ospf not load balance

[petercinvest]

in my router R1. I run show ip route 10.1.200.0

Routing entry for 10.1.200.0/24
Known via "ospf 65002", distance 110, metric 21, type intra area
Redistributing via bgp 65002
Advertised by bgp 65002
Last update from 10.1.10.2 on Ethernet0/3, 00:50:59 ago
Routing Descriptor Blocks:
* 10.1.12.2, from 10.1.0.4, 00:50:59 ago, via Ethernet0/2
Route metric is 21, traffic share count is 1
10.1.10.2, from 10.1.0.4, 00:50:59 ago, via Ethernet0/3
Route metric is 21, traffic share count is 1

I expect R1 -10.1.12.2-10.1.200.1 and R1-10.1.10.2-10.1.200.1, now R1 go to 10.1.200.1 only by 10.1.12.2, not load balancing.

I quite surprise because both 10.1.12.2 and 10.1.10.2 has metric 21, there is no ospf cost in 10.1.10.2 interface, so why it is not load balancing, thanks

[chrihussey]

By default the router will load share on a per destination basis. So if you don't have a large routing table with lots of destinations you may not always get exact load sharing. In instances where large data transfers occur an interface will show more utilization. Also, if there is generally a destination that is used a lot more and gets tied to an interface you will also see that interface used more.

If you want a truer more balanced load across the interfaces, and I assume CEF is running, you could try using the "ip load-sharing per-packet" interface command to see if this achieves a better balance.

[petercinvest]

My another question is why router choose 12.2 instead of 10.2 as the ospf cost metric is the same

[chrihussey]

It's actually using both and it really isn't choosing one over the other. The asterisk identifies the route it is currently using. If you repeatedly do the "show ip route 10.1.200.0" command you will see it move back and forth from route to route.

However, as stated, a lot will depend on the network routing table, amount of traffic to a particular destination and amount of varying traffic. 

[Joseph W. Doherty]

If you want a truer more balanced load across the interfaces, and I assume CEF is running, you could try using the "ip load-sharing per-packet" interface command to see if this achieves a better balance.

It often provides a very close to even balance, but a packets can now arrive at destination out-of-sequence, that may create other issues.  I.e. generally, it's best to avoid using this option