What is the best way to organize the fault tolerance of a linux server with two switches?

Question

linux

There are servers, each has 4 1GE network cards. The servers have Linux of the current version installed, a bridge is organized where LXC containers live. So far, all network ports are aggregated in bond LACP and connected to one switch. I would like to achieve a fault tolerance scheme with two switches. Switches are Cat2960G and they are very nervous about "jumping" poppies, so balance-alb does not really work (if one of the cords is disconnected, the server becomes completely unavailable for an indefinite time).
How to properly organize the inclusion of each server in two switches, so that if one of them disappears, everything continues to work? Ideally, when both are working, balance. Since the bond is essentially a bridge, if not LACP, then hypothetically you can catch a loop. To organize fault tolerance at the L3 level (quagga/bird OSPF)?

Answer 1

active-backup than does not suit?

Answer 2

In your case, I would opt for active-backup from STP. You can also do it with quagga, but there is a lot of trouble if there are a lot of IPs on the server, you will often have to redo the config on quagga, you can set up redistribution, but if there is a case where you don’t need to redistribute all IPs, but you only need specific ones, etc. Or on one physical server there will be many virtual machines with other subnets, or they will often change or be added, then you will often have to fiddle with the quaga. If the traffic is not more than 1G, then it is better to make an active-backup. In ideal, it would be necessary to put switches that can work in a stack and already pull aggregated channels from the stack to the server.