topic Re: CP Cluster bgp question in General Topics

CP Cluster bgp question

Gongya_Yu — Mon, 27 May 2024 02:08:43 GMT

Cluster member A with physical IP 192.168.1.1/24
Cluster member B with physical IP 192.168.1.2/24

Member A with the following for export

set routemap default id 100 on

set routemap default id 100 allow

set routemap default id 100 match network 0.0.0.0/0 exact

set routemap default id 100 match protocol static

set routemap default id 100 action nexthop ip 192.168.1.1

Member B with the following for export

set routemap default id 100 on

set routemap default id 100 allow

set routemap default id 100 match network 0.0.0.0/0 exact

set routemap default id 100 match protocol static

set routemap default id 100 action nexthop ip 192.168.1.2

When Cluster fails over, a blackhold will occur, right ?

Reason:

Before failover, member B does not have BGP peering, just synced with Member A for all the routes which points 192.168.1.1 as next-hop. When failover occurs and before member B establishes BGP relationship and advertises new next-hop (192.168.1.2), a blackhold occurs.

Am I right ??

Re: CP Cluster bgp question

emmap — Mon, 27 May 2024 02:13:30 GMT

When setting up dynamic routing on a CP cluster, all peers must use the cluster interface VIP to peer with and route to. Routing/peering to individual gateway IPs is not supported. Hence the configuration on each cluster member must be identical.

Tip: configure the router-id to a cluster VIP before configuring BGP, again identical across the cluster.

Re: CP Cluster bgp question

JozkoMrkvicka — Mon, 27 May 2024 05:45:03 GMT

Yes, without BGP Graceful Restart and/or BFD features enabled, once failover occurs, there is short outage for BGP during re-establishment of peering from new member. It is described within sk175923.

Best practise is to enable BGP Graceful Restart (sk100499) and/or Bidirectional Forwarding Detection (BFD) with cBIT detection (sk175923) to mitigate this problem.

Re: CP Cluster bgp question

Gongya_Yu — Mon, 27 May 2024 08:00:07 GMT

I think it will be blackheld even with Graceful Restart enabled.
Graceful Restart only helps when next-hop points to the VIP. This is the way we are doing.
But if standby node uses its own physical IP, a blackhold will occur during the failover.

Right ?

Re: CP Cluster bgp question

Gongya_Yu — Mon, 27 May 2024 09:36:02 GMT

thanks so much for clarification !!

Re: CP Cluster bgp question

JozkoMrkvicka — Tue, 28 May 2024 19:32:17 GMT

BGP should be configured to use VIP in cluster. You should also use the same VIP as router-id. Config for BGP within all cluster members should be the same.

I also dont get what is the goal here. You want to propagate only default IPv4 static route ? What is config for BGP peers (show configuration bgp) ?

You should use preference statement within routemaps to specify where to use routemap with "default" name.