hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ryan Rawson <ryano...@gmail.com>
Subject Re: Two regions with same start_key
Date Tue, 02 Mar 2010 20:44:45 GMT
Yes, I had some fun and messed up regions, and I managed to fix them
all using the technique I described.  I think I ended up doing
something like 20 pair-wise merges to accomplish everything.

Good luck, let us know how it goes!
-ryan

On Tue, Mar 2, 2010 at 12:17 AM, Manuel de Ferran
<manuel.deferran@gmail.com> wrote:
> Thanks !
>
> I was a bit worried merging regions which do not follow strictly the B
> -> C ; C -> D pattern.
>
> On Mon, Mar 1, 2010 at 8:58 PM, Ryan Rawson <ryanobjc@gmail.com> wrote:
>> I ran into these kinds of issues testing a pre-alpha hbase back in
>> "the day"... If you have multiple regions that are weirdly
>> overlapping, maybe as such:
>>
>> A -> B
>> B -> C
>> B -> D
>> C -> F
>> F -> G
>> G -> H
>>
>> Here we see there are 3 "weird" regions, the [B,C) [B,D) and [C,F)
>> regions... taken together they seem to form a reasonable continuous
>> collection of keys, with the regions before and after them perfectly
>> lining up.  The solution in this case is to do 2 merges: [B,C) & [B,D)
>> -> [B,D) & [C,F)
>>
>> You can merge a lot of regions, my record is about 10 or so.  The
>> normal system will split if necessary after booting up hbase again.
>>
>> One last tip, before you take hbase down, note the region names, since
>> it might be hard to do so when hbase is down.
>>
>> -ryan
>>
>> On Sat, Feb 27, 2010 at 5:04 AM, Stack <stack@duboce.net> wrote:
>>> Is it a failed split?  Its not an offlined parent with split
>>> daughters?  It doesn't seem so.  I'd grep logs to see if I could
>>> figure history of the regions but seems like you did that already.
>>> You could merge the regions without data loss.  You need to first take
>>> the table offline and then run the merge tool.
>>>
>>> hbase> disable "TABLENAME"
>>>
>>> Then once its offline:
>>>
>>> ./bin/hbase org.apache.hadoop.hbase.util.Merge
>>>
>>> That should printout usage.
>>>
>>> Good luck,
>>> St.Ack
>>>
>>> On 2/25/10, Manuel de Ferran <manuel.deferran@gmail.com> wrote:
>>>> Greetings,
>>>>
>>>> While browning a table, I noticed a strange thing in a couple of
>>>> regions. I have two regions with same start_key, and two others with
>>>> same end_key.
>>>>
>>>> Here are an extract of my regions list (s stands for 'start_key', and
>>>> e for 'end_key') :
>>>>
>>>> R0 s : 1263184838004\x2F354030030657380\x2F3873287876323769 e:
>>>> 1263192001699\x2Fa1000007c7afc6\x2F3880449775843972
>>>>
>>>> R1 s : 1263192001699\x2Fa1000007c7afc6\x2F3880449775843972 e :
>>>> 1263204796594\x2F354030030723539\x2F3893245091353305
>>>>
>>>> R2 s : 1263192001699\x2Fa1000007c7afc6\x2F3880449775843972 e :
>>>> 1263224608625\x2F358279017620731\x2F3913046476173954
>>>>
>>>> R3 s : 1263204796594\x2F354030030723539\x2F3893245091353305 e:
>>>> 1263224608625\x2F358279017620731\x2F3913046476173954
>>>>
>>>> R4 s : 1263224608625\x2F358279017620731\x2F3913046476173954 e:
>>>> 1263236230681\x2F351680031004439\x2F3924673427948082
>>>>
>>>> R5 s : 1263236230681\x2F351680031004439\x2F3924673427948082 e :
>>>> 1263246691649\x2F358279018935781\x2F3935132932584112
>>>>
>>>>
>>>> I expect contiguous regions. But I have smth like this :
>>>> R0 -> R1 and R2
>>>> R1 -> R3
>>>> R2 -> R4
>>>> R3 -> R4
>>>> R4 -> R5
>>>>
>>>> instead of R0 -> R1 -> R2 ...
>>>>
>>>> I could not find anything in logfiles that could explain this "failed
>>>> split". I was wondering how bad is this situation ? and how I could
>>>> fix it ?
>>>>
>>>>
>>>> Thanks
>>>>
>>>
>>
>

Mime
View raw message