The topology attached and explanation depicts the situation I have in production and I would like to confirm the following:
1) I see I can control traffic exit path via R2 just fine by increasing the OSPF metric. Then in case R2 is unavailable, R3 takes over as intended. I am not even touching BGP attributes in order to pick the exit path in this case. Does that make sense? I heard before that I should refrain from manipulating the OSPF metrics and use BGP (such as local-pref) to do this, but I am not sure if have an option in my case and I would like to confirm and it is acceptable to manipulate OSPF metrics to accomplish this instead of using BGP attributes. I have about 20 routers running OSPF already (and no iBGP) so if I can avoid doing iBGP internally that is better.
2) Given that I am doing redistribution from BGP-> OSPF internally, do I need or is there any benefit if I establish iBGP neighbor session between the two internal border routers R2<->R3?