* Ondrej Zajicek
Well, your use case wasn't the one i have in mind for 'trigger' feature. I expected trigger to be like default route. But your use case should work with multiple radv instances.
Indeed it does! My bug was (I think) that I had a hyphen in the protocol instance name. At least it works now, using quotes. Thanks for letting me know it ought to have worked. :-) For future Googlers' reference, here's the config I'm converging on, which seems to work and gives an acceptable fail-over time (30s): protocol radv 'radv-vlan100' { export filter { # Only consider on-link routes, if gw is set we've learned # the route elsewhere (probably from the other firewall # via OSPF) and in that case the RAs should be suppressed if gw = :: then accept; reject; }; trigger 2001:db8:100::/64; rdnss { ns 2001:db8::1; lifetime 2419200; }; interface "vlan100" { max ra interval 5; default lifetime 30; }; } [repeat as necessary] It's a bit more expressive than I would have liked, but I can live with that. It would have been nice if instead of a trigger prefix I could say for example that the trigger is "pulled" if, and only if, there is a prefix to announce in a RA Prefix Information Option (this information must necessarily be available to the radvd protocol, as it by default will automatically include a corresponding PIO if there's a globally scoped address/prefix on the interface). Best regards, Tore Anderson