Dear Oliver, Thank you so much for help. I should've guessed that trick myself)
Thanks once again and best wishes, Tigran Zakoyan. 03 нояб. 2014 г. 19:42 пользователь "Olivier Benghozi" < olivier.bengh...@wifirst.fr> написал: > Firstly, in classical BGP implementations (but not in BIRD), for eBGP > connections, router-id is never used as tie-break. Instead, route age is > used (the oldest one wins) as tie-break. It avoids frequent route changes > and probably improves a little routes distribution. > You'd better use "prefer older on". > > Secondly, a radical solution is (for incoming prefixes): > - upstream 1: match net ~ 0.0.0.0/1 and set MED at 1, set MED at 0 for > everything else; > - upstream 2: match net ~ 128.0.0.0/1 and set MED at 1, set MED at 0 for > everything else. > You can leave that config once for all, or remove it a few minutes later > if "prefer older on" is configured. > > > regards, > Olivier > > > Le 3 nov. 2014 à 14:44, Tigran Zakoyan <zako...@gmail.com> a écrit : > > > > Dear colleagues, > > > > My issue is best route selection in BGP protocol. > > > > I have two uplinks providing full view to me. The world connectivity > > of the First one is slightly better, than the Second's. But I get in > > result 90% of routes selected through the First. > > > > There are no LP or MED differences. And most, if not all, of the > > routes with the same AS path length get chosen from the First's table. > > > > I presume it to be the influence of lower router ID (77.x.x.x the > > First and 210.x.x.x the Second). > > If it's the case, is there any way to deviate the BIRD's decision? To > > make it choose more routes from the Second's? > > > > Thanks and cheers, > > Tigran Zakoyan. > >
