Hello @benolyndav ,
adding a new eBGP session with another peer in the same AS of the ISP does not cause the existing eBGP session to flap.
>> ours side is a stack switch one link from sw1 to isp router 1 link from sw2 to isp router
From a logical point of view your switch stack acts as a single device so the configuration suggested by @paul driver is correct you can play with weight to make the existing link the preferred exit link and you can use your own AS number AS path prepend to make the secondary link less attractive for return traffic coming from the internet.
>>You looking to Loadbalance BGP Traffic or Failover? (then you need to add preference which links to be primary and secondary.
>>>>> some traffic will be going over the second link.
You are likely receiving only a default route or a subset of routes and a default route being your device a switch.
There are different options to put some traffic on the secondary link and to receive some traffic on the secondary link.
First of all, at BGP level it works at prefix level and it is destination based ( normal routing)
So looking in the upstream direction the secondary link can be used for:
specific destination prefixes that are only advertised on the secondary link or that have better BGP attributes then the corresponding received on primary link.
if you want to perform this in BGP you can use a route-map.
ip prefix-list ROUTES-VIA-2PATH seq 5 permit 200.200.200.0/24
ip prefix-list ROUTES-VIA-2PATH seq 5 permit 201.201.201.0/24
and so on
route-map RISE-WEIGHT-2PATH permit 10
match address prefix ROUTES-VIA-2PATH
set weight 55000
route-map RISE-WEIGHT-2PATH permit 20
set weight 40000
router bgp 10
neighbor <secondary > route-map RISE-WEIGHT-2PATH in
This is for the upstream direction = to the internet.
For traffic coming from the internet if you have multiple public IP address blocks, each of them /24 you can make one path the preferred return path for one prefix and other one the preferred return path for the other address block.
To achieve this there are two ways in BGP :
a) being the upstream provider the same, with same AS number you can use MED to influence return path. A lower MED value is the preferred return path for an advertised prefix.
b) using AS-path prepending in a selective way .
For using effectively MED we have to remember that a missing MED is considered equal to zero that is the best possible value.
Let us suppose yo have two address blocks to advertise:
100.100.100.0/24
100.101.101.0/24
MED approach:
ip prefix-list BLOCK1 permit 100.100.100.0/24
ip prefix-list 2 BLOCK2 permit 100.101.101.0/24
! route-map to be used outbound primary link
route-map PRIMARY-OUT permit 10
match address prefix BLOCK1
set metric 50
route-map PRIMARY-OUT permit 20
match address prefix BLOCK2
set metric 1000
! for the secondary we do the opposite
route-map SECONDARY-OUT permit 10
match address prefix BLOCK2
set metric 50
route-map SECONDARY-OUT permit 20
match address prefix BLOCK1
set metric 1000
router bgp 100
neighbor <primary link address > route-map PRIMARY-OUT out
neighbor <secondary link address > route-map SECONDARY-OUT out
The use of selective AS path prepending is similar remembering that no AS path prepending would mean preferred return path
The route-map could be the following
route-map AS-PRI-OUT permit 10
match address prefix BLOCK1
route-map AS-PRI-OUT permit 20
match address prefix BLOCK2
set as-path prepend 10 10 10
Finally, if you have a single address block and you want to put some traffic on the secondary link you can use PBR policy based routing to match on specific sources to be routed via the secondary link.
In a scenario like this all return traffic will come from one link.
Hope to help
Giuseppe
