Mailing-List: contact user-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@couchdb.apache.org
Received-SPF: pass (nike.apache.org: domain of dave@muse.net.nz designates
 209.85.160.52 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAJyA54_18aUDRn9=E1cDZBSuD2FZ34mnxJb0+mjAYsuvd90Rqg@mail.gmail.com>
References: 
 <CAJyA54_18aUDRn9=E1cDZBSuD2FZ34mnxJb0+mjAYsuvd90Rqg@mail.gmail.com>
Date: Wed, 22 Aug 2012 20:28:21 +0200
Message-ID: 
 <CAHZBNKZr6a4vMqJBwEaMc6C8ZqV7XeQ7NHUjR6CS9Rq1ZUmWHA@mail.gmail.com>
Subject: Re: Changing the system clock underneath a multi-master setup
From: Dave Cottlehuber <dave@muse.net.nz>
To: user@couchdb.apache.org
Content-Type: text/plain; charset=ISO-8859-1

On 22 August 2012 20:13, Michael Parker <michael.g.parker@gmail.com> wrote:
> I have two servers running CouchDB in a multi-master setup, where each
> is performing a continuous pull replication on the other. I stupidly
> assumed that both servers had their system time set appropriately with
> ntpd, but I was wrong, and they're both off by a minute or two. Will
> running ntpd on each server affect CouchDB or the replication at all?
> Should I play it safe and stop replication, down CouchDB, change the
> clock, and then up CouchDB and start replication again?

I think you can change the clock quite safely. The erlang runtime will
likely emit
some warnings, but with a delta < 5 minutes this should not cause it
to restart[1].

With this short a time difference, I'd not get ntpd to forcefully
update, but let it
drift onto the corrected time. This will no doubt be platform dependent.

A+
Dave

[1]:http://www.erlang.org/doc/man/heart.html