IPv6 BGP default route preferring local instead of global route

[FortiCommunityUser67]

Hi,

Running a FortiGate on 7.2.5. We have BGP running over 2 seperate IPv6 interfaces connected to an ISP router over 2 VLANs to receive a default route; ISP dictates which link has preference. The BGP link is up and routes are received.  In the routing monitoring, I see both a next-hop-local and a next-hop-global. So far so good. But in the actual routing table, checking with get router info6 routing-table bgp/database, it sets the next-hop to the next hop local address, causing IPv6 routing not to work as that's a link-local address that's not usable for us.

I've been scouring manuals and options to prefer the global nexthop, but I can't seem to find anything. None of the routemap options that I find seem to do the trick. Closest I found is using a set-ip6-nexthop-local on the routemap, but according to the documentation that just refused the entire route in BGP. As both the linklocal and global are advertised in the same packet, that doesn't fix it either. The setup works perfectly in IPv4 because there is no linklocal/global concept there; there's just your default route.

Does anyone have any experience with this?

[glover33]

The advantage of getting a default from your providers is that it cuts down on memory use in your router because you are not holding full internet table. The advantage of receiving full table from your providers is that you can select the best path for each prefix via the most appropriate ISP. Remember that most specific route trumps all other routing decisions so the isp sending you longer prefixes (more specific routes) will likely get the majority of your traffic as it will be the most specific route and the provider sending you default probably isn’t getting much of the traffic. That’s fine if that’s your intention. Have a think about how you want to split your traffic over your different ISPs and feel free to post a follow up and I can give you some pointers on how best to engineer that.