DHCP failing on VMs

[jrb_proxmox]

I am experimenting with bringing up a VM server cluster. I have just two servers in it so far, and am having trouble getting DHCP working on the VMs.

The VM network has two bonded 10 Gb NICs, connected to the IT network. On a Windows-based test VM, if its IP and DNS servers are set statically, it can connect to the internet with no problems. However, if set to DHCP, it doesn't connect. More specifically, ipconfig /renew hangs for about 2 minutes, and ultimately returns this message:

An error occurred while renewing interface Ethernet Instance 0 : unable to contact your DHCP server. Request has timed out.

My interfaces file for the host with the problematic VM is set quite similarly to the very last example on this page (right above the disabling IPv6 section).

If I run this on the host:

tcpdump -vv -i bond1 udp port 67

I see DHCP requests coming from the Windows-based VM.

I'm not clear if this kind of problem could be due to something I'm missing in the interfaces file on my end, or represents something that needs to be corrected by my IT department. I'm working in coordination with them, but finding the people with the expertise to figure this out is proving challenging. I'm hoping to find someone who might be doing something similar and has already solved this. Thanks in advance!

 

[ph0x]

Do you NAT on the host or do you have a bridge? Any firewall active for the VM?

 

[jrb_proxmox]

Thanks for the reply ph0x! I am attempting to configure a bridge, much like that referenced example. It works for the VM accessing the network with a static IP, but not with DHCP. I want it to use the DHCP server available on the IT network. The VM has the firewall disabled. I could share all or part of my /etc/network/interfaces file if that would help and you'd be willing to take a look. Again though, I'm not sure if this might be beyond my control and something my IT folks need to enable.

 

[ph0x]

If the network works per se, there is nothing in /etc/network/interfaces of relevance.
I would rather guess that your IT department has configured the DHCP server to only talk to known hosts. So you should definitely talk to them first to rule out any configuration issue that is outside your influence.

 

[jrb_proxmox]

I think we got to the bottom of this. TL;DR is, it was indeed IT-side. In case anyone else runs into something similar, here are some terms that might help facilitate dialog with IT, or if you're managing this yourself, so you know what to look for.

Our VLAN's DHCP scope was configured for a "new" DHCP server. That new DHCP server's IP wasn't in the switch's "IP Helpers". We added it there, and things still didn't work. We then questioned if that DHCP server's routes were properly configured. We switched the VLAN's DHCP scope to a DHCP server that was more established and known working. Initially this also seemed to not help, but by morning the next day DHCP on the VMs worked like a champ. We believe there was a replication delay of some kind with the DHCP scope change, such that when I tested it shortly after it had been made, it just hadn't fully applied yet.

 

[ph0x]

Thanks for the feedback. Network helpers are indeed a pita source of problems.