I have the following setup in which I previously had problems with OSPF dropping in PTP or PTMP mode so I switch to NBMA for stability as these are 20+km wireless links (mixture of cambium and UBNT radios)
ROS Versions were 6.38-6.39
Traffic should flow over the primary link at all times unless the link goes down
R1 eth2 connected to netonix switch
VLAN10 -Primary link & VLAN11 Backup link on separate /30 subnets and both added as NBMA interfaces in OSPF setup (VLAN10 has lower interface cost than VLAN11 in OSPF)
VLAN10 ip = 10.1.1.1,
VLAN11 ip = 10.1.1.5
On switch the primary radio link is untagged on VLAN10 and seconary radio link untagged on VLAN11
On the remote end
R2 eth1 connected to Primary radio Link, Eth2 connected to Backup radio link.
Eth1 and Eth2 added as NMBA interface in OSPF setup. (Eth1 has lower interface cost than eth2 in OSPF)
Eth1 ip = 10.1.1.2
Eth2 ip = 10.1.1.6
R1 NBMA Neigbours
10.1.1.2 Prioriy = 1
10.1.1.6 Priority = 0
R2 NBMA Neighbours
10.1.1.1 Priority =1
10.1.1.5 Priority = 0
This setup seems to work but does not fail-over automatically, often requires manual restart of OSPF routing to make routes come back in the event of one of the routers rebooting.
In my test environment running 6.40.1 setting all Priorities to 1 seems to enable correct fail-over, however on 6.38 running this configuration in production caused the traffic to sometimes prefer routes from the backup path instead of the primary. I do not know why this is as the interface cost of the backup path is higher.
In essence using PTP or PTMP mode this functioned correctly and as expected using interface costs but I was forced to use NBMA due to instability over time. (Random routes dropping once per month)
What should my NBMA priorities be in this situation as I am unable to find a definitive answer.