OSPF Design Tips

[c.fuller]

Hello -

I have to break out my OSPF area configuration.   Right now one physical location is all one area.   The backbone routers are the ABR's (area0 of course) and the two distribution sets are both totally stubby and part of the same area1.   I want to keep one set of the distribution routers in the current area, but migrate the second distribution set to a new area and readdress things.  Is the best approach to stand up a new distribution set and then migrate of the old area1 distribution to the newly configured area2 distribution to include moving access closet uplinks, readdressing and changing vlans?   Or is it possible to configure the new area2 on the same distribution routers currently in area1 and migrate without the new distribution block.   I'd rather not stand up a new distribution pair, but am not confident on how I would do the conversion using the same distribution block.  Right now the distribution block's L3 uplinks to the backbone routers are in area1.   So Area0 does not extend down to the area1 distribution block.   If I created a new area2 on the area1 distribution block I am thinking I would need to extend area0 down to that distribution if I were to serve up two areas for the migration duration.   I've also considered a virtual link between the new area and area0 using area1 as the transit area.   If anyone has ever done anything like this before or has any insight/suggestions it would be appreciated.

Topology below.  Basically, want to reconfigure distb for a new area2.   Keeping dista in area1.  Best way to get er' done? 

            Backbone Area0

             |                 |

             |                 |

         CORE--------CORE            (abrs)

          |  |               |    |

          |  |               |    | --------> all are area1 L3 PTPs

          | dista     dista  |                (dista in area1)

          |                      |

        distb---------------distb             (distb in area1)

Thanks

Chuck

[lgijssel · Accepted solution]

Hi Chuck,

All non-zero area's need to border directly to the backbone, as you correctly assumed.

So you cannot create an area 2 behind area 1. Extension of the backbone area is the only viable solution.

In this case, the ABR functionality will move from the core to the dist layer. This may not be an improvement.

Factors to consider are:

- the size of the area's

- the performance of the hardware involved

Perhaps it is also possible to have a vlan trunk from core to dist and put area 1 and 2 in a different vlan?

The technically most fancy solution would of course be to have an extra pair of dist switches.

To provide a more detailed /accurate advice, we would need to look at the configs and have more info regarding the hardware used here.

regards,

Leo

[c.fuller]

Thank you Leo.  I appreciate your thoughts.  I am leaning toward extending area0 down to the distribution layer and having that serve as the ABR temporarily until we get everything converted to the new area's addressing.   Then pull area0 back to the core and readdress the uplinks and place in the new area2.  

What do you think about using a virtual link?   Basically, area2 would form an virtual link with area0, using area1 (which is directly connected to Area0 using L3 PTPs) as the transit area.   This would seem a fairly easy and quick interim solution during the migration phase as well.  

Thanks again.

Chuck

[lgijssel]

Actually I had written to consider using a virt link. Then I removed it again because I am actually never using them myself. I guess it is a matter of personal preference and mine is to keep my ospf as clean and lean as possible. But in these days of tightening funds, one may perhaps need to consider less conventional solutions as well.

For migration purposes it is probably ok, as long as the CPU's can keep up with the ospf processing.

As a definitive scenario, I would not recommend it.

Thnx for rating!

regards,

Leo

[c.fuller]

Leo, thank you.  I agree with your caveats about using a virtual link.   In doing more research I learned that virtual links can't pass through a stubby area.  Area1 in my topology is a totally stubby area.   So rather than get into changing the area to a non-stubby area and then adding a virtual link I'd rather just extend area0 to the distribution temporarily for the migration phase then pull it back.

Chuck