I'm getting a weird behaviour. When I run a VM, it gets the address of DHCP server (dnsmasq) as the default gateway !!!!
My install is as follow :
- two physical nodes, each with two NICs (one for management network, and the other for Data network (VMs) )
- OS : CentOS 6.4
- L2 plugin : Linuxbridge
- namespaces=False
- dhcp-agent is running on the controller node (all openstack services : nova, glance, cinder, quantum-server, ...)
- l3-agent is runnig on the compute node
Because of the misconfigured gateway, I can't access to my VMs using floating IPs, till I change the default gateway myself(I can access the VMs using their fixed IPs)
any idea what's wrong?
if you need any additional information, you can ask.
thanks in advance
yeah, i've seen this before.
My understand is that this happens on redhat/centos versions that have a version of dnsmasq that is older than what quantum is expecting. basically, the config we push to dnsmasq is not interpreted correctly, meaning dnsmasq does its default behavior of using its own IP as the gateway. i'll convert this to a bug, as we should either check the dnsmasq version and push the right config, or document the version requirements around dnsmasq better.