Hello All,

I am looking at the replication functionality in HBase from the
perspective of failover/failback mechanisms, i.e., fail over to a
slave cluster in case the master (or primary) cluster fails, and then
reverting back to the original master when it is up.

By looking at the code, and replication page [1], I think the current
replication functionality supports replicating a healthy cluster to a
backup cluster (irrespective of any cluster size). This is not a exact
failover/failback process, but a backup and recovery tool.
Please correct me if I am wrong.
[1] http://hbase.apache.org/replication.html

Thanks,
Himanshu

I would say that the primary function is being a "live backup" but
basically everything you do with MySQL slaves you can do with HBase
slave clusters:

 - Read-only store for colder data
 - Mining (MR or else)
 - Backup (and eventually we could add a delayed backup feature too)
 - Do Master-Master in between different sites

You could also use it as a failover, but you must plan how you want to
do the reconciliation.

J-D

On Fri, Feb 24, 2012 at 1:08 PM, Himanshu Vashishtha
<[EMAIL PROTECTED]> wrote:
> Hello All,
>
> I am looking at the replication functionality in HBase from the
> perspective of failover/failback mechanisms, i.e., fail over to a
> slave cluster in case the master (or primary) cluster fails, and then
> reverting back to the original master when it is up.
>
> By looking at the code, and replication page [1], I think the current
> replication functionality supports replicating a healthy cluster to a
> backup cluster (irrespective of any cluster size). This is not a exact
> failover/failback process, but a backup and recovery tool.
> Please correct me if I am wrong.
>
>
> [1] http://hbase.apache.org/replication.html
>
> Thanks,
> Himanshu