Mailing-List: contact user-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@couchdb.apache.org
Received-SPF: pass (athena.apache.org: local policy)
Date: Mon, 29 Jul 2013 12:14:46 -0700
From: James Marca <jmarca@translab.its.uci.edu>
To: "user@couchdb.apache.org" <user@couchdb.apache.org>
Subject: Re: Two way replication question
Message-ID: <20130729191446.GJ31164@translab.its.uci.edu>
Mail-Followup-To: "user@couchdb.apache.org" <user@couchdb.apache.org>
References: <20130729183736.GF31164@translab.its.uci.edu>
 <CABvT1DFJ1P_-qndRR4ewWnVYYSqutzeya42wJBE5fQ45X7Cf=Q@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <CABvT1DFJ1P_-qndRR4ewWnVYYSqutzeya42wJBE5fQ45X7Cf=Q@mail.gmail.com>
User-Agent: Mutt/1.4.1i

On Mon, Jul 29, 2013 at 07:47:10PM +0100, Robert Newson wrote:
> CouchDB will check that the current revision of every document is
> present on the target and will skip copying it if it is. In your case,
> the replication will not transfer any documents at all but will check
> every id/rev pair which will take some time.

Okay thanks.

Is this also true if I pull and push from the same machine.  That is,
if I set up pull replication on A to pull from B, in parallel with the
push from B to A?  

My understaing is that the two replication jobs (push A to B, pull B
to A, both running on A) would be totally independent and would not
share information about B's db state.  Correct?

Thanks,
James

> 
> B.
> 
> 
> On 29 July 2013 19:37, James Marca <jmarca@translab.its.uci.edu> wrote:
> > Hi,
> >
> > I have two machines, A and B.  I am running an analysis on A, saving
> > the output to couchdb, and then replicating by push from A to B.
> >
> > B's database started life from this push replication, and has never
> > been otherwise modified.  So the two are more or less identical.
> >
> > What I want to do now is parallelize my work flow and run more
> > analyses on B, save locally (to B's couchdb), and then replicate those
> > changes to A.  (The database is used both to save output, and also to
> > keep track of what has been done already, so it is useful to keep the
> > output db syncronized between all machines.)
> >
> > My question is whether replicating from B to A will require pushing
> > all of the docs to A.  This is an issue because my database is 21 GB
> > and growing, and I'd rather not push all that data from B to A when I
> > *know* the two are identical right now.
> >
> > Is there a way to set up replication to skip everything already there?
> > Or to copy the replication state from A to B so that B knows that
> > replication with A can start with new data only?
> >
> > If not, I can of course just save work done on B to the CouchDB on A
> > directly, but I'd rather set it up so that the computation process
> > always just hits couchdb on localhost, and let couchdb do the machine
> > to machine copying.
> >
> > Thanks for any insights or pointers to the correct docs page,
> >
> > James Marca