10-30-2013 12:27 PM - edited 03-04-2019 09:27 PM
At both the remote site, I've a Cisco 2811 connected to two different MPLS providers..say ProviderA and ProviderB running BGP using the same AS number. The same is at the core 7206, but ProviderA has a "weight" assigned as 55555. When I assign the same weight (or even different, higher weight number than the default 32678 - neighbor 10.xxx.xxx.xxx weight 55555) to the MPLS circuit from the ProviderA at the remote site, this circuit starts to bounce every 30 seconds or so. Remove the weight parameter from the remote site then the circuit becomes stable again.
My objective is to make ProviderA MPLS circuit to be the primary all the time while ProviderB MPLS will be the backup circuit.
R1 (core)
router bgp 11111
no synchronization
bgp log-neighbor-changes
network 192.168.xxx.xxx
neighbor 10.xxx.xxx.xxx remote-as 11x
neighbor 10.xxx.xxx.xxxdefault-originate
neighbor 10.xxx.xxx.xxx weight 55555
neighbor 68.xxx.xxx.xxxremote-as 12xxx
neighbor 68.xxx.xxx.xxxdefault-originate
default-information originate
no auto-summary
R2 (remote)
router bgp 22222
no synchronization
bgp log-neighbor-changes
network 192.168.xxx.xxx
neighbor 10.xxx.xxx.xxx remote-as 11x
neighbor 68.xxx.xxx.xxxremote-as 12xxx
no auto-summary
Any help is greatly appreciated.
Thanks.
10-30-2013 01:14 PM
I wouldn't think enabling a weight would cause the peer to flap. Can you run debugs (debug bgp all), enable the weight parameter, and then post the results of the debug?
HTH,
John
*** Please rate all useful posts ***
11-01-2013 03:03 AM
Hi,
a) The default weight attribute value is 0, not 32678.
32678 is the value assigned to the prefixes inserted to the BGP table by the router itself.
b) I can imagine using a value higher than 32678 could confuse your router under some conditions (your providers using as-override, e.g., and you advertising the prefixes received form one provider to the other provider in your core site) and it could accept its own LAN prefix 192.168.xxx.xxx advertised from the provider.
It would stop advertising it to the provider then and finally the prefix would start flapping from your core router point of view.
c) I'd recommend you to try using some weight value lower than 32768 on the remote site (and also in the core actually).
10000 should be good enough.
In that case, the router will still prefer prefixes received from provider A over the same prefixes received from provider B.
But this will not influence the prefixes advertised by the router itself any way.
Best regards,
Milan
11-05-2013 05:06 PM
Thanks Milan and John for your response. I stop using the weight parameter and but set local_preference value at 150 as recommended by cicso tech and it works fine for now.
Thanks.
Hieu
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