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

Posted to user@hbase.apache.org by Hanish Bansal <ha...@gmail.com> on 2013/11/13 13:10:26 UTC

Not able to get data if Master cluster goes down in case of data replication

Hi All,

My use case is to solution for HBase Disaster recovery. As Replication
is a HBase
Disaster recovery solution. I am using master-slave mode of replication.

I have two separate HBase clusters and separate zookeeper for each cluster:
Zookeeper for first hbase cluster is : zoo1:2181
Zookeeper for 2nd hbase cluster is : zoo2:2181

My *first hbase cluster is master cluster *and i have enabled replication
by referring the
http://blog.cloudera.com/blog/2012/08/hbase-replication-operational-overview/.
When i insert the data on master cluster by connecting to *zoo1:2181,* it
is replicated to slave cluster.
Till now there is no problem.

But as HBase Disaster recovery's main scenario is that
slave cluster will serve real time traffic in case the master site is down.

My application is querying HBase by connecting to *zoo1:2181.* Now if
master cluster goes down then queries should be served by slave cluster
because zoo1 holds replicated regionservers information. But i am not able
to get any data Instead i am getting MasterNotRunning exception.

Anything i am missing??

Any suggestion would really be helpful.

Thanks in advance !!

-- 
*Thanks & Regards*
*Hanish Bansal*

Re: Not able to get data if Master cluster goes down in case of data replication

Posted by Hanish Bansal <ha...@gmail.com>.

Thanks for Response Demai :)


On Wed, Nov 13, 2013 at 6:00 PM, Demai Ni <ni...@gmail.com> wrote:

> Hanish,
>
> I guess you are looking for hbase to automatically switch to the 2nd
> cluster. Unfortunately, that is not the architecture design of replication.
> And I think many of such design will rely on application layer for such
> 'smart' switch. for your application, once 'zoo1:2181' hit a
> MasterNotRunning exception, it can point its query to 'zoo2:2181' cluster.
> Some shops do that manually(that is by administrator), some implements in
> application logic. I am not aware of any hbase/server level switch.
>
> Demai
>
>
> On Wed, Nov 13, 2013 at 4:10 AM, Hanish Bansal <
> hanish.bansal.agarwal@gmail.com> wrote:
>
> > Hi All,
> >
> > My use case is to solution for HBase Disaster recovery. As Replication
> > is a HBase
> > Disaster recovery solution. I am using master-slave mode of replication.
> >
> > I have two separate HBase clusters and separate zookeeper for each
> cluster:
> > Zookeeper for first hbase cluster is : zoo1:2181
> > Zookeeper for 2nd hbase cluster is : zoo2:2181
> >
> > My *first hbase cluster is master cluster *and i have enabled replication
> > by referring the
> >
> >
> http://blog.cloudera.com/blog/2012/08/hbase-replication-operational-overview/
> > .
> > When i insert the data on master cluster by connecting to *zoo1:2181,* it
> > is replicated to slave cluster.
> > Till now there is no problem.
> >
> > But as HBase Disaster recovery's main scenario is that
> > slave cluster will serve real time traffic in case the master site is
> down.
> >
> > My application is querying HBase by connecting to *zoo1:2181.* Now if
> > master cluster goes down then queries should be served by slave cluster
> > because zoo1 holds replicated regionservers information. But i am not
> able
> > to get any data Instead i am getting MasterNotRunning exception.
> >
> > Anything i am missing??
> >
> > Any suggestion would really be helpful.
> >
> > Thanks in advance !!
> >
> > --
> > *Thanks & Regards*
> > *Hanish Bansal*
> >
>



-- 
*Thanks & Regards*
*Hanish Bansal*

Re: Not able to get data if Master cluster goes down in case of data replication

Posted by Demai Ni <ni...@gmail.com>.

Hanish,

I guess you are looking for hbase to automatically switch to the 2nd
cluster. Unfortunately, that is not the architecture design of replication.
And I think many of such design will rely on application layer for such
'smart' switch. for your application, once 'zoo1:2181' hit a
MasterNotRunning exception, it can point its query to 'zoo2:2181' cluster.
Some shops do that manually(that is by administrator), some implements in
application logic. I am not aware of any hbase/server level switch.

Demai


On Wed, Nov 13, 2013 at 4:10 AM, Hanish Bansal <
hanish.bansal.agarwal@gmail.com> wrote:

> Hi All,
>
> My use case is to solution for HBase Disaster recovery. As Replication
> is a HBase
> Disaster recovery solution. I am using master-slave mode of replication.
>
> I have two separate HBase clusters and separate zookeeper for each cluster:
> Zookeeper for first hbase cluster is : zoo1:2181
> Zookeeper for 2nd hbase cluster is : zoo2:2181
>
> My *first hbase cluster is master cluster *and i have enabled replication
> by referring the
>
> http://blog.cloudera.com/blog/2012/08/hbase-replication-operational-overview/
> .
> When i insert the data on master cluster by connecting to *zoo1:2181,* it
> is replicated to slave cluster.
> Till now there is no problem.
>
> But as HBase Disaster recovery's main scenario is that
> slave cluster will serve real time traffic in case the master site is down.
>
> My application is querying HBase by connecting to *zoo1:2181.* Now if
> master cluster goes down then queries should be served by slave cluster
> because zoo1 holds replicated regionservers information. But i am not able
> to get any data Instead i am getting MasterNotRunning exception.
>
> Anything i am missing??
>
> Any suggestion would really be helpful.
>
> Thanks in advance !!
>
> --
> *Thanks & Regards*
> *Hanish Bansal*
>