Comware

Hi all,

Can someone explain to me how intervlan routing works in IRF with multiple switches? For example, with two layer 3 switches in a network you would home the VLAN on one of the switches (switch1), and then if you needed to use the VLAN on the other switch (switch2) you would stretch the VLAN over, but whenever you needed to travervse VLAns on switch1 you'd need to the packet to be sent back to switch1 to be routed to the new vlan..

What I'm getting at is I want to know if both switches are aware of routes in an IRF configuration? I'm trying to decide if havinng one 4 switch IRF, or two 2 switch IRF fabircs would be better for my environment. If having two seperate IRF fabrics would cause a router on a  stick scenario as above, then I'm thinking a single IRF fabric for the four switches would be best. What's everyone thoughts on this?

Switches in an IRF stack have a shared configuration, control plane and management interface, for most intent and purposes you can consider them as a single, bigger switch.

Can you point me to any documentation which specifically talks about how interVLAN routing is carried out within IRF? I can't seem to find any.

So what you're saying is when my switches are stacked using IRF, when I create a VLAN and a VLAN interface, that each switch is aware of that VLAN interface and the VLAN interface and IP address is live on both switches at the same time? Which means either switch can process intervlan requests by utilising its own VLAN interface.

Let's say for example I have two switches in an IRF. So what you're saying is if I have a packet come in from VLAN 1 destined for VLAN 2, either switch can process that packet, route to the correct interface, because both switches have the VLAN interface active on them at the same time?

Hi!

IMHO you have to stop thinking to deal with two switches...IRF means one logical switch (the IRF Virtual Device) so your two VLANs are created into (and managed by) the same virtual device.

IIRC Every member device in the IRF is capable of L2/L3 forwarding...this sentence rests on these other two:

1. The IRF Master runs, manages and maintains the IRF virtual device, whereas the IRF Slaves process services as well as function as the backups (through the mechanism of processes placement on global Active MPU and on global Standby MPU). As soon as the IRF Master fails, the IRF virtual device elects a new IRF Master immediately from one of the IRF Slaves to prevent service interruption and implement 1:N backup.
2. All of the IRF members have the same routing table and can route packets received from the edge switches connected to the IRF virtual device. The IRF master will run the routing protocol for the entire IRF virtual device.

Given the above, the point of inter-VLAN routing would became instead: suppose you have Host 1 member of VLAN 10 dual homed on both IRF Members and Host 2 member of VLAN 20 dual homed on both IRF Members...then what will be the flow of packets between these two hosts if the IRF is responsible of VLANs IPv4 routing? will the IPv4 traffic traverse (preferentially or not) the IRF link(s)?

The answer is: it depends.