Hi!
Is there a way to build a layer 2 redundant ring with routeros? This is often used with industrial equipment and switches where fast fail over is needed and (r)stp can't be used. I have used hirschman switches with their Hiper ring protocol in the past which basically send a test telegram around in the ring and if the telegram suddenly is missing on the other ring port than it was sent from, it activates switching on both ports but if the telegram is instead received on the blocking ring port it will continue to block other traffic on it and stops it from creating a loop.
Many industrial devices like PLCs, remote I/Os, variable speed drives and so on comes with dual switched ethernet ports for this but you need a "master" in the ring to handle the redundancy so a complete loop isn't formed.
Would be cool if this could be done with routeros :)

It is called ERPS, Ethernet Ring Protection Switching.

As far as I know Mikrotik does not support it "yet", will be cool though

where fast fail over is needed

How fast is fast?

where fast fail over is needed

How fast is fast?

With ERPS, they aiming at 50ms

There is also MRP
https://en.wikipedia.org/wiki/Media_Redundancy_Protocol
Which should deal with up to 14 switches at 10ms (and more switches at higer switch over times)

I thought maybe it could be done with a looping script? One that sends a test packet and listens + count the timeout and if packet is received it triggers sending another packet. And maybe this could be done using a vlan or L2 fw rules so the "blocking" end can hear the test packet while all other traffic is blocked, just some quick ideas :)

G.8032 is definitely the way to go....would love to see this in the CRS3xx series.

G.8032 is definitely the way to go....would love to see this in the CRS3xx series.

Not sure that is meant for industrial use where all the dual port devices functions as "dumb" switches? I think G.8032 will require this support on all nodes while MRP could work on only the "ring master" switch but not 100% sure of this. As I said, I used hirschman hiper ring before and it works like that and MRP is a IEC standardisation and continued development of the hiper ring technology so I would guess it could work like I think it would with all other switches "dumb".

Hi!
I'm thinking of trying to script this using pings. Starting a ping every few ms using a script is not hard but and then how to catch it on the other end in the best way? I'm thinking of using L2 mac ping and specify which port the packets goes out on and ping the second ports mac address. And then use bridge firewall rules to capture the ping, as long as pings reach the second port (backup port) the port is member of some dummy bridge. If mac-ping packets stop coming in on that interface, the script switch the port to be part of the main bridge instead. And if pings start to make it through again port is moved back to dummy bridge. My biggest question right now is how to make the bridge firewall rules to execute scripts depending on if the mac-pings are coming in or not, any hints are welcome :)

Ya might want to take a look at spanning-tree and see what it can do. Spanning-tree will normally handle redundant L2 connections. In a ring enviornment, it is possible to have 1/2 the traffic go clock-wise and the other half go counter-clock-wise - and if the ring is broken then both directions will be used at the same time up to the break in the ring so that everything still talks.

Ya might want to take a look at spanning-tree and see what it can do. Spanning-tree will normally handle redundant L2 connections. In a ring enviornment, it is possible to have 1/2 the traffic go clock-wise and the other half go counter-clock-wise - and if the ring is broken then both directions will be used at the same time up to the break in the ring so that everything still talks.

Several problems with spanning tree, it is slow, I'm looking for like 50ms fail over to not cause communication problems in industrial system which uses near realtime communication.
And the ring is made of several two port devices which are not (r)stp compatible. If you build a ring using that my understanding is that either all devices has to understand (r)stp or only one can do it (main switch). If the ring has two or more (r)stp switches and dumb devices between them the "loop" will not be noticed.
So no, any spanning tree is not an option here.

If you are looking at sub 50ms, I doubt very much you will achieve this using scripts

If you are looking at sub 50ms, I doubt very much you will achieve this using scripts

You can ping with 1ms resolution so that part is fine I guess. And the receive monitoring script would have to loop since you can't execute a new script instance faster than 1s. But I still believe it is possible using a looping script. Question is how to "catch" the ping...
Another possibility could be using a single script if it was possible to evaluate if a ping was successful or not...

found this script which does about what I want. Seems ping will return number of successful pings as the return value? Manual for ping does not state this but maybe it is documented somewhere else...
viewtopic.php?t=125759#p619531

Some success :)
Failover to ether2 works when both consecutive pings fail but when ring is completed again it is not detected :(
There seems to be a problem when mac pinging ether2 from ether1 like this when they belongs to the same bridge, ping just don't go through then :(

G.8032v2 was on the roadmap. Maybe the Mikrotik guys can provide a status update ?

Seems like it could be the loop that is causing it, I had a slight idea the loop would cause problems but I thought if the switching back was fast enough the packet storm would be so fast it would not be noticeable. But it seems to completely block all traffic the exact same ms the loop is formed :(

G.8032v2 was on the roadmap. Maybe the Mikrotik guys can provide a status update ?

That would not help here anyway since it would be required on all devices in a ring.