|
Joseph Mack NA3T wrote:
> On Wed, 28 Sep 2005, Dan Trainor wrote:
>
>> Hello, all -
>>
>> Sorry to not reply with the correct headers here; I just signed up for
>> this list, and it just so happened that Jacob Coby asked a question that
>> was similar to the one which I wanted to ask, so here goes.
>
>
> and what is your question?
>
> The problem with DNS is not the servers but the resolvers AFAIK. The
> nameserver can be hung in a way that the client resolver interprets as
> there being no IP for the name, and it doesn't go onto the next
> nameserver in the list. A client should be able to differentiate no
> answer from one saying that there is no resolution for that name. I
> can't imagine how the broken clients (in all OSs) are going to be fixed
> and we aren't the people to fix it.
>
> The nameresolving process is totally broken
>
> http://www.austintek.com/LVS/LVS-HOWTO/HOWTO/LVS-HOWTO.services.general.html#name_resolution
>
>
> and no-one seems to have noticed. The only solution I can
> see is to present a perfect nameserver via something like
> LVS, where the client never sees a bad nameserver.
>
> Joe
>
Joe -
Yes, and I completely agree, which is again why I am hesitant to even
consider using DNS for any sort of load balancing.
My question was, how do people provide these high availability systems
that I speak of? Once a client is connected to one of my "pools" of
servers, which do the actual processing, that's cake - the problem I see
is a reliable connection to one of these "pools", and that's where I was
curious.
Thanks
-dant
|
