LAGG - Failover with different switches

[KiralyCraft]

I have a machine with multiple interfaces, and two of them are connected between different switches. The switches are configured in a redundant way, but not quite the best (STP only). I have two links connected one in each switch, and they are configured in a LAGG with failover mode. However, due to the ports being actively up, I suspect this causes some issues in the STP scenario. Is there a way of turning off the non-active links such that the switches do not see it?

Or, can I have both of them active and simply "enable STP" on OpnSense as well?

[Monviech (Cedrik)]

Spanning tree is needed on layer 2 between switches. The OPNsense is a Router and should be connected to Edge ports that do not send or receive STP packets.

Its interesting that a LAGG between switches that are only connected via STP (so non stacked) even works. For these scenarios there is MLAG.

https://en.m.wikipedia.org/wiki/Multi-chassis_link_aggregation_group

[KiralyCraft]

Thank you for the quick response. I hoped that OPNsense can bond the interfaces in a virtual switch. Indeed MLAG would be the best solution, but unfortunately the switches are from different vendors (Dell and FS respectively) and MLAG is therefore not an option.

The ports of the switches are indeed not configured as edge ports, so I'll tackle this first. Nevertheless, I was under the impression that the "failover" LAGG works by actively refusing (or appearing as "dead") on the port that is not primary. The interface is indeed active, although I don't know whether it's accepting packets or not.

EDIT: It seems that putting the ports of the switches where OPNSense connects in edge mode solved the problem, apparently. Thank you for your help - I'll reach back if the problem continues, at least now I'm on the right track.

EDIT2: The problem was very silly. In the end, the problem was that I've migrated an old router to this one, and the old router was still active at the same IP address, and was poisoning the ARP cache of the switches. It now works. Strangely, only Windows and Apple clients were affected, while Android (and Linux) worked correctly.