Steps to recreate the problem:

1. Create Network

2. Create instances in such a way there are instances created in control node as well as compute node

3. ping the dhcp IP address from the instances

local.conf :https://gist.github.com/gvranganvtn/ccbf76d20d18b6c670c7
(control
(compute)

ovs-vsctl show in both nodes: https://gist.github.com/gvranganvtn/f4cccecc76bb27195b06

dump-flows in both nodes:
https://gist.github.com/gvranganvtn/0d8dd22ba19491b39c61

Test:
ubuntu@odl-os-control-node:~/devstack$ sudo ip netns exec qdhcp-6b6f54de-f04a-4dd6-8baf-7725a51df8fa ping 1.2.3.6
PING 1.2.3.6 (1.2.3.6) 56(84) bytes of data.
64 bytes from 1.2.3.6: icmp_seq=1 ttl=64 time=5.63 ms
^C
— 1.2.3.6 ping statistics —
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 5.631/5.631/5.631/0.000 ms
ubuntu@odl-os-control-node:~/devstack$ sudo ip netns exec qdhcp-6b6f54de-f04a-4dd6-8baf-7725a51df8fa ping 1.2.3.5
PING 1.2.3.5 (1.2.3.5) 56(84) bytes of data.
^C
— 1.2.3.5 ping statistics —
1 packets transmitted, 0 received, 100% packet loss, time 0ms

1.2.3.5 is the instacne created in compute node and 1.2.3.6 is the one on the control node.