HBase Compaction Issues

[joseph <jo...@arssoftware.com>]

Hi,

 

We are using HBase 0.94 and Hadoop 2.0.0 .We are using CDH 4.7

 

We have a  4 node cluster with one master and the 3 region servers

 

We Have  time series data.  We have row key with the format of
Reversetimestamp_id. We have client connection timeout in the modules and
constantly facing compaction queue size alert in the Cloudera manager. We
disable compaction by changing the settings in the Cloudera HBase
configuration .Then we are compact tables manually using the cron scheduler.
But still compaction are triggering and the Cloudera manager always shows
compaction queue size alert .Modules still facing timeout with the HBase

 

 

Regards,

Joseph

 

Re: HBase Compaction Issues

[Dima Spivak <ds...@cloudera.com>]

Hey Joseph,

I suspect you're hitting hotspotting (see
http://hbase.apache.org/book.html#rowkey.design). Moving forward, you might
want to consider something like OpenTSDB for your time series data.

-Dima

On Mon, Feb 2, 2015 at 1:18 AM, joseph <jo...@arssoftware.com> wrote:

> Hi,
>
>
>
> We are using HBase 0.94 and Hadoop 2.0.0 .We are using CDH 4.7
>
>
>
> We have a  4 node cluster with one master and the 3 region servers
>
>
>
> We Have  time series data.  We have row key with the format of
> Reversetimestamp_id. We have client connection timeout in the modules and
> constantly facing compaction queue size alert in the Cloudera manager. We
> disable compaction by changing the settings in the Cloudera HBase
> configuration .Then we are compact tables manually using the cron
> scheduler.
> But still compaction are triggering and the Cloudera manager always shows
> compaction queue size alert .Modules still facing timeout with the HBase
>
>
>
>
>
> Regards,
>
> Joseph
>
>
>
>