Hello lovejit1313,
the customer has 8 sites and you provide an MPLS L3 VPN service to them.
In one of these 8 sites there is a firewall providing Internet access to all 8 sites.
This means that you are currently sending a default route 0.0.0.0/0 to the PE node connected to this site and the PE node is sending a VPNv4 route to all other sites announcing the default route with appropriate route target value so that all remote PE nodes import the route and can advertise it to the locally connected CE node.
Now, the customer is asking to provide internet access via a second site.
Ideally you should replicate the current infrastructure on this second site.
However, there are some considerations that need to be taken in account.
You are likely using route reflector servers and your PE nodes have MP iBGP sessions with route reflectors.
The PE node of the second VRF site with internet access must use a different route distinguisher to avoid competition between the two VPNv4 default routes on the RRS.
Example:
the VPNv4 routes
65000:100:0.0.0.0/0 with RT 65000:200
65000:121:0.0.0.0/0 with RT 65000:200
are considered not comparable by RRS and both are propagated to all remote PE nodes.
Remote PE nodes will import both routes if they import RT 65000:200.
However, each PE node will make a best path choice based on :
IGP metric to PE nodes PE1 and PE2 generating the default route
If IGP metric is the same the PE node with lowest BGP router-id is used.
Depending on your backbone topology the IGP can provide already some traffic load sharing to PE1, PE2.
However, if you want to choice the best route on each PE node you should use BGP standard communities or you can set an higher weight for the default route coming from a specific next-hop (PE1 or PE2 loopback address)
You cannot use local-preference because it would propagate to other nodes in same AS.
You can try to use a route-map applied within the vrf with an import-map command.
On each PEx with x different from PE1 and PE2:
access-list 11 remark PE1 loopback
access-list 11 permit host 10.11.11.11
access-list 12 remark PE2 loopback
access-list 12 permit host 10.22.22.22
prefix-list ONLY-DEFAULT permit 0.0.0.0/0
route-map SET-WEIGHT-DEF-ROUTE permit 10
match ip address prefix ONLY-DEFAULT
match ip next-hop 12
set weight 200
route-map SET-WEIGHT-DEF-ROUTE permit 20
match ip address prefix ONLY-DEFAULT
match ip next-hop 11
set weight 100
route-map SET-WEIGHT-DEF-ROUTE permit 30
vrf CUSTOMER
import-map SET-WEIGHT-DEF-ROUTE
Or in this way or using BGP communities you can make some sites to prefer one or the other exit point.
The last block in the route-map is to accept all other routes (permit 30 with no match or set conditions)
To be noted both firewalls must be able to serve all sites in case of fault of one site.
Half PE nodes can prefer PE1's route and half PE nodes can prefer PE2's route in this way
Hope to help
Giuseppe
