Mailing-List: contact user-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@hbase.apache.org
MIME-Version: 1.0
In-Reply-To: <D2745073-DA52-44BF-8752-9B340845E05C@infor.com>
References: <6DF64C36-790D-419F-AC4D-B24BF6AD831E@infor.com>
 <CALte62wCus3=KdWXC1sKr8BfR34=CLCja3oyAmrqf-9xNFHcHQ@mail.gmail.com> <D2745073-DA52-44BF-8752-9B340845E05C@infor.com>
From: Ted Yu <yuzhihong@gmail.com>
Date: Fri, 20 Jan 2017 15:25:45 -0800
Message-ID: <CALte62wh5sqY1n-tG3_PEsgLSjn+uT2pN5SE6ZGBAt=DSD-TdQ@mail.gmail.com>
Subject: Re: Managed region Splitting
To: "user@hbase.apache.org" <user@hbase.apache.org>
Content-Type: multipart/alternative; boundary=f403045f719a24058c05468ef84b
archived-at: Fri, 20 Jan 2017 23:25:53 -0000

--f403045f719a24058c05468ef84b
Content-Type: text/plain; charset=UTF-8

For #1, have you searched / polled phoenix mailing list ? You may get more
advice there.

For #2, see HBASE-4365.

Have you read

http://hbase.apache.org/book.html#disable.splitting
http://hbase.apache.org/book.html#arch.regions.size

Cheers

On Fri, Jan 20, 2017 at 2:53 PM, Pradheep Shanmugam <
Pradheep.Shanmugam@infor.com> wrote:

> HI Ted,
>
> Is this a common solution when using salted tables to change the policy
> back to ConstantSizeRegionSplitPolicy.?
> Any reason as why the default policy was changes to
> IncreasingToUpperBoundRegionSplitPolicy?
> When I am doing manual splitting how do I choose the threshold?
>
> Thanks,
> Pradheep
>
>
>
>
> On 1/20/17, 5:41 PM, "Ted Yu" <yuzhihong@gmail.com> wrote:
>
> >For #1, you can consider plugging in ConstantSizeRegionSplitPolicy
> >for hbase.regionserver.region.split.policy
> >
> >For #2, regions are spread across servers. There is no centralized control
> >for the underlying table that prevents region splits from happening at the
> >same time.
> >
> >For #3, KeyPrefixRegionSplitPolicy is not effective for salted tables.
> >
> >Cheers
> >
> >On Fri, Jan 20, 2017 at 2:29 PM, Pradheep Shanmugam <
> >Pradheep.Shanmugam@infor.com> wrote:
> >
> >> Hi,
> >>
> >> 1. I have couple of phoenix tables with salting. I am assuming that all
> >> the regions will grow uniformly across the region server.
> >> Based on above I expect the region splitting to happen across all the
> >> region servers at the same time which will impact my performance when
> the
> >> region size gets bigger.
> >> I am considering manual region splitting to avoid this. But the given
> that
> >> the default split policy is IncreasingToUpperBoundRegionSplitPolicy<
> >> https://hbase.apache.org/devapidocs/org/apache/hadoop/
> hbase/regionserver/
> >> IncreasingToUpperBoundRegionSplitPolicy.html>,
> >> I cannot really disable the splitting by increasing the
> >> hbase.hregion.max.file size to say 100Gb as the the new split size is
> going
> >> to be the one set by the policy
> >> which will be less than 100 Gb several times and the automatic splitting
> >> is going to continue.
> >>
> >> 2. Or altogether should I not consider manual splitting as the chances
> of
> >> all regions splitting at the same time is not possible if there is
> going to
> >> be some internal check which will not trigger multiple split(or a
> number)
> >> for regions of the same table
> >>
> >> 3. Would using KeyPrefixRegionSplitPolicy<https://hbase.apache.org/
> >> devapidocs/src-html/org/apache/hadoop/hbase/regionserver/
> >> KeyPrefixRegionSplitPolicy.html#line.34> will be of any better than
> >> IncreasingToUpperBoundRegionSplitPolicy<https://hbase.
> >> apache.org/devapidocs/org/apache/hadoop/hbase/regionserver/
> >> IncreasingToUpperBoundRegionSplitPolicy.html> for salted tables.
> >>
> >> Please advice..
> >>
> >> Thanks,
> >> pradheep
> >>
>

--f403045f719a24058c05468ef84b--