Hi all.
I've just started studying again after a long break, and as usual I've gotten myself a little confused on something I'd consider fine detail, but also pretty basic. I'm a bit embarrassed!
Unfortunately I don't have any hardware to lab this myself and I'm having issues getting GNS3 working, so I thought I'd run it past you fine folks and see what your opinions are.
I drew a diagram to help explain this:
A PC (on VLAN 10, IP address 10.10.10.50 /24) is connected to a layer 2 switch, ASW1. ASW1 has a trunk to a layer 3 switch, DSW1, which allows VLAN 10 over it. VLAN 10 is present in the VLAN databases of both switches. DSW1 has a standard, routed port off to R1 which sits in some random mystical cloud.
Question time:
Assume that DSW1 has, for whatever reason, no entries in its CAM or ARP tables for now. Say a packet comes down the routed port from R1 and reaches DSW1. It's destination IP is the host (10.10.10.50). DSW1 does a RIB lookup for 10.10.10.50 and knows it is directly connected via the SVI interface Vlan 10.
To forward the packet from here, DSW1 would theoretically need to ARP for 10.10.10.50 in order to build a frame with the correct destination MAC address. That frame would be sent over the trunk to ASW1. Now, the question:
With the assumption that DSW1 has nothing in its CAM/ARP tables for this host, would it tag the initial ARP request as vlan 10
based on the knowledge that 10.10.10.50 /24 exists via interface vlan 10?
That is how I would believe it to work, but I'm having trouble actually testing it. Is this correct, or does it do something else in order to correctly tag the initial broadcast? On a router-on-a-stick setup, you'd have subinterfaces with encapsulation dot1q blah commands, which makes more sense in terms of how it knows which vlan to tag it as.
I got to this point and realised how much I've written over something which probably isn't that complicated, but I get lost in the details sometimes. Sorry. I'd really appreciate any discussions on this!
