Hi all,
I'm currently quite baffled with the problem I'm encountering. I hope you can help me figuring this one out. My setup is as follows:
I've got four nexus 7k's as core switches. Two of them per datacenter. Connected to them are 4 5k's, again two per datacenter. The 5k's are redundantly connected with the 7k's on the same location. The 7k's are connected with each other and their namesakes in the other datacenter.
Connected to the 5k's is a UCS on which some VM's run. One of these VM's is a TFTP server. The other VM's attempt to boot using this TFTP server. The curious case is that some of these are able to boot from TFTP and other are not.
The server is in vlan 650 and the clients in vlan 251. Routing is done on the 7k's which are running hsrp. The current active hsrp node is called 1001. Its neighbor 7k is called 1002.
When tracing the traffic we see the TFTP requests leaving the UCS and arriving on one of the two 5k's, after this the traffic is spotted on one of the two 7k's. On 1001 the routing is performed and so the traffic leaves vlan 251 and enters 650. We can see the traffic on vlan interface 650's egress 'port'. However, it never arrives at the server. As far as I can see (using wireshark) it never even arrives on the 5k on its way back to the UCS.
So the traffic flows as follows:
VM --> N1000v --> interconnect --> N5k --> N7k, vlan 251 --> N7k, vlan 650 --> ???
We expected some deviation in the traffic that does not work in comparison with the traffic that does work. However, none is observed.
The interface counters on all switches remain at zero when looking at errors.
My suspicion is that the traffic is sent over the OTV links between the datacenters, even though the mac address table is correctly filled on the 7k's
I hope you can give me some pointers as to where I might look to figure out this one.
Thanks in advance.
