Sorry, what I was trying to say is that I would try putting both lines at the same preference level.
static {
route 0.0.0.0/0 {
next-hop 211.211.211.209;
qualified-next-hop 212.212.212.161 {
preference 20;
interface ge-0/0/2;
}
preference 20;
}
}
My issue on the netscreen device was fixed by this approach. The problem was the return on any traffic coming into the backup interface was going out the primary interface. This just didn't work for SNMP and ping traffic requests. When I put both primary and backup line at the same preference both interfaces worked normally.
The side effect will be the round robin your other outbound requests between these two interfaces. In my case it didn't matter since the traffic I wanted to tie to a specific interface was riding on a VPN tunnel. I could control what interface the VPN tunnel bound to and therefore control where that traffic went.
If you need to control what uses the primary interface you may have to create some policy based routing rules for that after you make the two preferences match.