I don't quite understand this so maybe someone can clear it up for me (could be a bug?)
I have 2 routers
- CCR1016 running 6.37.5 (R1) - 10.0.0.1/28
- RB2011 running 6.39.3 (R2) - 10.0.0.2/28

They are connected over a point-to-multipoint fixed wireless link (R1 is the hub, R2 is a spoke) that has other routers but they are not yet participating in OSPF
The wireless AP has traffic isolation turned on so R2 cannot directly talk to R3, hence the need to use a non broadcast traffic type for OSPF

If I use broadcast, point to point or NBMA it all works as expected (currently only using 2 routers so ptp and broadcast won't be suitable later), but NBMA is also very slow to converge and update LSA's - maybe 15 seconds. I find I can't set the poll interval any lower than about 3s either or there's flapping routes (despite only having 14ms latency between them)

If I use point-to-multipoint in OSPF between these 2 routers then R1 stops advertising the 10.0.0.0/28 subnet and instead ONLY advertises the host routes (10.0.0.2). I have 10.0.0.0/28 in the networks tab. This means that I cannot reach i.e. 10.0.0.8 which is a radio in that subnet.
I expected to see both the network of 10.0.0.0/28 as well as the specific host routes, but 10.0.0.0/28 doesn't get advertised at all
Is this functioning by design or is this a bug in RouterOS?

HI, have you tried updating the RouterOS version ?

I am also facing the same behaviour and would like to know if there is any workaround. I saw that PTMP is recommended over wireless links as being more stable, but at the same time if I cannot manage the radios, then it becomes problematic.

I don't remember where this was in our topology so I can't go and find what I did to fix it
But I do know that I don't ever use the PTMP network type anymore, and I exclusively use point-to-point so i'm guessing this is what I did with that link as well
All multi-point links now use individual VLAN's to each SM radio/router and then use the point-to-point network type

i.e. create 1 VLAN interface per SM
VLAN101 = SM1
VLAN102 = SM2 etc
Individual VLAN interfaces must be created on the AP side router, but depending on your network you don't necessarily need to add the VLAN interface on the SM side router. If you only ever carry all traffic over 1 VLAN then set the 'data port/network/customer VLAN' on the radio instead of the router. So the router thinks its just using VLAN1/Native on its side, but when it gets to the SM it tags it as i.e. VLAN102

I've found point to point to be by far the most reliable OSPF network type for MikroTik across wireless networks

I don't remember where this was in our topology so I can't go and find what I did to fix it
But I do know that I don't ever use the PTMP network type anymore, and I exclusively use point-to-point so i'm guessing this is what I did with that link as well
All multi-point links now use individual VLAN's to each SM radio/router and then use the point-to-point network type

i.e. create 1 VLAN interface per SM
VLAN101 = SM1
VLAN102 = SM2 etc
Individual VLAN interfaces must be created on the AP side router, but depending on your network you don't necessarily need to add the VLAN interface on the SM side router. If you only ever carry all traffic over 1 VLAN then set the 'data port/network/customer VLAN' on the radio instead of the router. So the router thinks its just using VLAN1/Native on its side, but when it gets to the SM it tags it as i.e. VLAN102

I've found point to point to be by far the most reliable OSPF network type for MikroTik across wireless networks

Aparently OPs issue is solved in some cases by changing to a p2p network..

HI, have you tried updating the RouterOS version ?

This is still an issue near the end of 2019.

Are the interfaces in a bridge, or is there any reason arp might get confused with changing Mac addresses?

I find that if I turn on horizon on the interfaces, ptmp works great.

In a static environment I just give in an run a vlan and do PTP mode over the vlan. Seems to be the simplest solution.

I do have a test hybrid mesh setup going with horizon bridging on WDS and a /24 in ospf and ptmp on all the interfaces and I don’t see this issue in their unique arrangement. I’m assuming the horizon setting is again saving the day.

I would recommend using individual VLAN interfaces to each router anyway for the reason you can then apply CoS priority based on DSCP high 3 bits. As many radios will not read DSCP tags. But applying CoS at every hop in the network ensures you have working QoS everywhere and the only way to do that is with VLANs. I don't really see any point in multipoint setup unless you really, really need to save IP addresses for some reason

Hello everyone, actually I've never seen that issue in 6.44.5 long-term again, I agree that the best solution is use point-to-point interfaces specially between routers, I'll use ptmp interfaces only at special occasions where there are lots of routers who try to establish adjacencies between them.