[ovs-dev] Meeting summary and action items (7/29)

[Nithin Raju]

hi Folks,
Alessandro might be writing up a more detailed transcript of the IRC meeting we had on Tuesday. I just wanted to cover things at a high level:

Attendees:
>From VMware : Justin Pettit, Ben Pfaff, Guru Shetty, Saurabh, Ankur, Nithin and Eitan
>From Cloudbase: Alin Serdean and SamuelG, Alessandro.

0. There was discussion on the code that got checked in.

1. Bug tracking/Feature(blueprint) tracking & Reviews:
- We decided to use github for this:
https://github.com/openvswitch/ovs/issues
- Reviews will continue to happen on dev at openvswitch.

2. Testing issues that Alin was facing:
- Alin said he'd try using the V3 patch that Saurabh sent out. If there are issues still, VMware folks would help debugging. Eitan Eliahu and Alessandro like Skype. We'll figure out which way to go in terms of tools.

3. Netlink implementation in the kernel
- At a high level, there are 3 sub-tasks:
  1) adding NL message parsing code in the kernel
  2) interpreting data structures defined in openvswitch.h in the kernel
  3) adapting the IO mechanism to the rest of the code
- Cloudbase folks pointed out that, #1 and #2 are already part of Cloudbase's kernel implementation sent out Tuesday morning (PST). Since there was not much time to look at it before the meeting, VMware folks will look at that code, and potentially look at leveraging code from it. We'll discuss about #3 after #1 and #2.

4. Architectural topic - port naming
- Everyone agreed that Cloudbase's powershell extension to set the "port name" for a VIF using WMI from userspace is the right way to go.
- Cloudbase would send out patches for the same.
- The kernel code will have to be modified to read the "Friendly Name" set in the port, rather than coming up with a name such as vmNICEmu1000048.

5. Architectural topic - multiple datapaths
- Everyone tried to get on the same page here to understand what the requirement and the use case is. There was a lot of discussion on this topic.
- It can be summarized that, Alessandro and team were concerned that there's no support today for binding the OVS extension to multiple Hyper-V switches at the same time. This was seen to be necessary, since multiple Hyper-V switches are a well known usage model on Hyper-V to ensure network isolation between switches. (ie. VMs on different switches)
- VMware folks tried to clarify the model to provide isolation on OVS to use bridges. Even though, in the kernel all the VMs are part of the same "datapath", OVS userspace ensures that network isolation is implemented according to the configuration defined in OVSDB. Userspace adds flows to the kernel to let the kernel know how to forward packets.
- Thanks Justin Pettit for explaining the OVS terminology and laying out the workflow.
- #Action-item: Nithin to provide a document mapping hyper-v concepts to OVS.
- #Action-item: Alessandro to provide a document explaining all possible use cases, Cloudbase foresees.

6. CI:
- Cloudbase was interested in setting up a CI along with Microsoft (Peter Pouliot). This is fine with VMware folks, and it seen as generally useful.

7. Release Dates
- There was discussion about the availability of OVS on Hyper-V for the Openstack summit during Nov 1st week. The session is jointly proposed by VMware (Justin Pettit) and Cloudbase.
- No commitments were made. But, folks from both teams have started working on the next immediate steps.

8. Future meetings:
- Everyone felt it was a very productive (and lengthy :)) meeting, and decided that Tuesday 10 AM PST would be a suitable time for future meetings as well.


I have tried to capture the salient points discussed. I might have missed out something. Pls. add on if it helps for any future reference.

thanks,
Nithin