On Wed, 12 Apr 2000, you wrote:
> No, this is not what I am trying to do. I am not trying to use round robin
> dns, and I am not interested in adjusting the TTL's, or worry about what the
> current good address is.
Dropping the TTL to just a few minutes would give finer granularity to the dns
failover plan, so it's a good idea. It just doesn't help you ATM.
I suspect the problem is (and if you've solved it, I could use your help) Linux
responds on the first route that qualifies, regardless of how the request
packet arrived. So, the server takes a request from either line, but always
sends the response out the 'default' line. I discovered this today, much to my
dismay, when I put three NICs in one network and everything went out eth0.
Apparently you can set two default routes on a box, though I haven't tested
that. It still sends everything out one route, but shifts over if that route
becomes unreachable.
Failing that, you would need a box on each connection pulling from a back-end
box. For NFS, you need decently sized disk caches on the two front line boxes.
If NFS isn't fast enough (it should be -- it only needs to supply DSL
capacity), you could try two squid proxy servers hitting a back end web server.
The proxy setup is a little twisted, since you need to supply the outside
world with one IP address for a given name, but give squid a different address.
-- Brian
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
