16:05:00 <rkukura> #startmeeting networking_ml2
16:05:02 <openstack> Meeting started Wed Oct 15 16:05:00 2014 UTC and is due to finish in 60 minutes.  The chair is rkukura. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:05:03 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:05:06 <openstack> The meeting name has been set to 'networking_ml2'
16:05:18 <rkukura> #topic: Agenda
16:05:30 <rkukura> #link: https://wiki.openstack.org/wiki/Meetings/ML2#Agenda
16:05:46 <rkukura> Does anyone want to add anything to the agenda?
16:06:20 <rkukura> Ok, lets proceed
16:06:22 <rkukura> #topic Announcements
16:06:37 <rkukura> Juno release is scheduled for later this week
16:07:06 <Sukhdev> I think it is happening tomorrow
16:07:09 <rkukura> #link https://wiki.openstack.org/wiki/ReleaseNotes/Juno#OpenStack_Network_Service_.28Neutron.29
16:07:33 <rkukura> Several new ML2 drivers are mentioned as featured content!
16:08:30 <rkukura> Now that Juno is just about out the door, our focus shifts to blueprints, specs, and summit preparation
16:08:49 <pgpus> good
16:09:41 <rkukura> We should try to get blueprints into review, and ideally approved, before the summit, so the summit can focus on the areas where we don’t have consensus before the summit
16:09:52 <rkukura> Any other announcements?
16:10:22 <rkukura> #topic Action Items
16:10:32 <rkukura> There were no recorded action items from last week
16:10:46 <rkukura> #topic Bugs/Reviews
16:10:59 <rkukura> shivharis:
16:10:59 <shivharis> hi
16:11:27 <shivharis> I need to get to manishg's bugs that needed help..
16:11:37 <shivharis> the patch sets are:
16:11:56 <shivharis> https://review.openstack.org/126360
16:12:01 <manishg> shivharis: thanks!
16:12:16 <shivharis> manishg: u are covered on this right, or do u still need help
16:12:59 <pgpus> Robert, Are there any specific plug-ins proposed for opnfv in Neutron ML2?
16:13:17 <shivharis> manishg:?
16:13:21 <manishg> shivharis: the discussion is going elsewhere on that.  everyone finally settled … but I need to get some more detail from paul michali
16:13:37 <rkukura> pgpus: Lets cover that under the Kilo planning topic
16:13:42 <manishg> shivharis: https://review.openstack.org/#/c/124917/
16:13:46 <pgpus> ok
16:14:04 <manishg> this one has a couple of +1 but sitting there for a week.
16:14:06 <shivharis> so i take it you are good on the first one.
16:14:16 <shivharis> next:another, https://review.openstack.org/124917
16:14:18 <rkukura> I’ll try to review both these
16:14:30 <manishg> rkukura: thanks !
16:14:36 <shivharis> folks need to get this reviewed, thanks for any help
16:14:37 <Sukhdev> manishg: it is tagged for K1 - so, we are good there
16:14:43 <shivharis> rkukura: thx
16:14:48 <shivharis> one more
16:14:59 <shivharis> one more reviewer needed here
16:15:06 <shivharis> anyone?
16:15:12 <manishg> Sukhdev: yes, not critical.
16:15:39 <shivharis> Ok, moving on
16:15:50 <romilg> Also I request to close this https://bugs.launchpad.net/neutron/+bug/1224978 asap ...I have to run behind the HEAD https://review.openstack.org/#/c/103546/
16:15:51 <manishg> shivharis: one question about a bug.
16:15:52 <uvirtbot> Launchpad bug 1224978 in neutron "port binding on multi segment networks could lead to agent misconfiguration" [Medium,In progress]
16:16:13 <shivharis> https://bugs.launchpad.net/neutron/+bug/1377346
16:16:14 <uvirtbot> Launchpad bug 1377346 in neutron "ML2: Invalid unit test case" [Medium,In progress]
16:16:33 <shivharis> manishg: we can close this one?
16:16:47 <manishg> shivharis: I believe romilg said he fixed it as part of another patch.
16:16:59 <shivharis> romilg: we'll get to the one u mentioned
16:17:09 <manishg> we can close if covered.  I'll take a look at the patch today.
16:17:15 <romilg> yeah it is part of https://review.openstack.org/#/c/103546/
16:17:16 <shivharis> so can we close that
16:17:21 <rkukura> romilg: I’ll review the multi-segment patch - want to see how it fits with the HPB changes I’m planning
16:17:24 <manishg> shivharis: yes, thanks.
16:17:39 <romilg> thanks :)
16:18:14 <manishg> shivharis: https://bugs.launchpad.net/neutron/+bug/1264608 - sorry to bring l2 agent related bug.
16:18:22 <uvirtbot> Launchpad bug 1264608 in neutron "openvswitch agent plugin should execute batched ovs-vsctl CLI statements on __init__()" [Medium,Triaged]
16:18:35 <manishg> but this is marked for later due to dependency on soft-restart.  which is delayed due to modular L2 agent work.
16:18:54 <manishg> but is there any ETA on modular L2 agent ?  I know we are going to discuss in Paris.
16:19:28 <manishg> but seems like there is still quite a bit of work left.  If the decision is to not touch the agents then it's okay.
16:19:29 <shivharis> romilg: why can u not close the bug, not sure i understood your point
16:20:11 <rkukura> manishg: What happened with the OVS agent work otherwiseguy was doing during Juno? This seems related.
16:20:30 <shivharis> manishg: can we take things in order, please hold your thoughts for a moment
16:20:40 <shivharis> romilg: ?
16:20:52 <manishg> shivharis: apologies.  will wait till later.
16:20:54 <romilg> one sec
16:22:07 <romilg> My point is the following patch set https://review.openstack.org/#/c/103546/ will close both the bugs 1224978 and 1377346
16:22:10 <uvirtbot> Launchpad bug 1224978 in neutron "port binding on multi segment networks could lead to agent misconfiguration" [Medium,In progress] https://launchpad.net/bugs/1224978
16:22:11 <uvirtbot> Launchpad bug 1377346 in neutron "ML2: Invalid unit test case" [Medium,In progress] https://launchpad.net/bugs/1377346
16:22:30 <romilg> so we need core reviewer to give +2
16:22:57 <romilg> I has been reviewed by other reviewer so many time and had +1's
16:23:17 <shivharis> rkukura: can you please look into this
16:23:36 <shivharis> I will also review this
16:23:48 <rkukura> shivharis: yes
16:24:00 <shivharis> Sukhdev: can you also please spend some time on this one
16:24:06 <Sukhdev> amotoki: can you look at this as well?
16:24:16 <shivharis> Any other, (manishg?)
16:24:23 <Sukhdev> shivharis: will do
16:25:13 <manishg> shivharis: re- bugs related to agents.
16:25:28 <romilg> Here is another bug that I completed https://bugs.launchpad.net/neutron/+bug/1179223
16:25:31 <uvirtbot> Launchpad bug 1179223 in neutron "Retired GRE and VXLAN tunnels persists in neutron db" [High,In progress]
16:25:39 <manishg> I believe we want to push them out until modular L2 agent is out there right?
16:26:11 <shivharis> rkukura: ml2 agent in the works? we will discuss in summit topics
16:26:18 <shivharis> moving on..
16:26:19 <manishg> but is there ETA for that?  if not, should there be any attempt to fix these? that was the question I wanted to clarify related to the bugs.
16:26:33 <manishg> shivharis: ok, thanks.  will talk at summit.
16:27:12 <shivharis> We have been collecting a lot of stale bugs, I am going to start pinging owners to this cleared
16:27:23 <rkukura> manishg: I’m not sure we want to delay fixes to the agents as long as they remain relevant if/when we have a plan for modular agent
16:27:25 <Sukhdev> manishg: I think we should not hold off fixes waiting for ML2 agent, what do you think rkukura ?
16:27:39 <rkukura> Sukhdev: agree
16:28:08 <rkukura> shivharis: any more bugs to discuss?
16:28:19 <shivharis> One more point
16:28:38 <manishg> rkukura: ok, will bring it up in open discussion.
16:28:39 <shivharis> If you are an owner of a bug please add a comment, i actually look at all the comments, please do so
16:28:56 <shivharis> thats all
16:29:20 <rkukura> I’ll add that, along with spec and summit planning, now is a great time to be squashing bugs
16:29:33 <rkukura> #topic Specs
16:30:07 <rkukura> Lets keep updating https://wiki.openstack.org/wiki/Tracking_ML2_Subgroup_Reviews#Under_Review as we move specs to kilo and/or add new ones
16:30:20 <rkukura> #topic Kilo Design Summit
16:31:23 <rkukura> So overall neutron design summit is in https://etherpad.openstack.org/p/kilo-neutron-summit-topics, but I’m not sure what the process is to assign sessions
16:31:57 <rkukura> So we’ve been discussing possible Kilo themes for ML2 here, with the intent to see which ones really need time at the summit
16:32:17 <rkukura> I’ve moved the list of ideas from the meeting agenda to an etherpad
16:32:28 <rkukura> #link https://etherpad.openstack.org/p/ML2_Kilo_Themes
16:32:57 <rkukura> Hopefully this will let us more easily collaborate both during and between the weekly ML2 meetings
16:33:04 <pgpus> ok
16:33:15 <manishg> rkukura: +1
16:33:26 <Sukhdev> We should discuss - for which topics it makes sense to request a session at the summit
16:33:35 <rkukura> Overall, we need to get some sort of agreement on what we think the major community goals are for ML2 in the Kilo cycle
16:33:56 <rkukura> Then make sure we are getting BPs and specs into review for them.
16:34:13 <pgpus> are we starting with DVR support discussions now?
16:34:36 <rkukura> If we don’t have enough consensus to do that, but think its important, we’ll ask for a session
16:35:23 <rkukura> I think we had some discussion on DVR and HPB last week, so maybe should start with backend sync / error handling topic
16:35:51 <rkukura> I also added an item on TaskFlow, which might be related to that, or might be separate
16:35:56 <otherwiseguy> rkukura: still working on OVS stuff. Will try to push an updated patch with some tests today or tomorrow.
16:36:18 <manishg> otherwiseguy: link please.
16:37:18 <pgpus> Ok I will look at past meetings to synch up with all before summit...thx
16:37:19 <otherwiseguy> manishg: https://review.openstack.org/#/c/114050/
16:37:31 <manishg> otherwiseguy: thanks.
16:38:00 <rkukura> sorry - had an interruption
16:38:49 <rkukura> I think at least me, Sukhdev, and manishg are looking into the backend sync and also TaskFlow.
16:39:07 <rkukura> I put some of the items from some emails into the etherpad
16:39:36 <rkukura> Do we need disussion of this here? Or do we agree its relatively high priority for kilo?
16:40:00 <manishg> rkukura: ack.  I'm trying to understand the issue, and if taskflow works there, etc. and if not, work with Sukhdev on other solutions.
16:40:45 <rkukura> So do we think we can have an agreed approach prior to the summit, or will we likely need to work on this at the summit?
16:40:53 <manishg> rkukura: I think something like taskflow is perhaps needed in some form for a decent error recovery and something that can be persistent (recover after restart).  So looks good to me.
16:41:46 <manishg> rkukura: we have two weeks.  we can atleast undersand the issues involved and present them in meeting.
16:41:54 <rkukura> My feeling is that we will make some progress on understanding TaskFlow and maybe some other options for sync/recovery, but will need to discuss and choose approaches at summit.
16:42:03 <shivharis> rkukura: we have been on this topic several times, we need to make decisions, I suggest do a pod at summit
16:42:23 <manishg> and hopefully we can have some consensus before summit but will need discussion there too.
16:42:35 <rkukura> shivharis: Agreed on pod and/or session
16:42:37 <manishg> shivharis: +1
16:42:38 <Sukhdev> rkukura manishg: it will be a good goal to have a good understanding of the solution - perhaps we can present the solution to wider audiance at the summut
16:42:57 <manishg> Sukhdev: agree.
16:43:06 <rkukura> Ok, bulk ops is next on the list
16:43:20 <Sukhdev> rkukura: This one may be a good one to ask for a summit session - markmcclain has some input on taskflow as well
16:43:30 <shivharis> banix is out
16:43:38 <rkukura> shivharis: right
16:43:58 <rkukura> I think this is current tracked as a bug - do we need a BP on it for Kilo?
16:44:01 <manishg> Sukhdev: agree.  markmcclain had also mentioned it in last summit .  so good to get his input too.
16:44:45 <rkukura> One possibility is that TaskFlow lets us make the entire plugin operation safe to call within transactions, which would mean banix’s changes would not be needed
16:45:31 <Sukhdev> manishg: Yes, once we have built our understanding, then will be appropriate to ping him - to cross check
16:45:36 <rkukura> I’m not thinking we’ll need a session specifically on bulk ops
16:46:00 <rkukura> Next items is imrproving DB transactions/locking
16:46:45 <shivharis> rkukura: what is the current issue with db transaction/locking?
16:46:53 <rkukura> The fact that we do with_lockmode(‘update’) but that it doesn’t really do global row locking with clustered DBs is a concern.
16:47:05 <rkukura> The semaphore locking also limits throughput.
16:47:25 <rkukura> I’d really prefer we take an optimistic approach rather than doing pessamistic locking
16:47:41 <pgpus> agreed
16:47:43 <shivharis> ah ha
16:47:50 <rkukura> I think some progress has been made in this direction, and we can possibly complete it in Kilo
16:48:22 <manishg> rkukura: any pointers?
16:48:53 <rkukura> manishg: I haven’t looked into this in detail for a while, so nothing specific right now
16:49:02 <manishg> rkukura: ok, thanks.
16:49:04 <rkukura> Just wondering if this is something that needs a BP
16:49:47 <rkukura> Anyone interested in working on this?
16:50:24 <manishg> rkukura: will take a look at the issue and let you know.
16:50:25 <rkukura> Please add ideas to the etherpad
16:50:30 <rkukura> manishg: thanks
16:50:50 <pgpus> Are there any BPs for bypassing DPDK like in OVS by Linoro community for ML2 drivers?
16:50:57 <HenryG> On the topic of DB operations, Mike Bayer filed a "bug" about optimizing joins or something.
16:51:08 * HenryG looks for the bug ...
16:51:13 <rkukura> Extension drivers is mentioned on the neutron etherpad as well
16:51:35 <rkukura> HenryG: Lets put links to bugs, related threads, etc. in the etherpad
16:51:41 <yamahata> I'm interested in extension drivers
16:52:07 <rkukura> yamahata: Do you have specific goals?
16:52:22 <HenryG> rkukura: Understood, but it looked more like a BP than a bug.
16:52:22 <Sukhdev> rkukura: I think we had a spec on the extensions during Juno, which got pushed to Kilo
16:52:30 <yamahata> rkukura: actually I started with port security extension.
16:52:57 <HenryG> rkukura: https://bugs.launchpad.net/neutron/+bug/1380823
16:52:58 <uvirtbot> Launchpad bug 1380823 in neutron "outerjoins used as a result of plugin architecture are inefficient" [Undecided,New]
16:53:00 <yamahata> during addressing it, issues were found. So I'd like to clear them first.
16:53:07 <rkukura> yamahata: right - there are specs and code in review for portsecurity
16:53:35 <rkukura> yamahata: ok
16:53:49 <yamahata> rkukura: I'm also willing to address semantic enforcement etc...
16:54:17 <rkukura> I’m interested in making ML2 port binding able to enforce the semantics required by the values of extensions, but not sure if this is important during Kilo
16:54:52 <yamahata> rkukura: I see. Okay
16:55:10 <rkukura> yamahata: sounds like we should discuss that, and see what makes sense to propose for kilo
16:55:23 <rkukura> OK, lets talk about modular L2 agent
16:55:33 <rkukura> We’ve got about 6 minutes left
16:55:56 <rkukura> So this has been discussed for several summit, but we’ve never converged on a plan
16:56:31 <Sukhdev> rkukura: Shall we create an action to discuss it next week?
16:56:32 <pgpus> Besides security extension can there not be a use case for performance improvements like Open Data Plane (ODP) efforts?
16:56:35 <shivharis> One question on ML2 agent: how many agents do we have that need to come under the ML2 agent umbrella?
16:57:30 <Sukhdev> shivharis: ideally all, but, start with one
16:57:38 <manishg> shivharis: I believe the idea is to avoid duplication of code between agents.  Some of work is perhaps common.  So I'd guess all.
16:57:56 <shivharis> i know the motivation, but there are not too many
16:58:00 <rkukura> Lets start with ML2 agent on this part of the agenda next week
16:58:10 <shivharis> rkukura: oj
16:58:11 <shivharis> ok
16:58:11 <pgpus> I  understood as we have one plugin and many derivers with ovs as default
16:58:16 <manishg> when's banix back?
16:58:28 <rkukura> I think extensibility of the agents is also important
16:58:50 <Sukhdev> manishg: he said he is out for several weeks - what ever that means :-)
16:58:51 <manishg> rkukura: +1
16:59:02 <rkukura> #topic Open Discussion
16:59:26 <pgpus> when you have only one plugin all extensions mergae and have same API, except common minimum rest vendors can implemnet what they think is needed for their hardware
16:59:35 <romilg> In the end I also request all to spend some time to review this https://review.openstack.org/#/c/121000/
17:00:20 <rkukura> pgpus: Not sure what you are getting at?
17:01:07 <rkukura> Ok, please add/update ideas for themes in the etherpad
17:01:32 <pgpus> I ma trying to see what support we can add for Carrier grade in ML2 if possible in Kilo to support nfv
17:02:06 <rkukura> pgpus: Please add something to the etherpad if you think this is a possible kilo theme
17:02:10 <pgpus> will do
17:02:15 <rkukura> #endmeeting