DHCP failing on Nexus with Server PXE Boot

[neilo]

Hi,

I have 2 x Cisco Nexus 9K 93180YC-EX units connected via vPC, plus a port-channel (also with vPC) back to our L3 Core (6509 VSS pair).  We have a new ESX Host trying to boot via PXE to get a DHCP address but it fails each time getting a 'No offers received' message each time.

From these switches we can ping the SVI at the Core acting as the gateway for the new ESX hosts VLAN and the Core SVI has IP help addresses configured and all works fine for other servers running across the other LAN switches.  The server switch port is now configured as an access port with the required VLAN.  We did have it set as a trunk port with a native VLAN set but I simplified the config because we're seeing these issues.

Is there a known method / requirements for getting DHCP to work across this kind of environment?  I've done this many times across a standard Catalyst setup and to be honest I assumed it would follow the same process with the Nexus units too.

Any ideas?

Many thanks

Neil

[paul driver]

Hello
I assume you have the correct vlans traversing the trunks with the correct native vlan specified, and these vlans are propergated into the vtp database of the switches?

Have you tried applying helper addressing be it on the core or the the nexus VPC (if applicable)

[neilo]

Hi Paul,

Thank you for your reply.

Yes, the correct VLANs are configured on the uplink to the core and the interface serving the Server has the correct VLAN.  The core is doing all the L3, we just have our mgmt SVI configured locally.  We can ping back to core on the mgmt and the server VLAN.  For testing purposes, I've tried the following:

• Changed edge interface from trunk to access, with just the server vlan configured - made no difference
• IP Helper already configured on Core SVI but I have now added a 3rd IP helper for the PXE Server - made no difference
• Added local SVI on the server VLAN - made no difference
• Switched off second switch, so we're running on a single switch - made no difference
• Removed vPC command from uplink port-channel back to core (eliminate vPC) - made no difference

Comms back to the core seems good but these servers are just not getting a response back.  Unless the Nexus are doing something strange with the DHCP broadcast, I can't see what else I can configure.

I did read a doc on DHCP config on the 9K's and 2 of the limitations were:

• Cisco Nexus 9000 Series switches do not support the relaying of bootp packets. However, the switches do support bootp packets that are Layer 2 switched.

• DHCP subnet broadcast is not supported.

Not sure if this will make a difference.

Thanks

Neil

[paul driver]

Hello

Trying disabling LACP within the PC if/when it doesnt receive any lacp packets this wont disable the port but puts it into a individual state thus may allow PXE to traverse through it.

int x/x
no lacp suspend-individual