OSPF split connected subnet into different processes

[hash2k2]

Hi,

on our branch offfices we have Cisco routers with two processes connected each to our HQ via a separate ISP provided MPLS.

In both processes we have the command:

OSPF process 1

redistribute connected subnets route-map MPLS-1

OSPF process 2

redistribute connected subnets route-map MPLS-2

Then we have two route maps saying something like:

route-map MPLS-1 permit 10
match interface GigabitEthernet0/0.200
set metric 90
set metric-type type-1
!
route-map MPLS-1 permit 20
match interface GigabitEthernet0/0.201
set metric 90
set metric-type type-1
!
route-map MPLS-1 permit 30
match interface GigabitEthernet0/0.202
set metric 90
set metric-type type-1

route-map MPLS-2 permit 10
match interface GigabitEthernet0/0.200
set metric 200
set metric-type type-1
!
route-map MPLS-2 permit 20
match interface GigabitEthernet0/0.201
set metric 200
set metric-type type-1
!
route-map MPLS-2 permit 30
match interface GigabitEthernet0/0.202
set metric 200
set metric-type type-1

By this we can stear the traffic from our HQ going to the branch office.

Now one of our branch offices is growing rapidly and we can not get higher bandwidth on short term.

The most traffic coming from the HQ is pulled by the subinterface 0/0.201 and is filling up the bandwith and the users start to complain.

Now my idea was to split the /24 subnet into two /25 subnets and split all clients over both subnets.

But how can I tell the router to redistribute the connected subinterface just by half?

So just redistribute 10.90.20.0/25 into ospf1 and 10.90.20.128/25 into opsf2?

Any idea on that?

Maybe there is way to filter on HQ side?

thank you

[balaji.bandi]

Do you need to put your network diagram to understand how these are connected ?

yes possible if you can play with metric and cost - between the links and Route-map 

You provided very minimal information, with that config may be possible. but we do not totally visibility.

is the subnet is stretched across DC?

[hash2k2]

the red routing table is what I want.

 

[Georg Pauwen]

Hello,

since you have two links, what is the current utilization of each link ? And are both links the same speed ? You cannot increase the bandwidth, so the best you can do is make sure both links are used to capacity.

[hash2k2]

Both links have 4Mbit, the first one is used mosty 100%, the second less than 10%.

[Georg Pauwen]

Hello,

4Mbit is (obviously) not much. Since that is a temporary solution, why not simply configure two default routes at the branch office, so that traffic is automatically load balanced over both links ?

ip route 0.0.0.0 0.0.0.0 10.22.100.254
ip route 0.0.0.0 0.0.0.0 10.23.100.254

[Georg Pauwen]

Hello,

a basic PBR would probably be the better option (PBR takes precedence over any static or dynamic routing). You can split the subnet into two access lists, and then route the traffic as desired.

[MHM Cisco World]

I read such like this case months ago, 
try 
split the /24 into two subnet
config the ip route null0 for each split subnet and /24 subnet.
redistribute the static no connect. 
this will give you 
each SP will carry one split and /24 "/24 in case of failover". 
this my opinion.

[hash2k2]

Yesterday I splitted the /24 into two /25 by creating a new subinterface.

Than I added an additional route-map entry for both route-maps with metric being 90 in first and 200 in second.

Now the router in the HQ knows to routes, one via the ISP line1 and one via line2.

Perfect.

[vencislav.metev]

Hi,

I think you can add a static route:

ip route 10.60.20.128 255.255.255.128 Gi2/0 10.23.100.20

PS. I think you have a mistake in the routing table in red. Second line "10.60.20.128/25 [110/91] via 10.22.100.20, gi2/0" should be "10.60.20.128/25 [110/91] via 10.23.100.20, gi2/0"

Regards,

Ventsi