couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Adam Kocoloski <kocol...@apache.org>
Subject Re: replication over internet
Date Fri, 25 Sep 2009 13:15:03 GMT
On Sep 25, 2009, at 6:42 AM, go canal wrote:

> hello,
> Looking at the replication wiki, I assume that the current  
> implementation is for supporting high availability - so machines are  
> likely in the same local network.
>
> What if I want to replicate to the remote sites ?

CouchDB's replication model is certainly geared for that as well.  In  
fact, it's designed to work between servers that are only occasionally  
connected, too.

> At least two requirements I can think of:
> * (must) first, we need to support HTTPS not only HTTP, unless the  
> data is already encrypted.

This is supported in the upcoming 0.10 release.

> * (want) secondly, schedule the replication - I do not know how  
> internally replication is done, with every change, or already use  
> some buffer for batch process. but better if we can configure when  
> to replicate, how often, etc. for example, hourly, etc. My customer  
> asked me to avoid as much network activities as possible during  
> certain period of time during the day. of cause, they maybe able to  
> look into their infrastructure and applications .... but nice if we  
> can configure something.

By default, CouchDB replication is "one-and-done" -- its replicates  
all the updates that are missing on the target at the moment you start  
the replication, then stops.  If you pass the "continuous":true option  
in the POST to _replicate, Couch will continue to replicate all  
updates immediately.

For the scheduled replication service you describe above, we recommend  
triggering normal replications with cron.  Down the line CouchDB might  
grow its own internal cron, but we didn't want to reinvent the wheel.   
Best,

Adam


Mime
View raw message