Feature request - link fault propagation

I'd like to request a feature where link down on one interface can set another interface to be disabled.

Many possible use cases:

  • addiing backup to LHG60 station: w60g link down -> force ether1 link down, other end of the cable connected to a port which is part of active/passive bonding (like in default config of a Cube but between two wired ports, this way the two radios don't have to be part of the same device)
  • 10G fiber link from upstream IXP to my router, want to put a CRS305 in between to allow some monitoring/filtering or even port mirroring to a separate DDoS detection system, but still want the router to see link down from the IXP (fiber cut by an excavator, happens too often here) immediately without waiting for BGP to time out (faster switching to backup routes)

It should react quickly, so polling interface state in a script every minute is too slow, it should be event-based with a set of rules like: if sfp1 is down, disable sfp2. Many target ports can be disabled if one source link fails (like, two routers behind one switch can see if the link upstream of the switch is down).

I've read about a possible solution with netwatch, but especially in the IXP case I may not have any way to ping the remote IXP from the switch put in the middle (managed by separate management port which is not part of the bridge, the bridge just bridging VLANs between ports but no access to CPU port, the IXP terms of service don't allow any extra MAC addresses on that L2 link other than the router).

Have you considered using VRRP as a script trigger to monitor ups&downs of interfaces?

I need this to work at the physical layer - link down from one interface propagated to another. It's not a new idea, I've seen hardware media-converters with this feature 25+ years ago (back then, usually activated by DIP switches). Just being able to run scripts on physical link down/up events (including the initial down/up state after boot) should be good enough.

Searching the forum, I've seen this already requested a few times, since 2010 or so.

Maybe one of the other possible probes Netwatch has, besides ICMP?

Or checking the log?
Link down should be logged.