Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Zookeeper >> mail # user >> Zookeeper Coodination Service

Copy link to this message
RE: Zookeeper Coodination Service
Yeah, we achieved something similar using a simple "VIP".  Once the leader election switches, the new leader can activate the VIP, while the old one deactivates it.  DNS with a very short TTL is another way to achieve this.


-----Original Message-----
From: Edward Ribeiro [mailto:[EMAIL PROTECTED]]
Sent: Thursday, January 17, 2013 12:06 PM
Subject: Re: Zookeeper Coodination Service

Yes, what you have just described is a sweet spot for Zookeeper.

Based on your description, you want one of the servers as the leader and while having  the others waiting. Take a look at "Leader Election Recipe"
in ZK Tutorials for how to implement this. If the leader goes offline then the other machines use ZK to run the election procedure and elect a new leader to serve requests. Of course, you will also need a "router" service that also listens to ZK so that it can redirect the requests to the right machine.


On Thu, Jan 17, 2013 at 2:02 PM, Saurabh Dutta <[EMAIL PROTECTED]> wrote:

> Hi,
> I've a a simple tcp service which will listen on a port for incoming
> requests from client. I'd like to have this same service run on
> multiple servers in a failsafe manner.
> There will be a client which will initially write to server1 and if it
> goes down i want the requests to be directed to server2.
> I'm tyring to write a a zookeeper coordinator which will keep my
> services up and if one service goes down i want the other service to
> take over and replace it.
> Please help me to explore if this can be done through zookeeper and if
> yes how this can be achieved.
> Thanks,
> Saurabh

No virus found in this message.
Checked by AVG - www.avg.com
Version: 2012.0.2221 / Virus Database: 2638/5539 - Release Date: 01/17/13