[75345] in North American Network Operators' Group
Re: Probe dns service - anycast network
daemon@ATHENA.MIT.EDU (Joe Abley)
Thu Nov 11 18:14:56 2004
In-Reply-To: <3dbfa3d3041111150213fd08f2@mail.gmail.com>
Cc: nanog@merit.edu
From: Joe Abley <jabley@isc.org>
Date: Thu, 11 Nov 2004 18:13:52 -0500
To: Gere geomag <geomag@gmail.com>
Errors-To: owner-nanog-outgoing@merit.edu
On 11 Nov 2004, at 18:02, Gere geomag wrote:
> We are thinking of deploying anycast in our network for dns servers.
> I have the following scenario:
> - 10 server DNS (isc-bind) , linux and zebra for propagating ospf ip
> anycast.
>
> Are there someone who has developed a "solid scripts (perl/c/ecc..)"
> that is used to probe a dns service (udp/tcp port 53) and in case of
> failure (or max lookup query time problem) automatic shutdown ospf
> annunce or remove a server from a anycast network?
This doesn't quite answer your question, but there's a wrapper script
here:
http://www.isc.org/pubs/tn/isc-tn-2004-1.html#appx.WrapperScript
which kind of does what you're asking. It relies on BIND 9 dumping core
upon experiencing an assertion failure (or any kind of failure which is
likely to cause it to fail). BIND 9 is quite good at dumping core when
it gets confused (by design, in case anybody thinks I'm being
sarcastic).
You could always augment that with a crob job which does some digs, and
creates ${stop_file} and kills named in the event that the digs fail.
Testing remotely may also be a good idea. There are allusions made in:
http://www.ietf.org/internet-drafts/draft-kurtis-anycast-bcp-00.txt
to some of the monitoring and measurement issues surrounding anycast
service deployments. In any case, there are many ways to skin this cat.
Joe
