-----Original Message-----
From: Andrew Grimberg [mailto:agrimberg@...]
Sent: Tuesday, November 21, 2017 11:40 PM
To: Xiao Ran(肖 然) <ran-xiao@...>; Sam Hague <shague@...>; Thanh Ha <thanh.ha@...>; Andrew Grimberg <andrew.grimberg@...>; integration-dev@...
Cc: ovsdb-dev@...; Matsumura Yasuyuki(松村 泰行) <yas-matsumura@...>; odlnw@...
Subject: [odlnw:03371] Re: [ovsdb-dev] Queries about changing network configuration of Jenkins deployed nodes in CSIT

ODL's systems currently, and in the future, run in a public cloud (essentially your diagram c below). We've got very little control over where nodes come up in the infrastructure as both Rackspace and our new provider are OpenStack on the underlying cloud infrastructure.

That is, we can't guarantee that any given node is not collocated on the same physical node or not.

Our new provider does give us a little more control over the network layouts, but we do not currently have anything in place that will allow us to dynamic lab bring ups that use more than one virtual network. I'll note that we have limits on the number of networks that we can have defined and having some method of properly allocating, and cleaning up, extra networks is work that would have to be figured out.

If you really want to stand up a configuration similar to your final diagram, it may be possible using one of our new providers "performance"
nodes with lots of CPU and RAM. We haven't tried virtualization in virtualization on their systems, but it would be what we have to do since we would still be launching into a VM managed by OpenStack that would then have to be configured to launch additional VMs inside of it.

-Andy-

On 11/21/2017 03:50 AM, Ran Xiao wrote:

Hi Sam,

Really appreciate for your help. And sorry for the late reply.

We would like to confirm the network structure of CSIT execution environment first.
Which of the following is the correct one?
a. OpenStack/ODL are running on physical environment
+------------------+ +------------------+ +------------------+
|Physical Machine#1| |Physical Machine#2| |Physical Machine#3|
| | | | | |
| Controller Node | | Compute Node | | ODL Node |
+------------------+ +------------------+ +------------------+
| | |
+------------------------------------------------------------------+ P
+------------------------------------------------------------------+ h
+------------------------------------------------------------------+ y
+------------------------------------------------------------------+ s
+------------------------------------------------------------------+ i
+------------------------------------------------------------------+ c
+------------------------------------------------------------------+ a
+------------------------------------------------------------------+ l
+------------------------------------------------------------------+ n
+------------------------------------------------------------------+ e
+------------------------------------------------------------------+ t
+------------------------------------------------------------------+ w
+------------------------------------------------------------------+ o
+------------------------------------------------------------------+ r
+------------------------------------------------------------------+ k

b. OpenStack/ODL are running on virtual environment
+-------------------------------------------------+
|Physical Machine#1 |
| |
| +---------------+ +------------+ +--------+ |
| |Controller Node| |Compute Node| |ODL Node| |
| +---------------+ +------------+ +--------+ |
| | | | |
| +-------------+ | +-----------+ |
| | | | |
| +----------+ |
| | Bridge | |
| +----------+ |
| |
+-------------------------------------------------+
|
+----------------------------------------------------+ Physical
+----------------------------------------------------+ network

c. OpenStack/ODL are running on virtual environment
+------------------------------------+ +------------------+
|Physical Machine#1 | |Physical Machine#2|
| | | |
| +---------------+ +------------+ | | +--------+ |
| |Controller Node| |Compute Node| | | |ODL Node| |
| +---------------+ +------------+ | | +--------+ |
| | | | | | |
| +-------------+ | | | | |
| | | | | | |
| +----------+ | | +----------+ |
| | Bridge | | | | Bridge | |
| +----------+ | | +----------+ |
| | | |
+------------------------------------+ +------------------+
| |
+-------------------------------------------------------------+ Physic
+-------------------------------------------------------------+ al
+-------------------------------------------------------------+ networ
+-------------------------------------------------------------+ k


The CSIT environment we want to deploy is as below. Is it possible?
+-----------------------------------------------------------------------------------+
|Physical Machine#1 |
| |
| +-----------------------------------------------------+ |
| | | |
| | +-----------+ | |
| | |Physical | | |
| | |Port | | |
| | |Bridge *1| | |
| | +-----------+ | |
| | (| L A G |)*2 | |
| | | | | |
| +---------------+ +------------+ +--------+ +-----------+ +-----------+ | |
| |Controller Node| |Compute Node| |ODL Node| |L2GW Node#1| |L2GW Node#2| | |
| +---------------+ +------------+ +--------+ +-----------+ +-----------+ | |
| | | | | | | | | | |
| | +-------------+ | +-----------+ | | | | | |
| | | | | +-------------------------+ | | | | |
| | | | | | +---------------------------------------+ | | |
| | | | | | | | | | |
| | +----------+ +----------+ | | |
| | |Ctrl Plane| | +----------------------+ | |
| | |Bridge *1| | | | |
| | +----------+ (|LAG|)*2 | |
| | +----------+ | |
| | |Data Plane| | |
| +----------------------------------------|Bridge *1|-------------------------+ |
| +----------+ |
| |

+---------------------------------------------------------------------
--------------+

*1 Connect each node with bridges divided as Ctrl Plane Bridge/Data
Plane Bridge/Physical Port Bridge
*2 Configure LAG for L2GW Node redundancy test
*Need to use Open vSwitch to create Data Plane Bridge.


Regards,
Ran


-----Original Message-----
From: Sam Hague [mailto:shague@...]
Sent: Tuesday, November 14, 2017 9:17 PM
To: Xiao Ran(肖 然) <ran-xiao@...>; Thanh Ha
<thanh.ha@...>; Andrew Grimberg
<andrew.grimberg@...>;
integration-dev@...
Cc: ovsdb-dev@...; Matsumura Yasuyuki(松村 泰行)
<yas-matsumura@...>; odlnw@...
Subject: Re: [ovsdb-dev] Queries about changing network configuration
of Jenkins deployed nodes in CSIT

Adding Andy and Thanh (and integration-dev) to thread for infrastructure questions since we are moving to a new provider and maybe things will be different.


On Tue, Nov 14, 2017 at 6:28 AM, Ran Xiao <ran-xiao@... <mailto:ran-xiao@...> > wrote:


Hi,

We are considering to change the network configuration of Jenkins deployed nodes in CSIT.
Could someone help to answer the following questions?
Is it possible to change the network configuration of nodes in CSIT test execution environment?
What we want to do are as below:
1. Divide the nodes' interfaces by usage (ctrl plane / data plane)


The infrastructure limits us to a single real interface. That is a limitation from Rackspace. This is why data and control go over a single interface. The current CSIT does have a workaround for the external networks where we add a br-physnet1 to simulate a different network. The new infra coming online at the beginning of the year might be different.

2. Use Open vSwitch for Bridge


This sounds like what we use as mentioned in 1. Looks like you just want to add another bridge and patch it to br-int?

3. Bonding I/F between Node and Bridge


Based on the limitation described in 1 I am not sure how this would look so need more details.

Thanks in advance.

Regards,
Ran Xiao


_______________________________________________
ovsdb-dev mailing list
ovsdb-dev@... <mailto:ovsdb-dev@...>
https://lists.opendaylight.org/mailman/listinfo/ovsdb-dev
<https://lists.opendaylight.org/mailman/listinfo/ovsdb-dev>

--
Andrew J Grimberg
Lead, IT Release Engineering
The Linux Foundation