1

Context

I am currently setting up a little cluster. For now, I only have one node configured and running (CentOS 7) : a management node. All the other nodes (compute nodes) aren't configured yet.

enter image description here

The compute nodes run on RedHat 6, and I am trying to use Cobbler on my management node to provision them with Centos 7.

I have installed Cobbler and configured it to PXE-boot one of my compute nodes and install Centos 7 on it.

The management node is on subnet 10.10.3.64/26, on VLAN n° 1301. The compute nodes' have to be provisionned with ip adresses in that subnet too.

Cobbler configuration

I have cobbler configured to manage dns & dhcp. I have created a system object representing the node I want to provision.

/etc/cobbler/modules.conf

[dhcp]
module = manage_isc

(whole file here)

/etc/cobbler/settings (relevant settings / settings I have changed)

manage_dhcp: 1
manage_dns: 1
bind_master: 10.10.3.65
manage_forward_zones: ['linuxrt']
manage_reverse_zones: ['10.10.3']
next_server: 10.10.3.65
restart_dns: 1
restart_dhcp: 1
server: 10.10.3.65

(whole file here)

(can't post more than 2 links, other configuration files linked in commentary)

Problem

When I'm booting the compute node on PXE, it starts looking for the DHCP server, and then I get that error :

PXE-E51: No DHCP or proxyDHCP offers were received.

No dhcp logs are generated in /var/log/messages. I also tried to sniff for incoming packages on the management node with tcpdump, but I see nothing coming from the compute-node. It seems like the node's dhcp queries don't reach the management node.

  • cobbler system – Elouan Keryell-Even Jun 4 '15 at 14:05
  • is the management node in the same VLAN as the clients? – Teun Vink Jun 4 '15 at 14:30
  • I haven't touched the client nodes, so they might have an old network configuration. Hence, likely not the same VLAN/subnet as my management node. But the whole point of dhcp is to configure the network interfaces of those client nodes, no? – Elouan Keryell-Even Jun 4 '15 at 14:35
  • The DHCP protocol by default relies on the server and clients being in the same broadcast domain. If they're not, DHCP requests won't reach the server unless you define a DHCP helper address. – Teun Vink Jun 4 '15 at 14:41
2

Make sure the switchports of the DHCP server and the clients are in the same VLAN, or use a DHCP helper.

The DHCP protocol by default relies on the server and clients being in the same broadcast domain. If they're not, DHCP requests won't reach the server unless you define a DHCP helper address.

  • 1
    It was a VLAN problem indeed. The node to be installed was connected on a switch port where trunking was activated, so it needed to tags its data with the good VLAN number, which it was not doing, so it was unable to reach the dhcp server. Thanks mate. – Elouan Keryell-Even Jun 5 '15 at 8:21
  • I just connected the node to-be-installed on a switch port where there was no trunking, and that solved the issue – Elouan Keryell-Even Jun 5 '15 at 10:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.