When deploying a baremetal node with Pike, neutron DHCP agent provides an incorrect IP address. After "openstack server create", I can found two ports are created, one is on provisioning network, and the other one is on tenant network. When I "cat /var/lib/neutron/dhcp/network_UUID/leases", I find sometimes the tenant work has the DHCP entry rather than the provisioning network. This cause the node get an IP address in tenant network and cannot reach the tftp server. I am using ipmi driver and the network interface is neutron. The provisioning network is flat while the tenant network is vxlan. This configuration sometimes works fine (when the provisioning network acts DHCP) but not stable. I believe the process should be provisioning network during deployment and tenant network after then. Could someone please give me some advice? Thanks in advance.