Hi, On Mon, May 18, 2015 at 08:21:10AM +0200, Marian ??urkovi?? wrote: > the fact that both entries are present in RIB and CEF tables is expected > behaviour. It's part of the new design, where both active and backup paths are > preinstalled into the CEF table, which allows much faster convergence when the > active path disappears. Note that the backup path has different CEF flags, > which > should instruct the forwarding engine *not* to use it when another regular > path > is present in CEF table. > > It should be investigated why your router uses the backup path for traffic - > this is clearly wrong. One possible reason might be that the affected > linecards > are running different firmware version than required by the IOS.
The answer I received in the TAC case (SR 635067397) is that "this is totally expected behaviour, because the CEF team thinks it needs to work that way". If I do not want load-sharing here, I am supposed to use labelled next-hops, so router sending prefixes *to* me can steer whether they want to use the "best-external" path or not. To me this sounds like total bullshit... (Yes, during reconvergence, microloops will happen if ebgp best-external is used. Otherwise, traffic blackholing happens if a prefix goes away and the peer router does not have an alternative path. So, right. Leave that decision to *me*, not force my network design because someone doesn't understand trade-offs) gert -- USENET is *not* the non-clickable part of WWW! //www.muc.de/~gert/ Gert Doering - Munich, Germany g...@greenie.muc.de fax: +49-89-35655025 g...@net.informatik.tu-muenchen.de
pgpXGEvJs0uDY.pgp
Description: PGP signature
_______________________________________________ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/