You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@hbase.apache.org by Gevorg Hari <ge...@gmail.com> on 2015/10/15 23:51:34 UTC

Verifying the distribution of data across region servers - hotspotting

Hello,

I'm afraid that my cluster is suffering of a bit of hotspotting, what's the
best way to verify that?

Is it enough to verify the data size handled by each regionserver with the
hdfs dfs -du command? Is there a way to easily count the number of records
handled by each regionserver? How about the number of read/writes handled
by each of them?

Thanks!
Gevorg

Re: Verifying the distribution of data across region servers - hotspotting

Posted by Ted Yu <yu...@gmail.com>.

Please read http://hbase.apache.org/book.html#_hotspotting , if you haven't.

Cheers

On Thu, Oct 15, 2015 at 3:15 PM, Ted Yu <yu...@gmail.com> wrote:

> Here're a few metrics (per server) to consider for finding hot spot:
>
> read request count
> write request count
> compaction queue size
> memstore size
>
> Cheers
>
> On Thu, Oct 15, 2015 at 2:51 PM, Gevorg Hari <ge...@gmail.com> wrote:
>
>> Hello,
>>
>> I'm afraid that my cluster is suffering of a bit of hotspotting, what's
>> the
>> best way to verify that?
>>
>> Is it enough to verify the data size handled by each regionserver with the
>> hdfs dfs -du command? Is there a way to easily count the number of records
>> handled by each regionserver? How about the number of read/writes handled
>> by each of them?
>>
>> Thanks!
>> Gevorg
>>
>
>

Re: Verifying the distribution of data across region servers - hotspotting

Posted by Ted Yu <yu...@gmail.com>.

Here're a few metrics (per server) to consider for finding hot spot:

read request count
write request count
compaction queue size
memstore size

Cheers

On Thu, Oct 15, 2015 at 2:51 PM, Gevorg Hari <ge...@gmail.com> wrote:

> Hello,
>
> I'm afraid that my cluster is suffering of a bit of hotspotting, what's the
> best way to verify that?
>
> Is it enough to verify the data size handled by each regionserver with the
> hdfs dfs -du command? Is there a way to easily count the number of records
> handled by each regionserver? How about the number of read/writes handled
> by each of them?
>
> Thanks!
> Gevorg
>