Mailing-List: contact user-help@zookeeper.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@zookeeper.apache.org
Received-SPF: softfail (nike.apache.org: transitioning domain of
 rgs@itevenworks.net does not designate 209.85.218.49 as permitted sender)
MIME-Version: 1.0
In-Reply-To: <EAC896C7-C6C8-4644-82DE-B1DDFE1D0A79@motum.be>
References: <0ABDD99C-B3C1-4278-B6E8-6A997658B988@motum.be>
	<CAMUR=WjxUaXbMD8tz07On_tGOdzYyK+0D4Dj93bd6+rf66wvEA@mail.gmail.com>
	<24927B98-DABB-4D6D-8BCB-9F8D60896EAC@motum.be>
	<CAMxCP3hQTymrqisR+5XCAAWxFdK+FghMBv9zFxt6RdHJRRZaXg@mail.gmail.com>
	<26F7CD6B-9381-4DD7-B612-6E506736A044@motum.be>
	<EAC896C7-C6C8-4644-82DE-B1DDFE1D0A79@motum.be>
Date: Thu, 25 Jun 2015 08:55:33 -0700
Message-ID: 
 <CAMUR=Wg-b6cZHFtiHyfXEUKGvNrjydRHyyy4CEUDKORg9QR1Mg@mail.gmail.com>
Subject: Re: Leader election problems
From: =?UTF-8?B?UmHDumwgR3V0acOpcnJleiBTZWdhbMOpcw==?= <rgs@itevenworks.net>
To: "user@zookeeper.apache.org" <user@zookeeper.apache.org>
Content-Type: multipart/alternative; boundary=001a113e12b8558e96051959a734

--001a113e12b8558e96051959a734
Content-Type: text/plain; charset=UTF-8

On 25 June 2015 at 08:12, Filip Deleersnijder <filip@motum.be> wrote:

>
> Some more information that we just got from our installation.
>
> - To check the EOF theory, we cleared all the version-2 data directories
> and restarted all vehicles
> - 2 of the 8 vehicles were unavailable at that time
> - The remaining 6 vehicles did not succeed in finishing leader election
> phase !
>

After how long?


>
> A log extract can be found at the end of this e-mail.
>
> - We then removed the 2 vehicles from the configuration just for testing
> purposes
> - Now leader election succeed without any problems !
>

How long did it take?

Having the logs for all servers (for both cases: for when the elections
fails and for when it succeeds) would be helpful. Please paste the logs on
gist.github.com or something similar since reading them in-line is a bit
hard.

I don't think the unclean shutdowns should be a problem, as long as quorum
can be formed (i.e.: they'll just be ignored, since ZooKeeper is meant to
be crash safe) (also, a Windows shutdown is probably not very different
then a shutdown through the init scripts).


-rgs

--001a113e12b8558e96051959a734--