[135215] in North American Network Operators' Group
Re: Software DNS hghi availability and load balancer solution
daemon@ATHENA.MIT.EDU (david raistrick)
Tue Jan 18 14:44:22 2011
Date: Tue, 18 Jan 2011 14:42:57 -0500 (EST)
From: david raistrick <drais@icantclick.org>
To: Rhys Rhaven <rhys@rhavenindustrys.com>
In-Reply-To: <4D35E50D.5060905@rhavenindustrys.com>
Cc: nanog@nanog.org
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
On Tue, 18 Jan 2011, Rhys Rhaven wrote:
> Having hit these issues myself, I heavily recommend a real frontend
> proxy like nginx or varnish.
A frontend proxy (nginx, varnish, haproxy, or anything else) doesnt give
you HA any more than any other loadbalancer solution does. You need a way
to send traffic to another frontend server when the primary frontend
server fails, or is overloaded, transparently.
The tools we have available these days to do this are VRRP-like solutions
(which all of the appliances use) that use multicast, some amount of
NAT and routing magic (which I've often not seen done sanely), or DNS
solutions (better known as GSLB) that dynamicly change the DNS responses
depending on conditions (which could be source location, or could be
server availability, or whatever).
Normally, VRRP would be the way to go. But these days multicast isn't
supported everywhere (major example - Amazon EC2), leaving DNS...
--
david raistrick http://www.netmeister.org/news/learn2quote.html
drais@icantclick.org http://www.expita.com/nomime.html
