Following official documentation, I'm trying to deploy a Devstack on an Ubuntu 18.04 Server OS on a virtual machine. The devstack node has only one network card (ens160) connected to a network with the following CIDR 10.20.30.40/24. I need my instances accessible publicly on this network (from 10.20.30.240 to 10.20.30.250). So again the following the official floating-IP documentation I managed to form this local.conf file:
[[local|localrc]]
ADMIN_PASSWORD=secret
DATABASE_PASSWORD=$ADMIN_PASSWORD
RABBIT_PASSWORD=$ADMIN_PASSWORD
SERVICE_PASSWORD=$ADMIN_PASSWORD
PUBLIC_INTERFACE=ens160
HOST_IP=10.20.30.40
FLOATING_RANGE=10.20.30.40/24
PUBLIC_NETWORK_GATEWAY=10.20.30.1
Q_FLOATING_ALLOCATION_POOL=start=10.20.30.240,end=10.20.30.250
This would lead to form a br-ex with the global IP address 10.20.30.40 and secondary IP address 10.20.30.1 (The gateway already exists on the network; isn't PUBLIC_NETWORK_GATEWAY parameter talking about real gateway on the network?)
Now, after a successful deployment, disabling ufw (according to this), creating a cirros instance with proper security group for ping and ssh and attaching a floating-IP, I only can access my instance on my devstack node, not on the whole network! Also from within the cirros instance, I cannot access the outside world (even though I can access the outside world from the devstack node)
Afterwards, watching this video, I modified the local.conf file like this:
[[local|localrc]]
ADMIN_PASSWORD=secret
DATABASE_PASSWORD=$ADMIN_PASSWORD
RABBIT_PASSWORD=$ADMIN_PASSWORD
SERVICE_PASSWORD=$ADMIN_PASSWORD
FLAT_INTERFACE=ens160
HOST_IP=10.20.30.40
FLOATING_RANGE=10.20.30.240/28
After a successful deployment and instance setup, I still can access my instance only on devstack node and not from the outside! But the good news is that I can access the outside world from within the cirros instance.
Any help would be appreciated!
Update
On the second configuration, checking packets on tcpdump while pinging the instance floating-IP, I observed that the who-has broadcast packet for the floating-IP of the instance reaches the devstack node from the network router; however no is-at reply is generated and thus ICMP packets are not routed to the devstack node and the instance.
So, with some tricks I created the response and everything works fine afterwards; but certainly this isn't solution and I imagine that the devstack should work out of the box without any tweaking and probably this is because of a misconfiguration of devstack.
After 5 days of tests, research and lecture, I found this: Openstack VM is not accessible on LAN
Enter the following commands on devstack node:
echo 1 > /proc/sys/net/ipv4/conf/ens160/proxy_arp
iptables -t nat -A POSTROUTING -o ens160 -j MASQUERADE
Cheers!