Hello @goyourmin ,
if the use of a dynamic routing protocol is not possible the configuration of static routes need to be careful to avoid issues with possible routing loops and with excessive ARP activiity.
a) ARP activity when using a
default static route
without an IP next-hop
A static route like the following one:
ip route 0.0.0.0 0.0.0.0 eth0/3 (ISP main route)
makes the local device to make an ARP request for each different destination in the internet instead of resolving the IP next-hop once. So if you have 10,000 different destinations you will have 10,000 ARP entries in the ARP table all of them resolving to the ISP router interface MAC address. This is a waste of resources and it can lead to problems if the ARP table size is too big.
So you need to use static routes where you use IP next-hops. I hope this point is clear now.
b) configuring static routes in a redundant environment requires attention to avoid possible routing loops
With the configuration that you have proposed there is a potential for routing loops
Your proposed config is the following:
- L3-switch-1
ip route 0.0.0.0 0.0.0.0 eth0/3 (ISP main route)
ip route 0.0.0.0 0.0.0.0 eth0/0 172.16.0.2
- L3-switch-2
ip route 0.0.0.0 0.0.0.0 eth0/3 100 (ISP backup route)
ip route 0.0.0.0 0.0.0.0 eth0/0 172.16.0.1
Configuration on L3 switch 1 should be :
ip route 0.0.0.0 0.0.0.0 eth0/3 100.100.100.1
+
ip route 0.0.0.0 0.0.0.0 eth0/0 172.16.0.2 100
the second static route must be floating in order to be used only when primary route is removed from routing table ( link to ISP router fails)
With your current proposed configuration you have two static routes on
L3switch-1
one pointing to ISP router and one pointing to the neighbor
L3Switch-2
but they are both used.
Depending on CEF load balancing decisions on
L3switch-1
if a packet is sent to
L3 switch-2
using the second
default route, L3switch-2
will send it back to
L3Switch-1
that will send it again to
L3switch-2
because the load balancing decision is based on an hash of IP source address and IP destination address, but it does not consider the input interface.
So you have a routing loop and you can solve it by making the route to
L3 switch2
floating ( adding an AD value greater then 1, 100 as in my example)
c) other considerations on your lab
In real world the ISP router would send a
default route
in BGP to the eBGP neighbors, with your current setup this does not happen and you use
default static routes on L3Switch-1 and L3switch-2
On router ISP you can advertise a
default route
in BGP
you need
ip route 0.0.0.0.0 0.0.0.0 A.B.C.D
+
router bgp 1
network 0.0.0.0
At this point you can remove the
default static routes
on
L3switch-1 , L3switch-2
toward ISP router and you can add an iBGP session between them
L3-switch1
route-map RISE-PREF permit 10
set preference 200
router bgp 2
neighbor 100.100.100.1 route-map RISE-PREF in
neighbor 172.16.0.2 remote-as 2
L3 switch-2
router bgp 2
neighbor 172.16.0.1 remote-as 2
and you can remove also the floating static routes pointing to each other.
L3switch-2
will select the iBGP path from
L3switch-1
for the higher local preference 200 over the
eBGP route via ISP router
Hope to help
Giuseppe
