Introducing OSPF into a static environment

[tommygunnah]

Hello everyone,

Im in a situation where I would like to present OSPF into my environment. Currently we have layer 3 switches and routers all over the place that have static routes pointing back and forth to each other which is creating slowness and routing loops. What is the best approach with little downtime to achieve this? I know some people have mentioned that I can bring up the OSPF process on all the routers and layer 3 switches and then build the neighborships without even touching the routes. Then once thats done I can move the static routes in one by one?

Thanks for all your input!

[Francesco Molino · Accepted solution]

Hi

If your static routes are correctly setup, I mean the forwarding traffic is well defined, you can configure your ospf, build your adjacencies and traffic won't be impacted as static routes will be preferred over ospf.
You can then have the possibility to validate that ospf is advertising correctly your subnet and the routing path is correct. As soon as you validate everything, you can start removing static routes and impact will be very low. When you remove a static route, the traffic will follow the ospf path. You can experience micro cuts when removing a static routes if you validated that ospf routing is aligned with what you have today in terms of routing

[Joseph W. Doherty]

I just want to re-emphasis what Francesco noted, that a smooth cut over assumes "static routers are correctly setup" and "ospf routing is aligned with what you have today in terms of routing".

As Francesco also already noted, normally statics would have preference over OSPF routes, so as you remove them, an OPSF route should take over. However, until all the static routes are removed, if the static routes don't agree with the OSPF routes, you might create a black hole.

For example, given R1<>R2<>R3<>R4<>R1. Statics could have R1 to R3 via R2, but OSPF might have R1 to R3 via R4 (or the converse). Depending what static you remove, OSPF might point to a next hop which still has a static point back to that device. If so, you create a routing loop, until the original static is reinserted or other device statics are removed.

To preclude the prior, you should insure your statics and OSPF will select the same path.

Lastly, as also noted by Francesco, as you remove statics, the route table needs to be updated. It's possible (?), for a moment, the static will removed and the OSPF route not yet placed in the route table. If this happens, for that moment, the router may not know how to route to some destination network, but such a moment should be very, very brief (assuming the OSPF topology has already been computed). If this is a concern, and you want to insure not even one packet is lost during the conversion, what you might try, is making the AD of the static the same, or higher, than the OSPF AD before you remove the static.