Manual:BGP Load Balancing with two interfaces

From MikroTik Wiki
Jump to: navigation, search
Version.png

Applies to RouterOS: 3, v4

NB: RouterOS version 3.13 or later with routing-test package is required for this to work

In these examples we show how to do load balancing when there are multiple equal cost links between two BGP routers. The "multiple recursive next-hop resolution" feature is used to achieve that.

The BGP session is established between loopback interfaces; update-source configuration setting is used to bind the BGP connection to the right interface.

Example with iBGP

Network Diagram

Ibgp load bal.png

Configuration

On Router A:

# loopback interface
/interface bridge add name=lobridge

# addresses
/ip address add address=1.1.1.1/24 interface=ether1
/ip address add address=2.2.2.1/24 interface=ether2
/ip address add address=9.9.9.1/32 interface=lobridge

# ECMP route to peer's loopback
/ip route add dst-address=9.9.9.2/32 gateway=1.1.1.2,2.2.2.2

# BGP
/routing bgp instance set default as=65000
/routing bgp add name=peer1 remote-address=9.9.9.2 remote-as=65000 update-source=lobridge

On Router B:

# loopback interface
/interface bridge add name=lobridge

# addresses
/ip address add address=1.1.1.2/24 interface=ether1
/ip address add address=2.2.2.2/24 interface=ether2
/ip address add address=9.9.9.2/32 interface=lobridge

# ECMP route to peer's loopback
/ip route add dst-address=9.9.9.1/32 gateway=1.1.1.1,2.2.2.1

# BGP
/routing bgp instance set default as=65000
/routing bgp add name=peer1 remote-address=9.9.9.1 remote-as=65000 update-source=lobridge

# a route to advertise
/routing bgp network add network=4.4.4.0/24

Results

Check that BGP connection is established:

[admin@B] > /routing bgp peer print status
Flags: X - disabled
0   name="peer1" instance=default remote-address=9.9.9.1 remote-as=65000
    tcp-md5-key="" nexthop-choice=default multihop=no route-reflect=no hold-time=3m
    ttl=255 in-filter="" out-filter="" address-families=ip
    update-source=lobridge default-originate=no remote-id=1.1.1.1
    local-address=9.9.9.2 uptime=28s prefix-count=0 updates-sent=1
    updates-received=0 withdrawn-sent=0 withdrawn-received=0 remote-hold-time=3m
    used-hold-time=3m used-keepalive-time=1m refresh-capability=yes
    as4-capability=yes state=established

Route table on Router A:

[admin@A] > /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        G GATEWAY                 DISTANCE INTER...
0 ADC  1.1.1.0/24         1.1.1.1                                   0        ether1
1 ADC  2.2.2.0/24         2.2.2.1                                   0        ether2
2 ADb  4.4.4.0/24                         r 9.9.9.2                 200      ether1
                                                                             ether2
3 ADC  9.9.9.1/32         9.9.9.1                                   0        lobridge
4 A S  9.9.9.2/32                         r 1.1.1.2                 1        ether1
                                          r 2.2.2.2                          ether2
[admin@A] > /ip route print detail
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
0 ADC  dst-address=1.1.1.0/24 pref-src=1.1.1.1 interface=ether1 distance=0 scope=10

1 ADC  dst-address=2.2.2.0/24 pref-src=2.2.2.1 interface=ether2 distance=0 scope=10

2 ADb  dst-address=4.4.4.0/24 gateway=9.9.9.2 interface=ether1,ether2
       gateway-state=recursive distance=200 scope=40 target-scope=30
       bgp-local-pref=100 bgp-origin=igp received-from=9.9.9.2

3 ADC  dst-address=9.9.9.1/32 pref-src=9.9.9.1 interface=lobridge distance=0 scope=10

4 A S  dst-address=9.9.9.2/32 gateway=1.1.1.2,2.2.2.2 interface=ether1,ether2
       gateway-state=reachable,reachable distance=1 scope=30 target-scope=10

The route 4.4.4.0./24 is installed in Linux kernel now with two nexthops: 1.1.1.2 (on ether1) and 2.2.2.2 (on ether2).

Example with eBGP

Network Diagram

Ebgp load bal.png

Configuration

Here the example given above is further developed for eBGP case. By default, eBGP peers are required to be directly reachable. If we are using loopback interfaces, they technically are not, so multihop=yes configuration setting must be specified.

On Router A:

/routing bgp instance set default as=65000
/routing bgp set peer1 remote-address=9.9.9.2 remote-as=65001 update-source=lobridge multihop=yes

On Router B:

/routing bgp instance set default as=65001
/routing bgp set peer1 remote-address=9.9.9.1 remote-as=65000 update-source=lobridge multihop=yes

Results

If we now print the route table on Router A, we see that the route from Router B is there, but it's not active:

...
2  Db  dst-address=4.4.4.0/24 gateway=9.9.9.2 interface="" gateway-state=unreachable
       distance=20 scope=40 target-scope=10 bgp-as-path="65001" bgp-origin=igp
       received-from=9.9.9.2
...

This is because eBGP routes are installed with lesser target-scope by default. To solve this, setup routing filter that sets larger target-scope:

/routing filter add chain=bgp-in set-target-scope=30
/routing bgp set peer1 in-filter=bgp-in

Or else, modify scope attribute of the static route:

/ip route set [find dst-address=9.9.9.2/32] scope=10

Either way, the route to 4.4.4.0/24 should be active now:

2 ADb  dst-address=4.4.4.0/24 gateway=9.9.9.2 interface=ether1,ether2
       gateway-state=recursive distance=20 scope=40 target-scope=10
       bgp-as-path="65001" bgp-origin=igp received-from=9.9.9.2

Notes

  • BGP itself as protocol does not supports ECMP routes. When a recursively resolved BGP route is propagated further in the network, only one nexthop can be selected (as described here) and included in the BGP UPDATE message.