I have a bigcouch cluster like this with haproxy front end that periodically checks health of the nodes. If one is down it stops routing there. Wasn't too difficult to setup. On Nov 6, 2013, at 1:45 PM, Robert Newson <[email protected]> wrote:
> Then have several haproxy servers and have DNS level failover (like > Cloudant does). > > B > > On 6 November 2013 19:05, Hank Knight <[email protected]> wrote: >> At the bottom of the page for Installing & Using BigCouch it says: >> "Cloudant recommends HAProxy." >> http://bigcouch.cloudant.com/use >> >> Sounds like a reverse proxy is the only way to accomplish this. >> Unfortunately then becomes the vulnerable link. >> >> On Wed, Nov 6, 2013 at 2:41 PM, Matthieu Rakotojaona >> <[email protected]> wrote: >>> Excerpts from Hank Knight's message of 2013-11-06 19:27:56 +0100: >>>> I would like to distribute a CouchDB database on a minimum of 3 >>>> servers for the purpose of redundancy. >>>> >>>> An attachment could be accessed like this: >>>> >>>> http://cdb1.example.com/abc/xyz/image.jpg >>>> http://cdb2.example.com/abc/xyz/image.jpg >>>> http://cdb3.example.com/abc/xyz/image.jpg >>>> >>>> Is it possible to access the database from a single domain name that >>>> would automatically resolve to an instance that is running? >>>> >>>> For example if the server running on of the instances was experiencing >>>> a power outage it should automatically resolve to an instance that >>>> works. >>>> >>>> If so, where can I find documentation about how to do this? >>> >>> This really should be the role of SRV records [0], but you'd need to ask >>> http clients to look out for a _http._tcp record. In the meanwhile, the >>> standard solution is a frontend proxy like HAProxy. >>> >>> [0] https://en.wikipedia.org/wiki/SRV_record >>> >>> -- >>> Matthieu Rakotojaona
