using observers for failover

[Brian Tarbox <br...@gmail.com>]

We have a 3 node cluster which as I understand it means we can survive the
failure of a single node.  We have a 4th machine available and I'm
wondering if I can run zk as an observer on that node and have it take over
in the case of the failure of one of the regular nodes?

If that doesn't make sense can anyone answer the more general question of
how to deal with multiple node failure?

Thanks,

Brian Tarbox

-- 
http://about.me/BrianTarbox

Re: using observers for failover

[Michi Mutsuzaki <mi...@cs.stanford.edu>]

Hi Brian,

Currently you cannot convert an observer to a participant without
modifying zoo.cfg and restarting all the servers. With dynamic
reconfiguration (new in 3.5), you can convert  an observer to a
participant without needing to restart the entire cluster.

I think by far the most common way to tolerate 2 node failure is to
deploy a cluster with 5 or more nodes.


On Thu, Sep 12, 2013 at 5:16 AM, Brian Tarbox <br...@gmail.com> wrote:
> We have a 3 node cluster which as I understand it means we can survive the
> failure of a single node.  We have a 4th machine available and I'm
> wondering if I can run zk as an observer on that node and have it take over
> in the case of the failure of one of the regular nodes?
>
> If that doesn't make sense can anyone answer the more general question of
> how to deal with multiple node failure?
>
> Thanks,
>
> Brian Tarbox
>
> --
> http://about.me/BrianTarbox