Hello I am experimenting with BGP PIC before deploying it to production and I have found an oddity: With "set routing-options protect core" local-preference stops working the way I would expect it to work.
If I configure "local-preference 120" on import from my IBGP neighbor the router will send traffic for some prefixes to my IBGP neighbor (where they will egress through EBGP) and for other prefixes directly to my EBGP neighbor. It seems like they're getting load-balanced per destination subnet. This is how it looks in the routing table: root@router1# run show route 1.0.143.0/24 table inet.0 inet.0: 705394 destinations, 2116034 routes (705393 active, 0 holddown, 2 hidden) @ = Routing Use Only, # = Forwarding Use Only + = Active Route, - = Last Active, * = Both 1.0.143.0/24 @[BGP/170] 00:08:21, localpref 120 AS path: 174 38040 23969 ?, validation-state: unverified > to 123.123.123.1 via ae1.0 to 55.55.55.2 via ae3.0 [BGP/170] 00:07:37, localpref 100 AS path: 174 38040 23969 ?, validation-state: unverified > to 55.55.55.2 via ae3.0 #[Multipath/255] 00:07:37, metric2 0 to 123.123.123.1 via ae1.0 > to 55.55.55.2 via ae3.0 to 55.55.55.2 via ae3.0 123.123.123.1 is my IBGP neighbor where I would like traffic to go 55.55.55.2 is my EBGP neighbor This is the part that is causing issues: #[Multipath/255] 00:07:37, metric2 0 to 123.123.123.1 via ae1.0 > to 55.55.55.2 via ae3.0 to 55.55.55.2 via ae3.0 With "local-preference 120" configured on my IBGP session I would expect all packets to go to my IBGP neighbor (123.123.123.1) - at least that's how it used to work. But for this specific subnet (and many others) the traffic will go directly to my EBGP neighbor. How do I make all traffic go to my IBGP neighbor? Is that "by design" or is it a bug in Junos 18.2R1? Regards Karl _______________________________________________ juniper-nsp mailing list juniper-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/juniper-nsp
