[geeks] Remote Failover

[Sridhar Ayengar]

Patrick Giagnocavo wrote:
>>> DNS round-robin is a really poor way to do failover/balancing,
>>> due to TTL and caching.  Why not use Ethernet load balancing, etc
>>> (LocalDirector and so forth)?
>> 
>> Because the machines are a thousand miles apart.  I'm not
>> particularly concerned with balancing, really.  And I can always
>> set the TTL really low for that one record, no?
> 
> Yes, but many larger ISPs will cache the DNS record anyways even if
> they are not supposed to.

Bleh!

> You might want to look into some other way to get what you want.  Do
> you have enough bandwidth to set up a box using NGINX or Pound (load
>  balancing web proxies that can detect when a machine is not
> responding) somewhere, then have it transparently switch between the
> two?

But then, aren't I still susceptible to going down if the proxy machine 
goes down?  The DNS infrastructure in question is already quite 
fault-tolerant.  Building a fault-tolerant web proxy infrastructure 
might be biting more off than I can chew, I think.

Peace...  Sridhar