Description
RouterOS allows to create multiple Virtual Routing and Forwarding instances on a single router. This is useful for BGP based MPLS VPNs. Unlike BGP VPLS, which is OSI Layer 2 technology, BGP VRF VPNs work in Layer 3 and as such exchange IP prefixes between routers. VRFs solve the problem of overlapping IP prefixes, and provide the required privacy (via separated routing for different VPNs).
It is possible to set up vrf-lite setups or use multi-protocol BGP with VPNv4 address family to distribute routes from VRF routing tables - not only to other routers, but also to different routing tables in the router itself.
Configuration
VRF table is created in /ip vrf menu. After the VRF config is created routing table mapping is added (dynamic table with the same name is created). Each active VRF will always have mapped routing table.
[admin@arm-bgp] /ip/vrf> print Flags: X - disabled; * - builtin 0 * name="main" interfaces=all [admin@arm-bgp] /routing/table> print Flags: D - dynamic; X - disabled, I - invalid; U - used 0 D name="main" fib
Note that order of the added VRFs are significant. To properly match which interface will belong to the VRF care must be taken to place VRFs in the correct order (matching is done starting from the top entry, just like firewall rules).
Lets look at the following example:
[admin@arm-bgp] /ip/vrf> print Flags: X - disabled; * - builtin 0 * name="main" interfaces=all 1 name="myVrf" interfaces=lo_vrf
Since the first entry is matching all the interfaces, second VRF will not have any interfaces added. To fix the problem order of the entries must be changed.
[admin@arm-bgp] /ip/vrf> move 1 0 [admin@arm-bgp] /ip/vrf> print Flags: X - disabled; * - builtin 0 name="myVrf" interfaces=lo_vrf 1 * name="main" interfaces=all
Connected routes from the interfaces interfaces assigned to the VRF will be installed in the right routing table automatically.
When interface is assigned to the VRF as well as connected routes it does not mean that RouterOS services will magically know which VRF to use just by specifying IP address in the configuration. Each service needs VRF support to be added and explicit configuration. Whether service has VRF support and has VRF configuration options refer to appropriate service documentation.
For example, lets make SSH service to listen for connections on the interfaces belonging to the VRF:
[admin@arm-bgp] /ip/service> set ssh vrf=myVrf [admin@arm-bgp] /ip/service> print Flags: X, I - INVALID Columns: NAME, PORT, CERTIFICATE, VRF # NAME PORT CERTIFICATE VRF 0 telnet 23 main 1 ftp 21 2 www 80 main 3 ssh 22 myVrf 4 X www-ssl 443 none main 5 api 8728 main 6 winbox 8291 main 7 api-ssl 8729 none main
Adding routes to the VRF is as simple as specifying routing-table parameter when adding the route and specifying in which routing table to resolve the gateway by specifying @name after the gateway IP:
/ip route add dst-address=192.168.1.0/24 gateway=172.16.1.1@myVrf routing-table=myVrf
Traffic leaking between VRFs is possible if the gateway is explicitly set to be resolved in another VRF, for example:
# add route in the myVrf, but resolve the gateway in the main table /ip route add dst-address=192.168.1.0/24 gateway=172.16.1.1@main routing-table=myVrf # add route in the main table, but resolve the gateway in the myVrf /ip route add dst-address=192.168.1.0/24 gateway=172.16.1.1@myVrf
If the gateway configuration does not have explicitly configured table to be resolved in, then it is considered, that gateway should be resolved in the "main" table.
Examples
Simple VRF-Lite setup
Lets consider setup where we need two customer VRFs that require access to the internet:
/ip address add address=172.16.1.2/24 interface=public add address=192.168.1.1/24 interface=ether1 add address=192.168.2.1/24 interface=ether2 /ip route add gateway=172.16.1.1 # add VRF configuration /ip vrf add name=cust_a interface=ether1 place-before 0 add name=cust_b interface=ether2 place-before 0 # add vrf routes /ip route add gateway=172.16.1.1@main routing-table=cust_a add gateway=172.16.1.1@main routing-table=cust_b
Static inter-VRF routes
In general it is recommended that all routes between VRF should be exchanged using BGP local import and export functionality. If that is not enough, static routes can be used to achieve this so-called route leaking.
There are two ways to install a route that has gateway in different routing table than the route itself.
The first way is to explicitly specify routing table in gateway field when adding route. This is only possible when leaking a route and gateway from the "main" routing table to a different routing table (VRF). Example:
# add route to 5.5.5.0/24 in 'vrf1' routing table with gateway in the main routing table add dst-address=5.5.5.0/24 gateway=10.3.0.1@main routing-table=vrf1
The second way is to explicitly specify interface in gateway field. The interface specified can belong to a VRF instance. Example:
# add route to 5.5.5.0/24 in the main routing table with gateway at 'ether2' VRF interface add dst-address=5.5.5.0/24 gateway=10.3.0.1%ether2 routing-table=main # add route to 5.5.5.0/24 in the main routing table with 'ptp-link-1' VRF interface as gateway add dst-address=5.5.5.0/24 gateway=ptp-link-1 routing-table=main
As can be observed, there are two variations possible - to specify gateway as ip_address%interface or to simply specify interface. The first should be used for broadcast interfaces in most cases. The second should be used for point-to-point interfaces, and also for broadcast interfaces, if the route is a connected route in some VRF. For example, if you have address 1.2.3.4/24 on interface ether2 that is put in a VRF, there will be connected route to 1.2.3.0/24 in that VRF's routing table. It is acceptable to add static route 1.2.3.0/24 in a different routing table with interface-only gateway, even though ether2 is a broadcast interface:
add dst-address=1.2.3.0/24 gateway=ether2 routing-table=main
The simplest MPLS VPN setup
In this example, a rudimentary MPLS backbone (consisting of two Provider Edge (PE) routers PE1 and PE2) is created and configured to forward traffic between Customer Edge (CE) routers CE1 and CE2 routers that belong to cust-one VPN.
CE1 Router
/ip address add address=10.1.1.1/24 interface=ether1 # use static routing /ip route add dst-address=10.3.3.0/24 gateway=10.1.1.2
CE2 Router
/ip address add address=10.3.3.4/24 interface=ether1 /ip route add dst-address=10.1.1.0/24 gateway=10.3.3.3
PE1 Router
/interface bridge add name=lobridge /ip address add address=10.1.1.2/24 interface=ether1 /ip address add address=10.2.2.2/24 interface=ether2 /ip address add address=10.5.5.2/32 interface=lobridge /ip vrf add name=cust-one interfaces=ether1 /mpls ldp add enabled=yes transport-address=10.5.5.2 /mpls ldp interface add interface=ether2 /routing bgp template set default as=65000 /routing bgp vpn add vrf=cust-one redistribute=connected \ route-distinguisher=1.1.1.1:111 \ import-route-targets=1.1.1.1:111 \ export-route-targets=1.1.1.1:111 \ label-allocation-policy=per-vrf /routing bgp connection add template=default remote.address=10.5.5.3 address-families=vpnv4 local.address=10.5.5.2 # add route to the remote BGP peer's loopback address /ip route add dst-address=10.5.5.3/32 gateway=10.2.2.3
PE2 Router (Cisco)
ip vrf cust-one rd 1.1.1.1:111 route-target export 1.1.1.1:111 route-target import 1.1.1.1:111 exit interface Loopback0 ip address 10.5.5.3 255.255.255.255 mpls ldp router-id Loopback0 force mpls label protocol ldp interface FastEthernet0/0 ip address 10.2.2.3 255.255.255.0 mpls ip interface FastEthernet1/0 ip vrf forwarding cust-one ip address 10.3.3.3 255.255.255.0 router bgp 65000 neighbor 10.5.5.2 remote-as 65000 neighbor 10.5.5.2 update-source Loopback0 address-family vpnv4 neighbor 10.5.5.2 activate neighbor 10.5.5.2 send-community both exit-address-family address-family ipv4 vrf cust-one redistribute connected exit-address-family ip route 10.5.5.2 255.255.255.255 10.2.2.2
Results
Check that VPNv4 route redistribution is working:
...
Check that the 10.3.3.0 is installed in IP routes, in cust-one route table:
[admin@PE1] > /ip route print Flags: X - disabled, A - active, D - dynamic, C - connect, S - static, r - rip, b - bgp, o - ospf, m - mme, B - blackhole, U - unreachable, P - prohibit # DST-ADDRESS PREF-SRC GATEWAY DISTANCE 0 ADC 10.1.1.0/24 10.1.1.2 ether1 0 1 ADb 10.3.3.0/24 10.5.5.3 recursi... 20 2 ADC 10.2.2.0/24 10.2.2.2 ether2 0 3 ADC 10.5.5.2/32 10.5.5.2 lobridge 0 4 A S 10.5.5.3/32 10.2.2.3 reachab... 1
Let's take closer look at IP routes in cust-one VRF. The 10.1.1.0/24 IP prefix is a connected route that belongs to an interface that was configured to belong to cust-one VRF. The 10.3.3.0/24 IP prefix was advertised via BGP as VPNv4 route from PE2 and is imported in this VRF routing table, because our configured import-route-targets matched the BGP extended communities attribute it was advertised with.
[admin@PE1] /routing/route> print detail where routing-table=cust-one ...
The same for Cisco:
PE2#show ip bgp vpnv4 all BGP table version is 5, local router ID is 10.5.5.3 Status codes: s suppressed, d damped, h history, * valid, > best, i - internal, r RIB-failure, S Stale Origin codes: i - IGP, e - EGP, ? - incomplete Network Next Hop Metric LocPrf Weight Path Route Distinguisher: 1.1.1.1:111 (default for vrf cust-one) *>i10.1.1.0/24 10.5.5.2 100 0 ? *> 10.3.3.0/24 0.0.0.0 0 32768 ? PE2#show ip route vrf cust-one Routing Table: cust-one Codes: C - connected, S - static, R - RIP, M - mobile, B - BGP D - EIGRP, EX - EIGRP external, O - OSPF, IA - OSPF inter area N1 - OSPF NSSA external type 1, N2 - OSPF NSSA external type 2 E1 - OSPF external type 1, E2 - OSPF external type 2 i - IS-IS, su - IS-IS summary, L1 - IS-IS level-1, L2 - IS-IS level-2 ia - IS-IS inter area, * - candidate default, U - per-user static route o - ODR, P - periodic downloaded static route Gateway of last resort is not set 10.0.0.0/24 is subnetted, 1 subnets B 10.1.1.0 [200/0] via 10.5.5.2, 00:05:33 10.0.0.0/24 is subnetted, 1 subnets C 10.3.3.0 is directly connected, FastEthernet1/0
You should be able to ping from CE1 to CE2 and vice versa.
[admin@CE1] > /ping 10.3.3.4 10.3.3.4 64 byte ping: ttl=62 time=18 ms 10.3.3.4 64 byte ping: ttl=62 time=13 ms 10.3.3.4 64 byte ping: ttl=62 time=13 ms 10.3.3.4 64 byte ping: ttl=62 time=14 ms 4 packets transmitted, 4 packets received, 0% packet loss round-trip min/avg/max = 13/14.5/18 ms
A more complicated setup (changes only)
As opposed to the simplest setup, in this example we have two customers: cust-one and cust-two.
We configure two VPNs for then, cust-one and cust-two respectively, and exchange all routes between them. (This is also called "route leaking").
Note that this could be not the most typical setup, because routes are usually not exchanged between different customers. In contrast, by default it should not be possible to gain access from one VRF site to a different VRF site in another VPN. (This is the "Private" aspect of VPNs.) Separate routing is a way to provide privacy; and it is also required to solve the problem of overlapping IP network prefixes. Route exchange is in direct conflict with these two requirement but may sometimes be needed (e.g. temp. solution when two customers are migrating to single network infrastructure).
CE1 Router, cust-one
/ip route add dst-address=10.4.4.0/24 gateway=10.1.1.2
CE2 Router, cust-one
/ip route add dst-address=10.4.4.0/24 gateway=10.3.3.3
CE1 Router,cust-two
/ip address add address=10.4.4.5 interface=ether1 /ip route add dst-address=10.1.1.0/24 gateway=10.3.3.3 /ip route add dst-address=10.3.3.0/24 gateway=10.3.3.3
PE1 Router
# replace the old BGP VPN with this: /routing bgp vpn add vrf=cust-one redistribute=connected \ route-distinguisher=1.1.1.1:111 \ import-route-targets=1.1.1.1:111,2.2.2.2:222 \ export-route-targets=1.1.1.1:111
PE2 Router (Cisco)
ip vrf cust-one rd 1.1.1.1:111 route-target export 1.1.1.1:111 route-target import 1.1.1.1:111 route-target import 2.2.2.2:222 exit ip vrf cust-two rd 2.2.2.2:222 route-target export 2.2.2.2:222 route-target import 1.1.1.1:111 route-target import 2.2.2.2:222 exit interface FastEthernet2/0 ip vrf forwarding cust-two ip address 10.4.4.3 255.255.255.0 router bgp 65000 address-family ipv4 vrf cust-two redistribute connected exit-address-family
Variation: replace the Cisco with another MT
PE2 Mikrotik config
/interface bridge add name=lobridge /ip address add address=10.2.2.3/24 interface=ether1 add address=10.3.3.3/24 interface=ether2 add address=10.4.4.3/24 interface=ether3 add address=10.5.5.3/32 interface=lobridge /ip vrf add name=cust-one interfaces=ether2 add name=cust-two interfaces=ether3 /mpls ldp add enabled=yes transport-address=10.5.5.3 /mpls ldp interface add interface=ether1 /routing bgp template set default as=65000 /routing bgp vpn add vrf=cust-one redistribute=connected \ route-distinguisher=1.1.1.1:111 \ import-route-targets=1.1.1.1:111,2.2.2.2:222 \ export-route-targets=1.1.1.1:111 \ add vrf=cust-two redistribute=connected \ route-distinguisher=2.2.2.2:222 \ import-route-targets=1.1.1.1:111,2.2.2.2:222 \ export-route-targets=2.2.2.2:222 \ /routing bgp connection add template=default remote.address=10.5.5.2 address-families=vpnv4 local.address=10.5.5.3 # add route to the remote BGP peer's loopback address /ip route add dst-address=10.5.5.2/32 gateway=10.2.2.2
Results
The output of /ip route print now is interesting enough to deserve detailed observation.
[admin@PE2] /ip route> print Flags: X - disabled, A - active, D - dynamic, C - connect, S - static, r - rip, b - bgp, o - ospf, m - mme, B - blackhole, U - unreachable, P - prohibit # DST-ADDRESS PREF-SRC GATEWAY DISTANCE 0 ADb 10.1.1.0/24 10.5.5.2 recurs... 20 1 ADC 10.3.3.0/24 10.3.3.3 ether2 0 2 ADb 10.4.4.0/24 20 3 ADb 10.1.1.0/24 10.5.5.2 recurs... 20 4 ADb 10.3.3.0/24 20 5 ADC 10.4.4.0/24 10.4.4.3 ether3 0 6 ADC 10.2.2.0/24 10.2.2.3 ether1 0 7 A S 10.5.5.2/32 10.2.2.2 reacha... 1 8 ADC 10.5.5.3/32 10.5.5.3 lobridge 0
The route 10.1.1.0/24 was received from remote BGP peer and is installed in both VRF routing tables.
The routes 10.3.3.0/24 and 10.4.4.0/24 are also installed in both VRF routing tables. Each is as connected route in one table and as BGP route in another table. This has nothing to do with their being advertised via BGP. They are simply being "advertised" to local VPNv4 route table and locally reimported after that. Import and export route-targets determine in which tables they will end up.
This can be deduced from its attributes - they don't have the usual BGP properties. (Route 10.4.4.0/24.)
[admin@PE2] /routing/route> print detail where routing-table=cust-one ...
References
RFC 4364: BGP/MPLS IP Virtual Private Networks (VPNs)
MPLS Fundamentals, chapter 7, Luc De Ghein, Cisco Press 2006