zookeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From pubudu gunawardena <pubud...@gmail.com>
Subject Re: Failover on a two-node cluster
Date Thu, 21 Jul 2011 11:26:14 GMT
On Thu, Jul 21, 2011 at 2:05 PM, Daniel Yazbek
<daniel.yazbek@abuzz.com.au>wrote:

> Hi all,
>
> I am a newbie to zookeeper, so please be patient :)
>
> I have setup two zookeeper servers on two separate machines (A and B).
> I can start servers A and B ok.
>
> I have a client C, running on machine A. I can start the client, connect to
> the server, and create/change/remove znodes, and everything works nicely.
>
> Now, when I stop server B, the client C on machine A throws a
> ConnectionLossException when I try to get a znode.
>
> I initially thought that server A would be fine to run in standalone mode,
> but unfortunately this is not the case.
>
> Can someone please explain this behaviour to me? Is there some bit of
> configuration I am missing? Or, is this how it works by design?
>
According to the
documentation<http://zookeeper.apache.org/doc/r3.3.2/zookeeperAdmin.html#sc_zkMulitServerSetup>"
*As long as a majority of the ensemble are up, the service will be
available.*" The above behavior should be because 1 out of 2 is not a
majority.

> The same scenario with three servers (and bringing one down) works just
> fine.
>
> Further, is there a way to configure zookeeper server to dynamically
> "sniff" out its peers (via a UDP broadcast maybe) instead of hard-coding
> them in the zoo.cfg?
>
> Thanks
>
> -Daniel.

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message