I suspect that the generousity needs to be over time - i.e. route flap during convergence is normal and you may see a burst over some minutes (details can be dredged out of update logs). I suspect that we need more think time here, and perhaps some more investigation - what we don't want is to simply generate more NOC calls to manually undamp routes, so where and how route propagation supression should be applied deserves some considerable thought. Geoff At 03:58 AM 29/04/2006, Rob Evans wrote:
i suggested that the rfd spec be modified, at least to be per-prefix as opposed to prefix+path, to allow the disaster prefixes to be easily detected and damped.
I get the impression from RFC2439 (sec. 4.4.3) that the authors intended it would be optional for the path to be included in the per-route state, so perhaps some vendor pressure to implement the relevant knob? Or am I missing some other aspect of the document?
Then the pathological prefixes could be damped by using some "generous" parameters that most prefixes would never reach in even the most severe circumstances.
Is that the idea?
Rob