Planning network topology is important for a successful deployment and functioning of the overcloud environment.
At the physical layer, two top of rack switches are deployed to provide redundancy and load balancing. Depending on switch vendor, either Virtual Link Trunking or Virtual PortChannel is configured. The required VLANs are created on the switches to separate the different network traffic types on the physical interfaces.
RHOSP uses the neutron networking service, which operates autonomously and manages software-based networks, static, floating and DHCP IP addresses. The director deploys the neutron service on each node in an overcloud environment.
In a typical RHOSP deployment, the number of logical network traffic types always outnumbers the physical networking links available on the nodes. To connect all the different network traffic types that overcloud provides, VLAN tagging is used to transmit more than one traffic on each network interface.
Note: For overcloud baremetal nodes, the VLAN selected for provisioning should be defined as the native VLAN on the switch port connected to the provision network port on the server. Also, PXE boot should be defined and mapped to the respective interface where the provision network is configured.
The following table provides details of all the network mappings used in OpenStack deployment with number of interfaces and VLANs:
Table 3. Network mapping details
- Network 1 – External Management
- Network 2 – Storage Management
- Network 3 – Provisioning
- Network 4 – PowerFlex data1
- Network 5 – PowerFlex data2
- Network 6 - Internal API
- Network 7 - Tenant
- Network 8 - Storage
- Network 9 – OpenStack External, Floating IP (mapped after overcloud creation)
| SAH Node | 4 (includes 2 bonded interfaces) | 9 (one for each network) |
- Network 1 – Provisioning
- Network 2 – External Management
| Undercloud VM | 2 | 2 (one for each network) |
- Network 1 – Provisioning (native VLAN, used for PXE)
- Network 2 - Internal API
- Network 3 - Tenant
- Network 4 – Storage
| Overcloud compute node | 3 (includes 2 bonded interfaces) or 2 (one for network 1 and other for remaining network flows using internal virtual bridge) | 4 (one for each network) |
- Network 5 - PowerFlex Data1
- Network 6 - PowerFlex Data2
| 2 (bonded interface) | 2 (one for each network) |
- Network 1 – Provisioning (used for PXE)
- Network 2 - Internal API
- Network 3 - Tenant
- Network 4 - Storage
- Network 5 - Storage Management
- Network 6 - OpenStack External, Floating IP (mapped after overcloud creation)
- Network 7 - PowerFlex Data1
- Network 8 - PowerFlex Data2
| Overcloud controller VM | 8 virtual network interfaces for respective network | 8 (one for each network) |
- Network 1 - Storage
- Network 2 - PowerFlex Data1
- Network 3 - PowerFlex Data2
| PowerFlex Gateway VM | 3 virtual network interfaces for respective network | 3 (one for each network) |
For more information about Planning networks, see Preparing your undercloud networking and Overcloud networks.
Note: All networks mentioned in the above table are isolated networks. In addition, overcloud networks are created automatically during the overcloud deployment using network environment files.
