Good morning everyone.

I'm experiencing a ospf issue i have never seen before.

We have a 1072 that was running a fairly old ROS running solid for the last past years.

There are 3 10G links to the datacentre and I use OSPF with BFD to peer to the loopback of the core router in the DC.

Now the issue I am facing is, if any of the 3 links to the DC goes down for what ever reason and the ospf link state changes All the BGP sessions are dropped even on interfaces not running OSPF.

It seems like what ever is happening is causing OSPF and BGP to reset completely resulting in all BGP sessions to be dropped.

I have updated the box and its firmware to 6.49.10 and the problem still persists.

The log is showing a ospf issue for all the interfaces that is running ospf " Discarding packet: no neighbor with this source address" BGP debug is impossible to read since it fills the log every time it happens.

Also the interfaces going to the DC is setup as ospf Point to Point with BFD enabled on all of them. No firewall, NAT or connection tracking on this device

Any suggestions on what may be going on here?

The very first small step ... try to catch logs with external syslog to spot what's happening.

The very first small step ... try to catch logs with external syslog to spot what's happening.

After sending ospf and bgp to a remote logging server I noticed something(no link has dropped yet, I need to manually simulate a failure to test)
some of the interfaces running ospf as point to point is sending multicast hello packets.

As I understand it ospf interfaces in point to point mode should only send hello packets to to its directly connected neighbour on the same subnet.

But still dropping all bgp sessions even on interfaces that is not running ospf is surprising.

Its seems like any dynamic route, ospf, bgp is affected when this happen.

I have created a support.rif file and will send to mikrotik for further investigation

Does BGP/OSPF propagate OSPF's/BGP's links routes somehow?
Recalculation of OSPF or BGP possibly breaks some connections -> neighbours change -> best paths change -> OSPF/BGP recalculates -> neighbours change -> best paths change -> .......

The ospf only propagates loopback addresses and the point to point addresses between routers.
BGP is handling our public /24s and announces a default route

My issue is quite similar to this as well. Let me know what mikrotik support says. In my circumstances when an ospf neighbor changes, that router will drop ALL ospf and BGP connections. In our case BGP is used for loopback, point to point addresses and routes within the network. BGP is handing default routes from route reflector and announces public IPs to upstream provider

So I may have found the culprit here.

Static blackhole routes with distance 1 to prevent routing loops.

Because the is public addresses on the ospf interfaces and I'm blackholing /24's. As soon as there is a link state change for a split second(i assume all the routes gets withdrawn) then it looks like the router ignores all dynamically applied routes and the BGP session gets blackholed and drops all sessions. I have moved the ifaces to private IP's and disabled the blackhole routes and tested by manually flapping a link and all sessions stays up.\

I will monitor in the coming days if the issue pops up again(one of our layer 2 providers is currently experiencing issues) so its a good test to see if the changes has solved the issue

@p3rad0x

So are you saying you removed those routes and then are using private IPs for your ospf interfaces? Not totally understanding what the solution was there, however it sounds somewhat similar to what i have been experiencing.