David Edmondson writes: > Solving this 'properly' requires teaching in.mpathd about layer 2 > reachability for arbitrary peers and then plugging in the various > underlying mechanisms (VNICs, LDoms direct interdomain > connections, ...). After that we'd have to somehow transform the layer > 2 reachability into layer 3 reachability - maybe by fiddling in the > forwarding table?
I don't think it really requires that. IPMP's response to failure is not to route around it; it's to shut down the broken links. It uses a pretty big hammer, so there's just no such thing as "partial failure" or "partial success" here -- local reachability like this should be considered to be useless (for IPMP's purposes) if the probe targets are dead. I get what you're saying, and there are more general solutions than IPMP, but since IPMP is a supported feature, it'd be nice to be compatible. I think it'd be more than sufficient to provide some means for the lower layer to say "I'm sick" to interested clients. Today, we do that by flicking off the IFF_RUNNING bit, and that has the obvious side-effect (in this case) of damaging local connectivity when the user doesn't care about IPMP. Perhaps there's some sort of new notification we could invent for this case. -- James Carlson, Solaris Networking <[EMAIL PROTECTED]> Sun Microsystems / 35 Network Drive 71.232W Vox +1 781 442 2084 MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677 _______________________________________________ networking-discuss mailing list [email protected]
