You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@zookeeper.apache.org by fededonna <fd...@gmail.com> on 2014/08/22 13:45:29 UTC

Cluster migration

Hi guys, we have the following scenario in production:

5 servers (1 leader, 4 followers), we want to migrate this cluster without
downtime. Can this be achieved adding one server and decommissioning other
one at the time? wouldn't this cause a "no leader scenario" when we have 4
of the new servers and 1 of the old? 

Do you have any advice for this scenario? thanks in advance.



--
View this message in context: http://zookeeper-user.578899.n2.nabble.com/Cluster-migration-tp7580215.html
Sent from the zookeeper-user mailing list archive at Nabble.com.

Re: Cluster migration

Posted by kishore g <g....@gmail.com>.
That should work, but keep in mind that when ever you add a server you will
have to restart every server with the new zoo.cfg that contains new quorum.

Do some validation every time a server is added and another is
decomissioned. Some of the things I remember
-- Ensure that all zookeepers have the same number of znodes
-- Count the number of connections on each server before and after. It
should be the same if no new clients have connected.

I suggest not to do this manually unless you are really confident. Its
better to write a script and test it on dummy servers before.

Other ZK experts please validate.

thanks,
Kishore G



On Fri, Aug 22, 2014 at 4:45 AM, fededonna <fd...@gmail.com> wrote:

> Hi guys, we have the following scenario in production:
>
> 5 servers (1 leader, 4 followers), we want to migrate this cluster without
> downtime. Can this be achieved adding one server and decommissioning other
> one at the time? wouldn't this cause a "no leader scenario" when we have 4
> of the new servers and 1 of the old?
>
> Do you have any advice for this scenario? thanks in advance.
>
>
>
> --
> View this message in context:
> http://zookeeper-user.578899.n2.nabble.com/Cluster-migration-tp7580215.html
> Sent from the zookeeper-user mailing list archive at Nabble.com.
>