hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ferdy Galema <ferdy.gal...@kalooga.com>
Subject Re: Harvesting empty regions
Date Tue, 31 May 2011 12:06:13 GMT
You can use the merge tool to combine adjacent regions. It requires a 
bit of manual work because you need to specify the regions by hand. The 
cluster also needs to be offline (I recommend to keep zookeeper running 
though). Check if merging succeeded with the hbck tool.

There are some jira issues that address the current drawbacks:
https://issues.apache.org/jira/browse/HBASE-1621

On 05/31/2011 08:20 AM, Arvind Jayaprakash wrote:
> My setup seems to have a lot of regions with no data that just keep
> accumulating over time. Here are some details:
>
> I have time-series data (created by opentsdb) being inserted into hbase
> every minute. Since the data has little value after say 15 days, I go
> ahead and delete all old data.
>
> When I look at /regionserver.jsp, I see a lot of regions being listed
> that I know has no data. To confirm this hypothesis, I ran a query on
> the shell with the start and end key of a given region and I get zero
> rows: http://pastebin.ca/2072627 . The value of the start&  end row is
> something that the regionservier reported for a specific region. I've
> run this test over multiple regions that I suspected to have zero rows
> and I get the same behaviour.
>
> Explicitly running a major compaction does not seem to help either. So
> my question is that once a region is created, does it live forever even
> if there are zero rows present in that region? I ask because I am
> worried that the master is having to carry the burden of empty regions
> as time passes and the list will just keep building forever.

Mime
View raw message