copy /hbase into new location (where 0.89 lives)
> 3. start 0.89
> 4. run heavy update task
> ....
>
> scan output:
> http://paste.ubuntu.com/501680/
>
> --fix output:
> http://paste.ubuntu.com/501684/
>
> Row count is bigger then should. ~6.3mil instead of 5.4mil
> (looks like this regions are copy of data was writer on update,
> but instead of write to existing region, hbase create two new regions)
>
> 2010/9/27 Stack <[EMAIL PROTECTED]>:
>> Can you scan your table? (Try a count from the shell or use rowcounter
>> mapreduce job).
>>
>> What happens when you do check_meta.rb --fix?
>>
>> Can you do a:
>>
>> $ echo "scan '.META.'" | ./bin/hbase shell &> /tmp/meta.txt
>>
>> ... and pastebin the /tmp/meta.txt so I can take a look?
>>
>> St.Ack
>>
>> On Mon, Sep 27, 2010 at 6:42 AM, Andrey Stepachev <[EMAIL PROTECTED]> wrote:
>>> After upgrade from 0.20.6 to 0.89 (stumbleupon version)
>>> I mentioned, that i have duplicate entries in .META.
>>> (encoded names are different, but keys are identical)
>>>
>>> check_meta.rb sad :
>>>
>>>
>>> What I can do to recover from this situation? Export/Import only?
>>> And if that, I must export/import production database to be able to
>>> use 0.89?
>>>
>>>
>>> Andrey.
>>>
>>
>