Mailing-List: contact user-help@zookeeper.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@zookeeper.apache.org
Received-SPF: pass (nike.apache.org: domain of raj.cassandra@gmail.com
 designates 209.85.213.170 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAKF1A=vXVtWR1HosFidnbh=NZTr0_Bnotkaxc1EQkM-M4pOGRg@mail.gmail.com>
References: 
 <CAF+j8akY3xDjMdNYqUxcF+-GVQkJVw3d95b1jE6ktqS3qoh0-Q@mail.gmail.com>
	<CAKF1A=vXVtWR1HosFidnbh=NZTr0_Bnotkaxc1EQkM-M4pOGRg@mail.gmail.com>
Date: Tue, 3 Jul 2012 15:48:10 -0400
Message-ID: 
 <CAF+j8a=99Aeph5UCJLE4kB2GedCDPk6UHuzpHBbuzJa1nFHqEA@mail.gmail.com>
Subject: Re: Zookeeper Leader Election Preference
From: Raj N <raj.cassandra@gmail.com>
To: user@zookeeper.apache.org
Content-Type: multipart/alternative; boundary=e89a8ff1c6d4ba734f04c3f2315f

--e89a8ff1c6d4ba734f04c3f2315f
Content-Type: text/plain; charset=ISO-8859-1

Thanks Camille. I was thinking about the same approach. But I was hoping
something could be built into Zookeeper to give a hint.

Thanks
-Raj

On Tue, Jul 3, 2012 at 3:28 PM, Camille Fournier <camille@apache.org> wrote:

> When I was doing a similar thing, we would start the 2 nodes in the primary
> first (which will get a leader elected), then start the 3rd in the backup.
> We also had a script set up to monitor the 3rd ZK and restart it if it
> became leader. Hacky but it works.
>
> The challenge is that some of the monitoring commands can timeout across
> data centers because our monitoring network connection logic is broken, so
> use the most lightweight monitor possible for this. stat is probably it.
>
> C
>
> On Tue, Jul 3, 2012 at 3:16 PM, Raj N <raj.cassandra@gmail.com> wrote:
>
> > Hi experts,
> >      I am trying to setup a Zookeeper ensemble of 3 nodes across 2 DCs. 2
> > of the 3 will run in the primary. And 1 will run in the secondary DC. Is
> > there a way to hint that the leader should be selected from the 2 nodes
> in
> > the primary DC. I am trying to avoid the case where the node in the
> > secondary DC is the leader since that means I will have to pay for the
> WAN
> > latency for every write request.
> >
> > Thanks
> > -Raj
> >
>

--e89a8ff1c6d4ba734f04c3f2315f--