VRF BGP Multipath - CE receiving one route - Need help

My ce (ce1) is receiving only one prefix from the pe (pe1) even though it's setup for bgp max-paths.  Could you guys/gals help me out?  Thank you in advance.

ce
ce1#sh ip bgp
*> 1.1.1.0/24       166.1.3.2                              0 100 65002 i

pe
* i1.1.1.0/24       193.254.254.3            0    100      0 65002 i
*>i                 193.254.254.2            0    100      0 65002 i


pe1# sh ip bgp v v ce1 nei 166.1.3.1 advertised-routes
   Network          Next Hop            Metric LocPrf Weight Path
Route Distinguisher: 100:1 (default for vrf ce1)
*>i1.1.1.0/24       193.254.254.2            0    100      0 65002 i

pe1# sh ip bgp v v ce1 1.1.1.0 255.255.255.0
BGP routing table entry for 100:1:1.1.1.0/24, version 38
Paths: (2 available, best #2, table ce1)
Multipath: eiBGP
  Advertised to update-groups:
     1
  65002, imported path from 100:3:1.1.1.0/24
    193.254.254.3 (metric 6) from 193.254.254.24 (193.254.254.24)
      Origin IGP, metric 0, localpref 100, valid, internal, multipath
      Extended Community: RT:100:3 RT:100:100
      Originator: 193.254.254.3, Cluster list: 193.254.254.24,
      mpls labels in/out nolabel/28
  65002, imported path from 100:2:1.1.1.0/24
    193.254.254.2 (metric 6) from 193.254.254.24 (193.254.254.24)
      Origin IGP, metric 0, localpref 100, valid, internal, multipath, best
      Extended Community: RT:100:2 RT:100:100
      Originator: 193.254.254.2, Cluster list: 193.254.254.24,
      mpls labels in/out nolabel/30

 

pe1#sh run | sec router bgp
router bgp 100
 no synchronization
 bgp always-compare-med
 bgp log-neighbor-changes
 neighbor 193.254.254.24 remote-as 100
 neighbor 193.254.254.24 update-source Loopback100
 neighbor 193.254.254.24 send-community both
 no auto-summary
 !
 address-family vpnv4
 neighbor 193.254.254.24 activate
 neighbor 193.254.254.24 send-community both
 exit-address-family
 !
 address-family ipv4 vrf ce1
 redistribute connected
 neighbor 166.1.3.1 remote-as 65001
 neighbor 166.1.3.1 activate
 maximum-paths eibgp 6
 no auto-summary
 no synchronization
 bgp suppress-inactive
 exit-address-family

 

Comments

  • One quick question, why you are using redistribute conncted? Any reason on it?

  • Simple way to provide reachability through the cloud.  I was just being lazy :) . Since this is my test lab, I really haven't taken the time to remove it.

  •  address-family ipv4 vrf ce1
     redistribute connected
     neighbor 166.1.3.1 remote-as 65001
     neighbor 166.1.3.1 activate
     maximum-paths eibgp 6
     no auto-summary
     no synchronization
     bgp suppress-inactive
     exit-address-family

    You have maximum-paths towards CE, where you have only one link and only one BGP peering, this doesn't make sense of using single path (which is default in BGP) or maximum-paths and even on PE configuraiton you have only one BGP peering with 193.254.254.24, is there any sense of using maximum-paths? When you have multiple BGP peering and have same BGP attributes , to install those routes  via multiple BGP next-hop, we use maximum-paths (by default is equal load-sharing) and if you enable BGP dmzlink-bw, can use for unequal load sharing.

    If you have more config paste here:

  • I agree with Narayan. Can you please post the topology you are working on? Are you peering over the loopback? 

    If you have a single connection, it does not make any sense to use max-path.

  • I actually had a brain fart.  The load balancing is done in the vrf, not on the ce.  The pe will only send the best ip to the ce.

     

     

     

    image

  • I agree with Narayan. Can you please post the topology you are working on? Are you peering over the loopback? 

    If you have a single connection, it does not make any sense to use max-path.

    I had a simple oversight, but multipath is working correctly.  I was
    looking at the multipath in the wrong direction.  The VRF will send the
    best prefix to the CE.  Loadbalancing is accomplished through the mpls
    vpn network.  I was thinking the two prefixes would appear in the CE's
    BGP table.  They appear in the vrf table, where i have the max-path
    eibgp setting configured.

    In response to your question, the loadbalancing in my lab is via the
    mplsvpn network, not the CE to PE connections; so the multiple paths are
    via the different labels to each FEC in the mpls network.

     

    image

  • Got it!! Nice drawing by the way. Thanks for clarifying..

  • Got it. That makes perfect sense now, nice drawing by the way. Thanks for clarifying :)

  • peetypeety ✭✭✭

    So...you may want to ensure that each PE marks routes in the VRF with a unique RD.  With unique RD, the "final" PE (in this case, PE1) gets to make the bestpath selection, as the multiple paths are considered unique within VPNV4 space (the RD is 48 bits of the 96-bit route).  Without unique RD, route reflectors can thin out the choices mid-network.

Sign In or Register to comment.