Apparently your hbase starts to complain when the data to bulk load is big.
The client side log does not tell much. The real error should be in HBase
master and region servers logs. Try find there.

On Tue, Jun 23, 2015 at 10:39 PM, hongbin ma <[email protected]> wrote:

> can you share your hbase settings(number of region servers, heap size for
> each, etc.) and the size of your cube?
>
> On Tue, Jun 23, 2015 at 9:36 PM, alex schufo <[email protected]> wrote:
>
> > Hi,
> >
> > Thanks for the answer, actually it was maybe misleading from me to
> > formulate it like that but when I said "small", "medium" and "large"
> cubes
> > I was not talking about the advanced setting that can be set when
> creating
> > the cube but only about the cube actual size in terms of data (few days,
> > some weeks or several months of data).
> >
> > Anyway I tried to build the cube by setting the property to a "small
> cube"
> > profile and I still see the exact same issue.
> >
> > On Tue, Jun 23, 2015 at 3:14 AM, hongbin ma <[email protected]>
> wrote:
> >
> > > ​hi alex,
> > >
> > > "small​","medium" and "large" are three profiles which differs from
> each
> > > other in terns of hbase region split size.
> > >
> > > if I'm remebering correctly, small profile will generate htable with
> 10G
> > a
> > > region, 20G for medium profile, and 100G for large profile. (in v0.7.1)
> > >
> > > if your cube is not very big, you should stay with the small profile,
> if
> > > your cube is very large, you should think about the other two. In your
> > > case, when you use medium profile or large profile, your regions
> servers
> > > seemed not prepared for so large regions(maybe short of memory), please
> > > take a second look at:
> > >
> > > [LoadIncrementalHFiles-1]
> > >
> > > mapreduce.LoadIncrementalHFiles: HFile at
> > >
> > > hdfs://xxx:8020/tmp/kylin-b68bac23-ea82-4471-bc70
> > >
> > > -c144991fbbe0/smallCube/hfile/F1/e6ed300e4d9c41938d1ee474536c4fbf no
> > >
> > > longer fits inside a single region. Splitting...
> > >
> >
>
>
>
> --
> Regards,
>
> *Bin Mahone | 马洪宾*
> Apache Kylin: http://kylin.io
> Github: https://github.com/binmahone
>

Reply via email to