Remove/add [A] records based upon server availability

[Danny Mayer]

Mark Andrews wrote:
> In message <12831C89-2438-4A84-B81F-14A2ED000DDF at menandmice.com>, Chris Buxton 
> writes:
>> On Dec 27, 2009, at 7:16 AM, Rick Dicaire wrote:
>>> On Sun, Dec 27, 2009 at 3:16 AM, Ryan S <ryan332211 at hotmail.com> wrote:
>>>> Some web browers and applications will fail in a round-robin A record
>>>> configuration such that if the first A record returned is unavailable, the
>> n
>>>> the browser will not bring up the page.
>>> So fix the application instead of bending the protocol to suit a
>>> broken applications need?
>>> Specifically, what web browsers and applications are you referring to?
>>> On what OS's?
>> All web browsers, pretty much. Round robin does not provide failover except f
>> or protocols and applications that specifically make it work, such as the DNS
>>  and SMTP protocols (only between servers, in each case).
>>
>> Using DDNS to remove unresponsive or overloaded web servers from the rrset wo
>> rks OK in situations where solutions at the HTTP and routing layers are not a
>> ppropriate, such as web servers in different physical locations; there are ap
>> pliance vendors out there offering such solutions. This is also similar to on
>> e part of the Akamai solution for global traffic management.
>>
>> However, this strategy should be avoided when possible, and buffered with hig
>> hly available solutions at each point to minimize the use.
>>
>> Chris Buxton
>> Professional Services
>> Men & Mice
> 
> Applications that fail to try multiple address are broken.  RFC 1123
> said as much back in October 1989 (over twenty years ago now).  With
> IPv6 coming along almost every host will be multihomed and if a
> application doesn't cope then you should report it to the vendor now.

While I agree with you in principal, this is almost universally ignored
by just about every internet application that I've ever used including
every one of the web browsers I use.

Danny