Ask Your Question

What interface do I assign physical_interface_mappings for provider network?

asked 2019-02-04 21:25:08 -0600

mevry gravatar image

updated 2019-02-04 23:49:12 -0600

I am building a small cluster using the Rocky installation tutorial, however, my cluster design is a bit different. The following is how I have it connected: (OpenStack PoC Cluster)

On the controller, I set the physical_interface_mappings = provider:enp4s0, which is the connection to the provider network in the diagram above. I am currently configuring Neutron on a compute node. Every source seems to say that when configuring Neutron on a compute node, you should set physical_interface_mappings = provider:PROVIDER_INTERFACE_NAME. That would seem to imply that the compute node should be connected to the provider network directly. However, much of what I have read implies that North-South traffic is routed out to the Provider network through the network node (which is also the controller in my setup).

Am I confusing the network topology? If not, what interface should be assigned to physical_interface_mappings?

edit retag flag offensive close merge delete


What happens if you simply set physical_interface_mappings to an empty value on the compute node?

jsm gravatar imagejsm ( 2019-02-05 06:03:30 -0600 )edit

I tried as you said. Neutron didn't complain and I was able to complete installation. When I finish configuration, I will post an update about whether tenant networks were able to route out to the provider network

mevry gravatar imagemevry ( 2019-02-05 20:53:09 -0600 )edit

I am getting the following error 2019-02-09 22:40:42.725 32681 ERROR nova.api.openstack.wsgi NetworkNotFoundClient: Network ec3c94d7-50fb-41cc-843e-e55074dbdfc4 could not be found

mevry gravatar imagemevry ( 2019-02-10 01:15:41 -0600 )edit

2 answers

Sort by ยป oldest newest most voted

answered 2019-02-13 22:04:37 -0600

mevry gravatar image

I believe that if you are not using provider networks for your instances, you do not need to specify the physical_interface_mappings value on compute nodes. You would only need this if you wished to allow the provisioning of instances with direct access to the provider network (eg trusted instances). More likely, your instances would access the provider network through the network node, which can be set to route tenant traffic if you configure it to do so. Another thing to note is that no ip address needs to be assigned to interfaces designated for access to provider networks.

Take what I say with a grain of salt, as I am new to OpenStack. Someone please correct me if I am wrong.

edit flag offensive delete link more

answered 2019-02-05 07:46:31 -0600

novainfinite gravatar image

if you use provider all traffic go throw provider if you use selfservice all traffic go throw controller i test it and i think selfservice is useless and we should use provider

edit flag offensive delete link more


My goal is to allow tenants to create and manage their own isolated networks, so I'm assuming I need to enable self-service networks to accomplish this.

mevry gravatar imagemevry ( 2019-02-05 20:55:22 -0600 )edit

so all traffic throw controller and no way to solve it beacuse the router is in controller

novainfinite gravatar imagenovainfinite ( 2019-02-06 11:46:25 -0600 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2019-02-04 21:25:08 -0600

Seen: 87 times

Last updated: Feb 05