10-08-2004 07:39 AM
Hi,
I set up a lab enviroment at home to practice MPLS/VPN. I have dual PE routers (2611) and dual CE routers (2500). These routers form a square topology.
R2 -------- R1
| |
| |
| |
| |
R4 -------- R3
\ /
\ /
10.10.1.0/24
10.10.2.0/24
10.10.3.0/24
10.10.4.0/24
10.10.5.0/24
10.10.6.0/24
I have OSPF + iBGP running between PEs and OSPF only between a PE and a CE. Both CE routers announce 10.10.1.0 - 10.10.6.0/24 subnets to PEs via OSPF. On PE R2, I aggregated these 10.x.x.x subnets into 10.10.0.0/21, but not on R1. The RID for R1 is 1.1.1.1, for R2 is 2.2.2.2, for R3 is 3.3.3.3 and for R4 is 4.4.4.4.
When I pull the link between R2 and R4, I expected OSPF on R2 would detect the link status change and perform SPF calculation. As a result, the OSPF routes in R2's VRF table will be removed and R2 will begin to use the route announced by PE R1 for the 10.10.x.x network.
However, when I looked at the following logs. it seems like BGP(1) process kicked in first and realize all routes disappeared. OSPF didn't react to the link pull until 5 seconds later. Also, R2 withdrawned the route of 3.3.3.3 and 4.4.4.4.
*Mar 1 04:58:07.465: %LINEPROTO-5-UPDOWN: Line protocol on Interface Ethernet0/0, changed state to down
*Mar 1 04:58:07.597: BGP(1): route 65000:2:172.16.2.0/24 down
*Mar 1 04:58:07.601: BGP(1): route 65000:2:3.3.3.3/32 down
*Mar 1 04:58:07.601: BGP(1): route 65000:2:4.4.4.4/32 down
*Mar 1 04:58:07.605: BGP(1): route 65000:2:10.10.1.0/24 down
*Mar 1 04:58:07.605: BGP(1): route 65000:2:10.10.2.0/24 down
.....................
*Mar 1 04:58:07.613: BGP(1): add request for 65000:2:3.3.3.3/32
*Mar 1 04:58:07.617: BGP(1): Add request installing 3.3.3.3/32 -> 1.1.1.1 to Home IP table
*Mar 1 04:58:07.617: BGP(1): add request for 65000:2:4.4.4.4/32
*Mar 1 04:58:07.617: BGP(1): Add request installing 4.4.4.4/32 -> 1.1.1.1 to Home IP table
*Mar 1 04:58:07.621: BGP(1): add request for 65000:2:10.10.1.0/24
*Mar 1 04:58:07.621: BGP(1): Add request installing 10.10.1.0/24 -> 1.1.1.1 to Home IP table
*Mar 1 04:58:07.625: BGP(1): add request for 65000:2:10.10.2.0/24
....................
*Mar 1 04:58:07.641: BGP(1): nettable_walker 65000:2:172.16.2.0/24 route sourced locally
*Mar 1 04:58:07.641: BGP(1): 1.1.1.1 computing updates, afi 1, neighbor version 136, table version 147, starting at 0.0.0.0
*Mar 1 04:58:07.645: BGP(1): 1.1.1.1 send unreachable 65000:2:3.3.3.3/32
*Mar 1 04:58:07.645: BGP(1): 1.1.1.1 send UPDATE 65000:2:3.3.3.3/32 -- unreachable
*Mar 1 04:58:07.645: BGP(1): 1.1.1.1 send UPDATE 65000:2:4.4.4.4/32 -- unreachable
*Mar 1 04:58:07.649: BGP(1): 1.1.1.1 send UPDATE 65000:2:172.16.0.0/24 -- unreachable
...................
*Mar 1 04:58:12.818: - OSPF db timer at 16 min 47 sec 0 9 -
*Mar 1 04:58:12.970: OSPF: running SPF for area 0.0.0.0
*Mar 1 04:58:12.970: OSPF: Initializing to run spf
*Mar 1 04:58:12.970: OSPF: No new path to 172.16.2.1
*Mar 1 04:58:12.970: It is a router LSA 172.16.2.1. Link Count 0
*Mar 1 04:58:12.970: OSPF: Adding Stub nets
*Mar 1 04:58:12.970: OSPF: Entered old delete routine
*Mar 1 04:58:12.970: OSPF: No ndb for STUB NET old route 10.10.1.0, mask /24, next hop 172.16.2.2
*Mar 1 04:58:12.970: OSPF: No ndb for STUB NET old route 4.4.4.4, mask /32, next hop 172.16.2.2
*Mar 1 04:58:12.974: OSPF: No ndb for STUB NET old route 3.3.3.3, mask /32, next hop 172.16.2.2
*Mar 1 04:58:12.974: OSPF: Delete path to router 172.16.0.1 via 172.16.2.2 spf 13
So my questions are:
1. What is BGP (1)? Why this process detect link failure immediately?
2. Why didn't OSPF run SPF immediately after link pull? Is there built-in delay or timer on Cisco to intentially hold on the SPF run for OSPF inside VRF?
3. Why R2 only withdrawn 3.3.3.3 and 4.4.4.4, but not 10.10.x.x network?
Your help is really appreciated!
SShang
10-08-2004 09:44 AM
Hi again!
1) Looks like "bgp fast-external-fallover"
2) OSPF waits 5 seconds to run SPF. This is not depending on the existance of a VRF.
3) Maybe I misunderstood: where do you aggregate? In R2 or R1? In BGP or in OSPF? Configs would be more heplful ;-)
regards
Martin
10-08-2004 10:08 AM
Hi, Martin:
1) bgp fast-external-fallover is what I was thinking about too. However, I didn't have any EBGP session in this case. For iBGP, the session was built by loopback IP.
2) So which parameter determines that 5-second delay for SPF run? How can I tweak it? I am in the impression tha OSPF should run SPF immediately after link status change, right?
3) I tried to aggregate the route on R2 by "aggregate 10.10.0.0/21 summary-only" command in BGP. I will post my config tonight after I go back to home.
Thank you very much again!
SShang
10-08-2004 07:34 PM
Hi,
I copied and pasted the config on R2 as follow:
ip vrf Home
rd 65000:2
route-target export 65000:100
route-target import 65000:100
interface Ethernet0/0
ip vrf forwarding Home
ip address 172.16.2.1 255.255.255.0
half-duplex
!
router bgp 65000
no synchronization
bgp log-neighbor-changes
neighbor 1.1.1.1 remote-as 65000
neighbor 1.1.1.1 password punk
neighbor 1.1.1.1 update-source Loopback0
neighbor 1.1.1.1 timers 1 3
no auto-summary
!
address-family ipv4 vrf Home
redistribute ospf 400 match internal external 1 external 2
no auto-summary
no synchronization
aggregate-address 10.10.0.0 255.255.248.0 summary-only
exit-address-family
!
address-family vpnv4
neighbor 1.1.1.1 activate
neighbor 1.1.1.1 send-community extended
no auto-summary
exit-address-family
!
router ospf 400 vrf Home
log-adjacency-changes
redistribute bgp 65000 metric 10 subnets
network 172.16.2.0 0.0.0.255 area 0.0.0.0
I also noticed that, when I typed "show ip bgp vpnv4 vrf Home" command, there are two entries for each route. One is learned from its iBGP peer (1.1.1.1) and another shown as local, but is identified by physical IP address (172.16.2.2). When I pull the link, the IP address of 172.16.2.2 will be lost:
C2600-R2#show ip bgp vpnv4 vrf Home
Network Next Hop Metric LocPrf Weight Path
Route Distinguisher: 65000:2 (default for vrf Home)
*> 3.3.3.3/32 172.16.2.2 75 32768 ?
* i 1.1.1.1 11 100 0 ?
*> 4.4.4.4/32 172.16.2.2 11 32768 ?
* i 1.1.1.1 75 100 0 ?
*> 10.10.1.0/24 172.16.2.2 74 32768 ?
* i 1.1.1.1 74 100 0 ?
*> 10.10.2.0/24 172.16.2.2 20 32768 ?
* i 1.1.1.1 20 100 0 ?
*> 10.10.3.0/24 172.16.2.2 20 32768 ?
* i 1.1.1.1 20 100 0 ?
So is this normal?
Thanks!
SShang
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide