Install OpenStack

1. Review these documents for instructions on installing OpenStack on OVM:

   http://www.oracle.com/technetwork/server-storage/openstack/linux/documentation/oracle-openstack-user-guide-2296176.pdf

   http://www.oracle.com/technetwork/server-storage/openstack/linux/documentation/openstack-hol-virtualbox-2326937.pdf

2. If you are installing Oracle Fabric Manager on an Oracle VM controller, make sure the controller has an appropriate amount of memory by editing /boot/grub/grub.conf to set the dom0_mem variable to a high-enough value for Oracle Fabric Manager and the other controller processes.

   This consideration applies only if Oracle Fabric Manager is installed on an Oracle VM—for example, in a small 2-node install.

3. Verify that each node has:

   • A management port to the management network

   • A data port to the Oracle Ethernet Switch, which forms an internal data network also known as a physical network

   For external network connectivity, the controller node also needs:

   • An external port, which provides tenant networks with access to the outside world.

   For more information, refer to OVM documentation at https://docs.oracle.com/cd/E38500_01/.

4. Note the host to port mapping.

   Determine which host data ports connect to which ports on the ES2 switches.

5. Read the following to understand the packstack command.

   On each OpenStack node, in order for the OpenvSwitch (OVS) agent to properly manage the data port connected to the ES2 switch, the data port must plug into an OVS bridge, known as the physical bridge. The physical bridge name is typically the data network interface name prefixed by br-.

   For instance, the physical bridge name for eth1 is br-eth1.

   Each physical bridge also maps to a physical network name. You can create VLAN allocation pools based on a single physical network name to avoid using multiple actual interface names on each node.

   By default, the packstack command assumes the OVS agent on all nodes have homogeneous OVS-bridge-to-physical-interface mappings. For heterogeneous deployments, these mappings can be specified manually after the packstack completes.

   packstack --install-hosts=<controller
   IP>,<compute1
   IP>,<compute2
   IP... etc> 
    --neutron-l2-plugin=ml2 --neutron-ovs-bridge-mappings=<physical
   network>:<ovsbridge
   name> --neutron-ovs-bridge-interfaces=<ovs
   bridge 
   name>:<interface>
   

   When you have installed OpenStack, use the following steps to set up bridges.

6. (As needed) For each node, add the integration bridge (e.g. br-int), and map the physical bridge (e.g. br-eth1) to the physical port (e.g. eth1):

   ovs-vsctl add-br br-int
   ovs-vsctl add-br br-eth1
   ovs-vsctl add-port br-eth1 eth1
   

7. (As needed) Verify that OpenvSwitch bridges were created correctly:

   ovs-vsctl br-exists br-int
   ovs-vsctl br-exists br-eth1
   ovs-vsctl port-to-br-eth1 eth1
   ovs-vsctl show
   

8. (As needed) Edit the file /etc/neutron/plugins/openvswitch/ovs_neutron_plugin.ini and add the physical network name (e.g. default) to the physical bridge (e.g. br-eth1):

   [ovs]
   bridge_mappings = default:br-eth1
    
   [SECURITYGROUP]
   firewall_driver=neutron.agent.linux.iptables_firewall.OVSHybridIptablesFirewallDriver
   enable_security_group = True
   

9. Then restart the OpenvSwitch agent:

   service neutron-openvswitch-agent restart