Hi all,
I'd like to check an implementation I made on a SOHO network to manage WAN failover using OSPF.
Based on attached diagram, R1 has 2 routing tables (main and table2) and I'm managing the default route through OSPF+BFD for fast switching between R2 and R3 (connected to internet) in case of failure.
I have an OSPF instance on R1 linked to the main routing table and for this I get the default route installed from R2 (main WAN) which switches over to R3 in case of fail; I'm using cost to prefer R2 over R3. For this route I'm working on the default backbone area 0.0.0.0 involving R1-R2-R3.
In order to manage the second routing table on R1 (where the default route is primarily based on R3 with R2 as backup/failover), I have installed on R1 a second OSPF instance tied to routing table2; on both R2 and R3 this second instance is tied to the main routing table (R2 and R3 have only the main routing table).
Then I have created a separate area1 (area id 1.1.1.1) on R1, R2 and R3 attached to this instance2 and I use this to install the default route on table2 in R1, based on costs.
Everything works fine and switching from R2 to R3 is quite fast based on BFD for both routing tables, main and table2.

As I'm not an expert with OSPF I was wondering whether defining 2 areas to manage the 2 routing tables on R1 is a normal choice or there is a better solution to achieve the same result.
Thanks in advance

My 2 cents: if it works for what you do and does what you want, then it is okay.

Now, do I think this is overengineered for a WAN failover? Absolutely. For a "simple" WAN monitoring, I would likely use the "check-gateway" option with a single router and two uplinks. If you got /29 from both ISPs, you may even have a failover router and use VRRP to protect against a device failure. But again, you may have other use for OSPF than "just" WAN failover.

Thanks for sharing your opinion on this.
Yes I agree with you that engaging OSPF is too much for just failover, but I wanted to improve my setup after some test.
Initially I had used the recursive routing approach with couple of target IPs, but time to time I have got some false positive results leading to switch over to the other WAN unnecessarily.
Also the switching time to get back to the previous WAN was longer than with OSPF+BFD; now I get almost immediate switch and the time clients get back internet is much faster.
I too conclude that the setup to feed a second routing table is ok with a second area, so after all I don't see issues with it (as it's working fine).
Without this second routing table needs everything would have been done only with a standard backbone area between the 3 routers.