provider network vlan external dhcp

asked 2017-11-14 08:45:37 -0600

stackopen gravatar image

updated 2017-11-22 13:09:34 -0600

rbowen gravatar image

Hi all,

I come from a vSphere environment with DVS. I tested an openstack deployment using RDO packstack and was able to create a network setup similar to vSphere but now I'm stuck because we already have a Windows DHCP server on our provider network and I don't want the dnsmasq to interfere with it.

I have deployed a 4 node (1 controller + 3 compute) setup. On all nodes 2 10Gb nics are bonded to bond0 which is then mapped to br-ex. For testing I created a network and a subnet with DHCP enabled and allocated a range of free IPs that is excluded from the existing Windows DHCP. Using this I can deploy instances that can reach machines on the same vlan outside of openstack just fine by adjusting the security group.

I thought this dnsmasq DHCP is somehow isolated so that only instances can reach it but in /var/log/messages I can see that all sorts of machines on the same provider vlan reach it to:

Nov 14 15:32:53 gedaspl03 dnsmasq-dhcp[6116]: DHCPREQUEST(tapf01d75bb-1e) 172.28.19.178 00:50:56:b7:38:81
Nov 14 15:32:53 gedaspl03 dnsmasq-dhcp[6116]: DHCPNAK(tapf01d75bb-1e) 172.28.19.178 00:50:56:b7:38:81 lease not found

So I stopped my test for now as I don't think it is good to have more then one DHCP. :P I'm now wondering If a setup similar to vSphere is even possible?

My question is, can I simply ignore the messages regarding dnsmasq because it is really somehow isolated and only instances will get IPs or do I have to use tenant networks with routers and floating IPs?

Best Regards, Oliver

edit retag flag offensive close merge delete