Routing

Hello,

I try to test with configuring Multihoming on the PE Router for VPLS.

PE1:

routing-instances {
VPLS-TESTING {
instance-type vpls;
vlan-id 782;
interface xe-0/0/1.782;
vrf-target target:111:111;
protocols {
vpls {
site SITE1 {
site-identifier 2;
multi-homing;
site-preference primary;
interface xe-0/0/1.782;
}
}
}
}
}

PE2:

routing-instances {
VPLS-TESTING {
instance-type vpls;
vlan-id 782;
interface ae4.782;
vrf-target target:111:111;
protocols {
vpls {
site SITE1-BKUP {
site-identifier 2;
multi-homing;
site-preference backup;
interface ae4.782;
}
}
}
}
}

i have tested without including multi-homing statement, the primary and backup site still work well. The RR will chose the routes of  primary site because its local preference is higher than backup site. So, the question is wheather i should include multi-homing statement or not  ?

i have read the information on https://www.juniper.net/documentation/en_US/junos/topics/usage-guidelines/vpns-configuring-vpls-multihoming.html :

When a CE device is connected to the same VPLS site on more than one PE router, include the multi-homing statement on all associated PE routers. Configuration of this statement tracks BGP peers. If no BGP peer is available, VPLS deactivates all active interfaces for a site.

But i acctually dont understand its function !

Thanks for any answer !

The multi-homing is used to prevent certain scenarios that could result in a layer 2 loop, such as if your backup PE loses its connection to the core.  In that case, your backup would assume primary role due to it no longer receiving label blocks from the actual primary PE.

thanks for your answer,

But if the primary router lost connection to the core, how the loop can occur ? Because at that sittuation, the others router just communicate with the backup router (now it became the primary). Althoght the link between CE and old primary router still up, but this router did not connect to core network anymore.  

I assume it has more to do with preventing black holes than preventing loops. Even if the primary PE fails, the CE could still send traffic to it for a short amount of time because the MAC/ARP tables have entries pointing to the failed PE. If the link is deactivated after a failure, those entries are flushed immediately.

Thanks so much !

I understand carefully now ! 🙂