Hello,
For all OpenStack setups I've followed there is a common requirement to have at least 2 NICs on the controller and compute nodes.
One for management and one public.
What's the reason for this requirement ? Is it possible to do an OpenStack setup with one interface only (for controller and compute nodes) ?
In case Controller/Network Node && N*Compute Node Setup
Compute :
NIC1 - MGMT
NIC2 - VXLAN (GRE) either VLAN connectivity to Controller
Controller
NIC1 - MGMT
NIC2 - VXLAN (GRE) either VLAN connectivity to Compute
NIC3 - External Network
Using same NICs for MGMT and VXLAN tunnelling is good only for POC deployments.
Thanks. I'm trying to understand why mixing management and external traffic is a bad idea. Why do you say "PoC deployments" only ?
That's not the point, Boris. It's just understanding the requirements: I have a few physical machines used for development purposes. I plan to move them to a private OS cloud. This is to get a better allocation of H/W, faster response time in terms of development infra requirements and so on.
Now, the machines already have a management network so getting access to a terminal is guaranteed. The only issue I see is that in case N/W congestion there is an impact on OS services and their dependencies (MariaDB , MongoDB, RabbitMQ, Apache server).
Usually one NIC for management network and one NIC for the external network used. Hope that helps...
Asked: 2016-10-08 15:33:56 -0600
Seen: 347 times
Last updated: Oct 09 '16
