Hello Stephen,
multiple OSPF routing processes are executed with a "ships in the night" model: both routing processes are in competition to propose IP prefixes to populate the IP routing table.
The most important aspect is that OSPF type of route is not compared between different routing processes.
Within a single OSPF process OSPF internal route are preferred over OSPF inter area route then are preferred OSPF exernal type 1 over OSPF external type 2.
This information is lost when the comparison is performed in the IP routing table by the IP routing table manager process.
So the risk is that an OSPF external route in process 2 that was originated by redistribution of OSPF process 1 into process 2 is selected and placed in the IP routing table instead of the original route ( that was internal to process 1)
The recommended best practice to have a deterministic behaviour is to use the distance OSPF command to differentiate the Administrative distance for the different type of OSPF routes in both routing processes building a clear hierarchy.
The clear ip route gives a chance for competition between the OSPF processes
see
http://www.cisco.com/en/US/tech/tk365/technologies_white_paper09186a0080531fd2.shtml#singlered
What you can do is something like this
router ospf 1
distance ospf external 120
router ospf 2
distance ospf external 130
Where ospf 1 is the core protocol (WAN side) and ospf 2 is the edge protocol.
This is specially useful when multiple points of mutual redistribution exist in the network.
With a configuration like this you get a deterministic behaviour with each OSPF process that is able to install its own internal routes and with a clear preference of OSPF process 1 external routes over OSPF 2 external routes.
the distance ospf command allows also to modify the cost of internal OSPF routes and of OSPF inter-area routes if needed.
Hope to help
Giuseppe
