On Fri, 8 Oct 2004 12:05:56 -0700 (PDT), Mike Mestnik <[EMAIL PROTECTED]> wrote: > > --- Albert Ulmer <[EMAIL PROTECTED]> wrote: > > OK, sounds reasonable. Can this scheme be adapted to a scenario where > > 1 NIC (no pppd) and 1 ISDN-card are involved? Putting > > up-/down-statements into the NIC-setup won't do much good, as the > > NIC-link doesn't go down, even when the connection on the WAN-side is > > lost. > > > What happens is the routing code detects that the metric 0 path isn't > working and it starts using the metric 1 part. The thing that may be > tricky for you, I don't know ISDN only NICs, is that you must keep the > interface bound to an IP and have routes(that have a higher metric) > attched to it. This means that the interface has to be "ifconfig xxxy > up", but maby not dialed and working.
The ippp0 interfaces pretty much behaves like a normal eth-interface. I should be able to assign a static IP-address for routing purposes, albeit using a virtual ippp0.0 interfaces if necessary. > > Is there a way to alter the metric of an interface based on RTT of the > > ISP-router on the other side of the link? I guess this would solve > > part of the problem.... > > > That's what metric is, it's the admins perseption of what routes are > better then others. You could base this only on RTT, but that may be > narrow minded. That would basically be what I want: If the RTT goes up, try the alternative uplink. It may be narrow minded, but it would certainly get the job done. ;) Best regards, Albert.
