load balance of DNS

[Mark Andrews]

In message <EB2AF3E3-5C8F-47FB-9D0A-FB82B8A42EA2 at kumari.net>, Warren Kumari wri
tes:
> 
> On Jan 16, 2012, at 2:58 PM, Todd Snyder wrote:
> 
> >>> do you propose he specify the ratios with BIND?
> >>> =
> 
> >>> One (icky) solution is to hand out more addresses for one server than =
> 
> >>> the other=8A
> >>> =
> 
> >>> www.example.com  IN  A  192.168.1.1
> >>> www.example.com  IN  A  192.168.1.2
> >>> www.example.com  IN  A  192.168.1.3
> >>> www.example.com  IN  A  192.168.2.1
> >>> =
> 
> >>> Bind 192.168.1.[1-3] to server1 and 192.168.2.1 to server2.
> > =
> 
> >> Unless things have changed since I last checked this (many years ago), B=
> IND ignores the duplicates.
> > =
> 
> > In this case, there are no dupes - I thought this at first, but noticed i=
> t's 1.2 and 2.1 after my dyslexia turned off for a moment.  =
> 
> > =
> 
> > The idea is novel, if ugly - bind multiple unique addresses and the BIND =
> daemon won't know they're all for the same physical box.  It'd work, but ug.
> 
> Oh, yeah -- I'm the first to admit that it is ugly, but it *does* work -- I=
>  used this for a while at a previous company (before GSLB solutions existed=
> ) and while it made me want to throw up every now and then, it did allow me=
>  to perform unequal load-sharing=85

If you want unequal load sharing design your protocol to use SRV records.

If you are runnning your protocol on top of http/https complain to WC3 to
get SRV support added.

Mark
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org