16:02:49 <Sukhdev> #startmeeting networking_ml2
16:02:50 <openstack> Meeting started Wed Oct 22 16:02:49 2014 UTC and is due to finish in 60 minutes.  The chair is Sukhdev. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:02:51 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:02:53 <openstack> The meeting name has been set to 'networking_ml2'
16:03:05 <Sukhdev> #topic: Agenda
16:03:54 <Sukhdev> #link https://wiki.openstack.org/wiki/Meetings/ML2#Meeting_October_22.2C_2014
16:04:14 <Sukhdev> #topic: Announcements
16:04:40 <Sukhdev> rkukura and banix are off today for personal reasons
16:05:07 <Sukhdev> Juno Release is out - in case you did not know
16:05:23 <Sukhdev> #link - https://launchpad.net/neutron/juno/2014.2
16:05:49 <Sukhdev> Anybody has any announcement?
16:06:13 <shivharis> in celebration of this release lets have a short meeting :-)
16:06:33 <Sukhdev> shivharis: sure - I am all for it...
16:07:02 <Sukhdev> There are couple of things I have in mind to cover -
16:07:14 <Sukhdev> First - discussion about TaskFlow
16:07:38 <Sukhdev> and second - Kilo summit session for ML2
16:07:39 <shivharis> i missed the meeting at yahoo :(
16:08:27 <Sukhdev> but, before we get into those topics, lets get the usual stuff out of the way
16:08:46 <Sukhdev> #topic: Bugs
16:09:02 <Sukhdev> shivharis: do you want to cover anything here?
16:09:03 <shivharis> hi
16:09:33 <shivharis> We have a lot of bugs that need assignment, folks please feel free to pick them up
16:09:44 <romilg> sure
16:09:54 <shivharis> All have target of kilo-1
16:09:59 <manishg> ok.  will do.
16:10:20 <shivharis> We have 4 high bugs - 3 have good progress and 1 i am not sure - so i will ping
16:10:37 <amotoki> if you find bugs which affect juno please add tag juno-backport-potential.
16:10:39 <shivharis> romilg, manishg: thanks
16:10:45 <romilg> https://bugs.launchpad.net/neutron/+bug/1179223
16:10:49 <uvirtbot> Launchpad bug 1179223 in neutron "Retired GRE and VXLAN tunnels persists in neutron db" [High,In progress]
16:10:53 <shivharis> amotoki: good point
16:11:10 <romilg> I posted the patch set the same need more reviewers
16:11:45 <absubram__> hi.. I have a low priority bug in review.. https://review.openstack.org/#/c/105514/ it has two +2s.. but one -1 from Yamamoto Takashi which is holding it up..
16:11:49 <manishg> shivharis: I see bugs for midonet plugin there too.  these bugs I assume generally get (and should be) picked up by vendors, right?
16:11:53 <shivharis> romilg: folks please jump on the bug that romilg has pointed  - it is one of the high pri bugs
16:12:06 <shivharis> manishg: yes and no
16:12:26 <shivharis> manishg: once the plugin/driver comes into the common base
16:12:44 <shivharis> if belongs to the community, and the community own supporting it
16:13:06 <amotoki> here is ML2 meeting :-)   and vendor specific bugs should be dealt by plugin maintainers.
16:13:07 <shivharis> usually/mostly it is taken care of by the vendor, due to hardware etc.
16:13:12 <manishg> shivharis: but it only impacts those using midonet plugin, right?
16:13:24 <Sukhdev> absubram__: that needs workflow checked and it should merge
16:13:58 <absubram__> Sukhdev: his -1 comments have already been debated and addressed in previous versions.. anyway to get in contact with him about his blocking this merge?
16:14:06 <shivharis> we also have a list of plugin and their primary maintainers and we expect active support from them
16:14:07 <manishg> shivharis: and that's why the question.  I'll leave them alone for now.  there seems to be other bugs out there that need owners.
16:14:29 <shivharis> manishg: fair statement
16:14:33 <absubram__> Sukhdev: anything else I need to do in order for the workflow check to happen?
16:14:57 <amotoki> re: absubram__ review, I will check if we can override -1 from yamamoto and it is worth +A.
16:15:16 <absubram__> amotoki: thank you very much!
16:15:22 <Sukhdev> absubram__ Have Sumit or Edger check the workflow and it should merge
16:15:36 <shivharis> have you contacted yamamoto?
16:15:53 <Sukhdev> absubram__: you do not need to worry about -1, after you have two +2
16:15:53 <absubram__> shivharis: yes I sent him an email
16:15:58 <amotoki> i don't think we need to reach him. just review!
16:16:11 <shivharis> amotoki: can you override the -1
16:16:20 <SumitNaiksatam> absubram__: we can +A if we dont get a response within the next day
16:16:26 <SumitNaiksatam> absubram__: just ping me
16:16:31 <Sukhdev> amotoki: thanks
16:16:38 <absubram__> Sukhdev: Summit had mentioned in his comment he'd +A pending on removal of the -1.. I'll contact both again to see what to do
16:16:50 <shivharis> absubram_: thanks, anything else on this?
16:17:09 <absubram__> SumitNaiksatam: Thank you.. that sounds reasonable to me
16:17:12 <absubram__> thanks all
16:17:20 <shivharis> we are doing "quite average" in terms of the bug count
16:17:40 <shivharis> I will start assigning (if unassigned) bugs after the summit
16:17:50 <shivharis> thats all from me for now
16:18:27 <Sukhdev> OK lets move on to next topic
16:18:40 <Sukhdev> #topic: TaskFlow Update
16:19:04 <Sukhdev> manishg arranged a meeting at Yahoo last week to discuss about TaskFlow
16:19:39 <Sukhdev> The goal was to see how can we use it for ML2 Sync as well as error recovery
16:20:07 <Sukhdev> #link: http://docs.openstack.org/developer/taskflow/index.html
16:20:17 <shivharis> what was the takeaway?
16:20:39 <Sukhdev> So, we looked at various possibilities
16:21:13 <Sukhdev> Considered using it at Neutron API level, ML2 Plugin level, or at individual driver level
16:22:05 <Sukhdev> the best part was that the author of TaskFlow was present in the room to answer questions
16:23:01 <Sukhdev> Currently Cinder project is planning on using it with zookeeper and jobboard
16:23:32 <manishg> shivharis: the meeting was more from the ml2-sync-blueprint point of view originally and how taskflow
16:23:35 <Sukhdev> we felt for ML2 level, it will be too heavy and hence decided to come up with simpler light version
16:23:36 <manishg> could potential be used.  so folks working on on that blueprint were involved.
16:24:14 <Sukhdev> manishg has volunteered to work on a lighter version
16:24:22 <Sukhdev> for our ML2 use
16:24:34 <Sukhdev> At this point we are interested in two things:
16:24:53 <Sukhdev> 1) to get a better feeling about it - hence put together a POC
16:25:18 <Sukhdev> 2) to come up with an architecture/design to deploy it
16:25:35 <Sukhdev> manishg is planning on working on 1)
16:25:52 <Sukhdev> manishg: please jump in and fill in what ever I missed
16:26:10 <manishg> all: if you are interested in discussions related to ml2-sync and/or taskflow usage etc. let us know and we can be in touch via mails, etc. for updates.
16:26:32 <amotoki> is there any target driver for POC?
16:27:17 <shivharis> manishg: have the decisions been made or are we still at discussion stage?
16:27:18 <Sukhdev> amotoki: not really, we figured we use a mocked one at this point
16:27:26 <manishg> Sukhdev: for now I think the plan is like what you mentioned.  To experiment with the idea of taskflow and understand the scope, and do some POC to see how applicable it would be.
16:27:46 <manishg> amotoki: we'll use mock.  and then it should be applicable to all if that works out well.
16:28:20 <manishg> there were a few ideas considered.  we'll start with the simple ideas and hope to gather some experience by summit time.  so we can have more useful discussion there.
16:28:21 <Sukhdev> shivharis: nothing is decided - only POC - and all in discussion stage
16:28:58 <Sukhdev> All - if interested, please speak up - we can use more brain power
16:29:09 <manishg> shivharis: no decisions.  but it would be good to be familiar with it before Paris summit so we can all dig in deeper together.
16:30:01 <Sukhdev> we figured we get some understanding and then use PODs at Paris summit to do more discussions
16:30:19 <shivharis> manishg: I am still trying to sync up to where you guys are already .. summit we be a good place to further the conversation
16:30:33 <amotoki> it is the next topic :-)
16:30:53 <Sukhdev> amotoki: you mean summit discussion - yes
16:31:18 <manishg> shivharis: cool.  I can ping you (and other who are interested) offline abut the details.
16:31:19 <Sukhdev> manishg: want to add anything before we move on?
16:31:53 <manishg> Sukhdev: I think I'll have more update once I get something going.  Hope to write some intro doc too
16:31:56 * Sukhdev waiting
16:32:10 <Sukhdev> manishg: cool - thanks
16:32:18 <Sukhdev> OK - so lets move on
16:32:53 <Sukhdev> All - feel free to join us at the PODs in Paris summit
16:33:16 <Sukhdev> #topic: Kilo Design Summit
16:33:48 <Sukhdev> As you know number of sessions at summit are reduced this time
16:34:04 <Sukhdev> #link https://etherpad.openstack.org/p/kilo-neutron-summit-topics-distilled
16:34:30 <Sukhdev> See the etherpad with proposed sessions
16:34:34 <amotoki> 7 slots + Friday whole day pod
16:35:06 <amotoki> woops 4+7 slots + ...
16:35:15 <Sukhdev> If you look at it, we have a part of one session
16:35:32 <Sukhdev> amotoki: I thought something was wrong ;-)
16:36:03 <Sukhdev> https://etherpad.openstack.org/p/neutron-l2
16:36:18 <Sukhdev> ML2 topics are covered in sub-pad
16:36:22 <amotoki> Sukhdev: sorry if it was wrong....
16:36:35 <Sukhdev> amotoki: np
16:36:57 <Sukhdev> So, I wanted to cover the ML2 part here
16:37:22 <Sukhdev> We have been covering this for past few sessions
16:37:31 <amotoki> i think ML2 topic is not so general and I believe POD session fits us. we need more detail discussions I blieve.
16:37:49 <Sukhdev> amotoki: correct
16:38:48 <Sukhdev> Does anybody believe we are missing anything?
16:39:32 <Sukhdev> One of the thoughts is that we use PODs for detailed discussions
16:40:22 <Sukhdev> and use part of the main session to announce/give overall run down (and intro) of the topics to be covered in PODs
16:40:50 <amotoki> in the past summits ML2 session was very busy and just presented the essense...
16:41:11 <amotoki> i feel it was small input.
16:41:34 <Sukhdev> amotoki: correct, we had to squeeze too much into a session and not much time for Q&A
16:41:59 <Sukhdev> or discussion
16:42:14 <amotoki> I believe we can announce rough pod scheduel in the main session.
16:42:32 <amotoki> so we can announce all of them.
16:43:27 <Sukhdev> amotoki: yes, I think that will be a good idea
16:44:08 <Sukhdev> Focusing back on the etherpad, is every item clear?
16:45:29 <Sukhdev> Do we need to list them in any specific order - or priority?
16:46:02 <shivharis> we can do that at the pod
16:46:43 <Sukhdev> shivharis: sure - or this will be good forum as well - perhaps everybody may not make it to summit
16:47:16 <shivharis> Sukhdev: +1
16:47:24 <Sukhdev> Also, feel free to update etherpad
16:47:57 <manishg> Sukhdev: Re bulk-ops item.  that is for the agent, right?
16:48:09 <manishg> "Complete the bulk ops support"
16:48:11 <amotoki> I am not sure who can lead the topics. is it clear?
16:48:39 <amotoki> manishg: feel free to add it
16:48:56 <manishg> amotoki: +1 - I think it might be good idea to put some links or owner/moderator there per item.
16:49:00 <amotoki> i think it is not specific to agnet support
16:49:06 <Sukhdev> We need volunteers to work on these items
16:49:36 <Sukhdev> manishg: Neuron API supports bulk ops
16:49:52 <manishg> Sukhdev: need clarification on what this is about.
16:50:02 <amotoki> topics in the latter half (except agent refactor) seems BP topic.
16:50:05 <Sukhdev> we at ML2 level have issue with it
16:50:26 <Sukhdev> banix is leading that effort - he is missing for personal reason
16:50:32 <manishg> Sukhdev: yep, aware of neutron part.  ah, ok.  got it.
16:51:02 <manishg> Sukhdev: so ML2 gets status (and port details) one at a time via the rpc from neutron?
16:51:08 <manishg> I mean the agent.
16:52:17 <Sukhdev> manishg I think it is the other way around
16:52:33 <Sukhdev> a bulk operation is issued by the API, ML2 executes one at a time
16:52:49 <Sukhdev> how do we deal with it if we fail in the middle - some driver failures
16:53:08 <Sukhdev> it is all done in a transaction -
16:53:19 <Sukhdev> how to roll back, etc...
16:53:26 <manishg> ah, just looked at the code.  yeah, don't see bulk operation methods in ml2 plugin. thanks.
16:53:34 <shivharis> manishg: prior IRCs have loads of info on this (lots of discussions and subtle points)
16:54:00 <manishg> shivharis: thanks much.  will take a look.
16:54:13 <manishg> Sukhdev: so banix would be the contact point for this ?
16:54:32 <Sukhdev> manishg: when looking at taskflow, keep it in mind - we may be able to solve it in a much cleaner way
16:54:52 <manishg> Sukhdev: right.  will do.
16:55:28 <Sukhdev> manishg: correct….but, he will be unavailable for a while - he will most likely miss the summit as well
16:55:54 <amotoki> async ops vs bulk ops... difficult topic and they are not orthogonal.
16:55:57 <Sukhdev> manishg: he may respond to the emails - he was in touch with me via email
16:56:19 <manishg> Sukhdev: ok thanks.  is anyone else working with banix who may be aware of current state of the affairs?
16:56:26 <manishg> Sukhdev: will send emails.  thanks.
16:56:41 <Sukhdev> manishg: shivharis is good source as well
16:57:10 <manishg> Sukhdev: ok great.  will bug him too ;-)
16:57:32 <Sukhdev> amotoki: are we doing anything about async at Neutron API level?
16:58:23 * Sukhdev just realized we are almost out of time
16:58:30 <amotoki> Sukhdev: I am not sure at now task topic will be covered in REST API sessions
16:59:00 <Sukhdev> #topic: Open Discussion
16:59:16 <Sukhdev> Any body want to bring up anything here?
17:00:15 <Sukhdev> amotoki: we considered async ops as a part of taskflow discussion as well - and, it seems we could easily support it
17:00:31 <Sukhdev> amotoki: but, requires lots more investigation
17:00:58 <shivharis> we are meeting next week? will that be the the last meeting prior to the summit
17:01:12 <Sukhdev> Well, Folks thanks for attending - please update the etherpad with your input
17:01:15 <manishg> thanks!
17:01:22 <Sukhdev> shivharis: correct
17:01:24 <ChuckC> o/
17:01:37 <romilg> thanks
17:01:39 <Sukhdev> Thanks folks - we are out of time
17:01:41 <Sukhdev> bye
17:01:45 <shivharis> thanks all, thanks Sukhdev for chairing
17:01:46 <Sukhdev> #endmeeting