Multiple default records

[Kevin Darcy]

MX records were designed with failover and load-balancing in mind, which 
is why they have a "preference" field. But MX records are only for mail. 
SRV records are a more generalized form of service-location record, with 
not only a "preference" field, but also a "weight" field (in case you 
want some hosts to get more connections than others), but unfortunately 
the browser developers haven't picked up SRV-record support yet, despite 
significant prodding to do so.

                                                                         
                                                               - Kevin

Kvetch wrote:

>Thanks, I was looking into this because we use hardware load balancers for 
>some of our sites but for the default records I wanted to see if there was a 
>simpler way of achieving this type of failover.
>That makes sense that it is up to the user's browser but I wasn't sure if 
>there was anything in the round robin distribution method that accounted for 
>this. I thought MX records were handled this way. If the preferred MX record 
>fails it attempts to delivery to the next MX record.
>Thanks,
>Nick
>
>On 8/31/05, John Wobus <jw354 at cornell.edu> wrote:
>  
>
>>The DNS has nothing to do with whether there is
>>actually anything at the address. It's just a database
>>that (in this case) looks up names and gives
>>out corresponding numbers. However, there
>>are products that do what you suggest, checking
>>webservers, and reconfiguring DNS servers
>>(such as bind) accordingly.
>>
>>The actual user experience is dependent upon
>>what clients (IP stacks and browsers) do. Client
>>software does whatever the coder thought
>>worked best (as opposed to following any
>>agreed-upon protocol), and the coders are
>>often concerned with the avoiding huge numbers
>>of lookups (e.g. gif images), thus they cache.
>>Typical browser cache-time is a half hour.
>>
>>Balancing load and handling fallback is a
>>whole field, with lots of lively
>>discussion, and products that cost
>>substantial amounts of money. Probably
>>more than you can imagine.
>>
>>John
>>
>>On Aug 31, 2005, at 2:42 PM, Kvetch wrote:
>>
>>    
>>
>>>Hello, I was wondering exactly what happens with a zone that has
>>>multiple
>>>default records and one of those records are unavailable.
>>>For example if I have
>>>foo.net <http://foo.net>. 60 IN A 192.168.1.1 <http://192.168.1.1>
>>>foo.net <http://foo.net>. 60 IN A 192.168.1.2 <http://192.168.1.2>
>>>foo.net <http://foo.net>. 60 IN A 192.168.1.3 <http://192.168.1.3>
>>>
>>>If a user visits foo.net <http://foo.net> <http://foo.net>'s website and 
>>>      
>>>
>>gets the
>>    
>>
>>>192.168.1.1 <http://192.168.1.1><http://192.168.1.1>server, then 15 
>>>      
>>>
>>minutes they return to
>>    
>>
>>>foo.net <http://foo.net> <http://foo.net> but 192.168.1.1<http://192.168.1.1><
>>>      
>>>
>>http://192.168.1.1> had
>>    
>>
>>>crashed
>>>during this 15 minute interval would they automatically be sent over to
>>>192.168.1.2 <http://192.168.1.2> <http://192.168.1.2> or 192.168.1.3<http://192.168.1.3><
>>>      
>>>
>>http://192.168.1.3> or
>>    
>>
>>>would they continue to round robin between the 3 IP's and might get a
>>>404
>>>error if they hit 192.168.1.1 <http://192.168.1.1> <http://192.168.1.1>? 
>>>      
>>>
>>So basically does
>>    
>>
>>>DNS
>>>recognize that one of the IP's is unavailable and send the user to
>>>another
>>>IP. Kind of like MX record's preference values.
>>>
>>>Since the TTL is 1 hour they would not have to return to our DNS server
>>>since they received all 3 IP's, correct?
>>>
>>>Thanks,
>>>Nick
>>>
>>>
>>>      
>>>
>>
>>    
>>
>
>
>
>
>
>  
>