Thanks Matt peters,

Why we don't need data interface in controllers ?.
I think neutron dhcp and router namespaces will be in controller node, So packets should go to controller to route it as per old OpenStack release.
I don't know how its working in Stralingx , I think as per new OpenStack release packets don't go to controller.
Does Starlingx create dhcp and router namespaces in compute nodes ?.


Thanks,


On Thu, Jan 17, 2019 at 8:06 PM Peters, Matt <Matt.Peters@windriver.com> wrote:

The data network provides the physical infrastructure for the OpenStack guest tenant networks, which is used for both inter-compute and external network access from Virtual Machines (VMs).  The underlay configuration (VxLAN attributes, VLAN ranges, etc) is managed by the cloud administrator and is made available to the OpenStack tenants (applications).  The application requirements drive the topology of this network since the cloud operator must be able to support whatever application is being deployed within the VMs.

 

The virtual switch, OVS in the case of StarlingX, implements the OpenStack tenant networks and acts as the bridge between the physical infrastructure and the virtual networks.

 

For additional background information on OpenStack networking, please refer to the following:

https://docs.openstack.org/neutron/rocky/admin/intro.html

 

Hope that helps.

 

Regards Matt

 

From: Quick Convey <quickconvey@gmail.com>
Date: Thursday, January 17, 2019 at 2:55 AM
To: "Peters, Matt" <Matt.Peters@windriver.com>
Cc: "starlingx-discuss@lists.starlingx.io" <starlingx-discuss@lists.starlingx.io>
Subject: Re: [Starlingx-discuss] Starlingx network requirement

 

Thanks Matt peters,

 

Compute nodes communicate via this Data interface, right ?

VMs in different compute nodes also use this data interface for communication, right ? (communication between VM in CP1 -to- VM in CP2)

OVS also use this data interface to make tunnel between the compute nodes, right ?

 

You have mentioned that application requirements decide the required number of ports and network topology. Applications will be running in the VM and doesn't aware about the physical topology, right ? Could you please explain it.

 

Thanks,

 

 

On Mon, Jan 14, 2019 at 10:50 PM Peters, Matt <Matt.Peters@windriver.com> wrote:

See inline.

 

From: Quick Convey <quickconvey@gmail.com>
Date: Monday, January 14, 2019 at 3:56 AM
To: "starlingx-discuss@lists.starlingx.io" <starlingx-discuss@lists.starlingx.io>
Subject: [Starlingx-discuss] Starlingx network requirement

 

Dear All,

 

I am planing to setup Starlingx in bare-metal (controller-storage deployment)

 

I have couple of questions

 

Q1) What is the network requirements for this setup. All nodes should be in same network, that is the only requirement, right ?

The Management network is on all hosts and also serves as the PXEBoot network for booting other hosts from the Controller hosts.

The OAM network is required for controller hosts only.

The Data network is required for compute hosts only.

 

Q2) In "Hardware Requirements" section, I seen "Data: n x 10GE Compute", what that means, is it number of physical interfaces needed for data ? what that "n" indicate ? is it number of compute nodes ?

The ‘n’ indicates you can have more than 1 port if required for your application deployment.  The data networks are not used by the platform, so it is up to the application requirements to decide the required number of ports and network topology.

 

Q3) What is the number of physical interfaces needed in controller and compute bare-metal nodes ?. From the document I understand that only 2 physical interfaces are enough, right ?

Controller: 1 Mgmt, 1 OAM

Compute: 1 Mgmt, N Data (where N>=1)

 

Q4) Is there any picture which shows Management, OAM and Data interface connections between controller and compute nodes ?

I don’t think there is a StarlingX document that shows the interconnection.

 

Thanks,