my peer1 and peer2 switches are both CRS326-24S+2Q+ and my final scope is to have 2 groups of ports in HA on the same switch that will not comunicate among them. One group is for a proxmox ceph cluster and the other for proxmox lan, so I don’t want to mix those two kinds of traffic.
I know how to do this with vlans without MLAG but I don’t how to do this with MLAG on top.
/interface bridge vlan
add bridge=inter-chassis-bridge tagged=(all group one ports plus ICCP port) vlan-ids=100
/interface bridge vlan
add bridge=inter-chassis-bridge tagged=(all group two ports plus ICCP port) vlan-ids=120
but I can still ping from group one ports to group two ports and viceversa, furthermore I didn’t applied any kind of vlan on client side, I just made the bond interface, so i would expect to loose connectivity on client side but it still works.
I know that I’m missing some basics here but I cannot find something similar around to have a starting point..
Are you sure the traffic isn’t going to layer 3 from a gateway that the two port groups share? did you leave the default vlan 1 on all ports? the configuration is correct as the ports of both switches will only communicate on the tagged vlan. try posting your switch settings
Leave the PVID on 1 and set admit all or admin only VLAN Tagged on Frame Types, in this way only packets with the VLAN TAG you set on /interface/bridge/vlan will pass through the port
Dear I tested everything now is working, your suggestions were very usefull, problem was this
Leave the PVID on 1 and set admit all or admin only VLAN Tagged on Frame Types
and the two groups are now isolated beetween them. So the from the original MLAG tutorial the more important part is to add tagged VLANs in /interface bridge and add the same VLAN on top of the bonding interface for the client side.
Just one question, will this kind off setup take advantage of hardware offload since I have switch chips in my CRS? Is there a way to test hardware offload?
Notable exceptions of this, ccr2**116 and ccr22**16, the bigger models on the current product line
so i think is a very relevant thing in routers too, in fact today i use bridge vlan filtering model in most routers even without integrated switch chip
i have embraced bridge vlan filtering as my defacto standard in configuration, in that way if i have to migrate from for example ccr1036 to ccr2116 i dont have to almost any changes