#openstack-meeting log

21:02:45 <markmcclain> #startmeeting Networking
21:02:46 <openstack> Meeting started Mon Aug 12 21:02:45 2013 UTC and is due to finish in 60 minutes.  The chair is markmcclain. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:02:47 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
21:02:49 <openstack> The meeting name has been set to 'networking'
21:02:52 <markmcclain> #link https://wiki.openstack.org/wiki/Network/Meetings
21:03:11 <markmcclain> #topic Announcements
21:03:32 <markmcclain> #info arosen is now a member of the stable team
21:03:42 <garyk> yay!
21:04:32 <markmcclain> Will definitely be nice to increase our review bandwidth there.
21:05:03 <arosen> Yup :)
21:05:51 <markmcclain> The Havana Feature Proposal Freeze is 11 days away.  Any approve blueprint that does not have code proposed by then will not make the Havana.
21:06:35 <markmcclain> #topic Bugs
21:07:29 <markmcclain> Last week was a bit of a bumpy right getting items approved and merged through the gate.  We had several critical items pop up
21:07:45 <markmcclain> and solved most of the them.  The lone exception is this one:
21:07:45 <markmcclain> https://bugs.launchpad.net/neutron/+bug/1210664
21:07:48 <uvirtbot> Launchpad bug 1210664 in neutron "neutron can't setup basic network connnectivity in gate jobs" [Critical,Confirmed]
21:08:04 <markmcclain> Currently, tempest skips this test.  I know marun has been looking into it
21:08:35 <markmcclain> Another important bug changes the way the db schema is created:
21:08:36 <markmcclain> https://bugs.launchpad.net/devstack/+bug/1207402
21:08:37 <uvirtbot> Launchpad bug 1207402 in neutron "neutron should automatically stamp the database version when it is deployed" [High,In progress]
21:09:01 <salv-orlando> I am keeping the patches in WIP at the moment
21:09:25 <markmcclain> salv-orlando is working on this, so if there are problems switching from auto creation to running migrations like the other integrated projects please let the team know
21:09:37 <salv-orlando> I've heard from nati-ueno that removal of DB_ENGINE global breaks metaplugin, so I want to make sure no plugin is broken before really proposing theam
21:09:41 <salv-orlando> * them
21:10:01 <markmcclain> salv-orlando: makes sense
21:10:26 <salv-orlando> One more thing is that if you're not sure your plugin fully leverages migrations, this is the time to do so.
21:10:35 <salv-orlando> emagana already did this for the plum grid plugin
21:11:47 <emagana> if somebody is interested, this is the patch: https://review.openstack.org/#/c/41005/
21:13:16 <markmcclain> emagana: looks like you've been unlucky with merge conflicts.. hopefully the third time will be the last
21:13:28 <markmcclain> Any bugs the team needs to discuss?
21:13:50 <markmcclain> #topic Docs
21:14:04 <emagana> Nothing new about Docs, working on some cleaning
21:14:24 <markmcclain> We have 3 major categories of new docs we're tracking: VPN, Firewall and ML2
21:14:42 <emagana> nati_ueno and sumitNaiksatam working on FwaaS and VNPaaS
21:14:52 <markmcclain> rkukura: I assigned you to the ML2 doc bug, but feel free to reassign to someone on the ml2 subteam
21:14:58 <nati_ueno> hi! sorry late
21:15:03 <rkukura> markmcclain: ok
21:15:35 <salv-orlando> nati_ueno: np. We've assigned you 1 bug for each minute you've been late :p
21:16:36 <nati_ueno> salv-orlando: ha ha
21:17:14 <markmcclain> Any other doc items?
21:17:24 <emagana> nothing from my side!
21:17:51 <SumitNaiksatam> markmcclain: for fwaas, i would prefer to put the admin patch after the driver patch gets approved
21:18:16 <markmcclain> SumitNaiksatam: ok
21:18:20 <SumitNaiksatam> similarly for the API patch, i would prefer to get confirmation on the commit operation
21:18:57 <SumitNaiksatam> driver patch I believe is very close - we can discuss in the FWaaS section
21:19:13 <markmcclain> ok.. I wanted to discuss commit during the FWaaS report
21:19:28 <markmcclain> #topic Stable
21:19:40 <markmcclain> moving up because garyk has a conflict at the bottom of the hour
21:19:55 <markmcclain> #link http://tarballs.openstack.org/neutron/quantum-2013.1.3.tar.gz
21:20:12 <markmcclain> garyk: Anything else to add?
21:21:11 <garyk> last week the stable grizzly branch was release
21:21:18 <markmcclain> Ok looks like he's left
21:21:32 <markmcclain> oops.. you're still around :)
21:21:39 <garyk> nothing else much to add. guess we need to start to work on the new backports now
21:21:55 <garyk> yeah, sorry i am preparing a list of the open nova reviews regarding neutron
21:22:09 <garyk> but nothing imporant to add regarding the stable. sorry
21:23:02 <pcm_> VPNaaS API doc is almost ready to review.
21:23:11 <markmcclain> Yeah.. the week after a stable release is always quiet
21:23:20 <markmcclain> #topic Nova
21:23:39 <garyk> there are quite a lot of open reviews - would be nice if other can also jump in:
21:23:54 <garyk> https://review.openstack.org/#/c/31061/
21:24:19 <garyk> https://review.openstack.org/#/c/38803/ - i think that this will not work as the default device_id is ''
21:24:22 <dkehn> yep
21:24:31 <garyk> https://review.openstack.org/#/c/36138/
21:24:46 <garyk> and last but not least: https://review.openstack.org/#/c/40561/
21:25:46 <garyk> if others can take a look it would be great. thanks
21:26:32 <markmcclain> Thanks for sharing these
21:27:30 <markmcclain> Any other nova questions?
21:27:54 <dkehn> wondering if I can get some help with test_neutronv2.py
21:28:23 <garyk> dkehn: sure. i can help you offline if you wnat
21:28:36 <dkehn> garyk, that would be great
21:29:04 <markmcclain> #topic API
21:29:11 <markmcclain> salv-orlando: hi
21:29:12 <salv-orlando> hellp
21:29:23 <salv-orlando> hello - nothing major to report.
21:29:26 <garyk> i am going to call it a night. sorry guys.
21:29:37 <markmcclain> oh I thought things were melting down on the API side :)
21:29:52 <markmcclain> garyk: bye
21:29:53 <salv-orlando> is there something I don't know?
21:30:02 <salv-orlando> I am melting down, but this is an unrelated matter
21:30:14 <salv-orlando> or perhaps something I forgot because of the late hour
21:30:17 <markmcclain> haha.. I thought there was something I didn't know
21:30:38 <markmcclain> #topic VPN
21:30:40 <salv-orlando> no everything is quiet luckily
21:31:20 <markmcclain> yes.. always nice when the API is quiet at the end of h3
21:31:37 <nati_ueno> OK ipsec driver is still in review.
21:31:59 <nati_ueno> In this week, I posed status update code https://review.openstack.org/#/c/40993/ (WIP)
21:32:06 <markmcclain> ok.. I see if was updated Friday.. I'll checkout the new code
21:32:18 <nati_ueno> I'm also working on supporting service type framework
21:32:20 <nati_ueno> markmcclain: Thanks
21:32:29 <nati_ueno> client patch is still in review
21:32:34 <nati_ueno> That's all from me
21:32:55 <markmcclain> nati_ueno:  thanks for updating
21:33:07 <markmcclain> #topic LBaaS
21:33:25 <enikanorov> hi. nothing much changed from last week
21:33:39 <markmcclain> looks like 2 patches in review at the moment
21:33:43 <enikanorov> support for service type framework review got +2 from nati_ueno
21:34:41 <enikanorov> this one is to be reviewed https://review.openstack.org/#/c/40381/
21:35:20 <enikanorov> today I've posted https://review.openstack.org/#/c/41396/ as a first step towards more flexible vip-pool relationship
21:35:48 <enikanorov> probably that's all significant code for lbaas in H-3
21:36:00 <enikanorov> and that's all from my side for today
21:36:05 <markmcclain> great thanks for getting it in early!
21:36:19 <markmcclain> #topic FWaaS
21:36:28 <SumitNaiksatam> hi
21:36:33 <SumitNaiksatam> Agent patch was merged today, thanks to SridarK for putting in work on this patch
21:36:42 <markmcclain> So the iptables driver is that is left correct?
21:36:44 <SumitNaiksatam> CLI patch was merged last week, thanks to KC
21:36:56 <SumitNaiksatam> markmcclain: yeah, driver is left
21:37:06 <SumitNaiksatam> https://review.openstack.org/#/c/34074
21:37:14 <SumitNaiksatam> its pretty close, already has Salvatore's blessing (and Armando's +1)
21:37:23 <SumitNaiksatam> anyone else has any thoughts/objections on this patch?
21:37:33 <SumitNaiksatam> RajeshMohan is here
21:37:34 <armax> I'm btw
21:37:38 <armax> sorry I'm late :)
21:37:51 <RajeshMohan> I am here Sumit
21:37:55 <nati_ueno> SumitNaiksatam: I'll re-review today
21:38:02 <SumitNaiksatam> I believe RajeshMohan will be doing a rebase shortly
21:38:06 <SumitNaiksatam> nati_ueno: thanks
21:38:07 <salv-orlando> I think Armando's +1 should be read as a -1 with the caveat that you can improve exception handling in another patch
21:38:27 <RajeshMohan> My rebase will be in soon
21:38:50 <armax> yeah…I think if we can improve the exception handling I'd be a happy bunny :)
21:39:09 <SumitNaiksatam> armax, salv-orlando: it will be good to point out what exactly is expected
21:39:44 <salv-orlando> I really don't know that :) Perhaps RajeshMohan has a better idea
21:39:46 <SumitNaiksatam> i think the current model is preserving the abstraction between the driver and the agent
21:40:18 <RajeshMohan> I can check with Armando to understand what he expects
21:40:19 <markmcclain> As currently written I'm −1
21:40:33 <armax> yeah…let's take this offline
21:40:45 <armax> I
21:40:45 <markmcclain> LOG.error and then converting the exception to a string loses lots of context
21:40:54 <RajeshMohan> Since driver will be vendor specific, we did not want to pass those exceptions to agent
21:40:55 <armax> am around after the call if
21:41:02 <armax> I can be of help
21:41:05 <markmcclain> I'll add my comments to the review
21:41:17 <SumitNaiksatam> The horizon patch is complete but is waiting on Akihiro (who is on vacation) and also needs one more Horizon core
21:41:33 <SumitNaiksatam> I pushed the patch for the "explicit commit operation" yesterday night - https://review.openstack.org/#/c/41353/
21:41:49 <SumitNaiksatam> this has a little more work to do on the plugin side (which I should be able to complete by today), but from an API perspective it has everything to give you a feel for what we are doing, would appreciate your feedback
21:42:00 <SumitNaiksatam> Aaron, I did not see your email earlier, but I have responded to it a couple of hours back
21:42:28 <salv-orlando> I think we mentioned the ability of doing a bulk create several times doing the API patch review.
21:42:48 <salv-orlando> Sumit, can you refresh my mind on why we did not follow that route but preferred a two step process?
21:43:09 <SumitNaiksatam> salv-orlando: bulk create is different from the requirement for commit
21:43:36 <SumitNaiksatam> salv-orlando: i also don't recall discuss this to the extent of several times
21:44:05 <salv-orlando> sorry several times for me is >1
21:44:49 <SumitNaiksatam> the two step process is the approach we have followed from day one (even while discussing the in the summit) to allow policy auditing
21:45:02 <SumitNaiksatam> bulk operations are kind of complementary to this discussion
21:45:44 <SumitNaiksatam> but different
21:46:42 <salv-orlando> ok, perhaps the discussion on bulk operations derails from the focus of the topic
21:46:59 <markmcclain> I'll need to look at the implementation
21:47:09 <SumitNaiksatam> salv-orlando: yeah, i think we can have bulk operations regardless but those will be on the rules/policies
21:47:25 <markmcclain> but I had a few concerns over agent restarts and which version of the policy actually gets applied
21:47:32 <salv-orlando> I don't know if anybody has an opinion on committing rules with an explicit action. As already said, I am not a big fan of it, but since I understand the need for it, and I don't have a better alternative, I will accept it.
21:47:45 <salv-orlando> Or I will reserve the right to look at the implementation as markmcclain said
21:47:59 <SumitNaiksatam> markmcclain: that part will be handled in the implementation
21:48:18 <nati_ueno> I have concern about consistency of API model.
21:48:33 <SumitNaiksatam> markmcclain, salv-orlando: essentially we will follow the current model of pending states
21:48:39 <nati_ueno> if all resources will have commit mode, it sounds big change..
21:48:52 <salv-orlando> yup, but i guess you need to save 'committed' and 'desired' state for each rule
21:48:52 <nati_ueno> because current implemenation looks have additional pending change table
21:49:12 <SumitNaiksatam> nati_ueno: why should all resources have commit mode?
21:49:39 <nati_ueno> SumitNaiksatam: because needs for commit operation is generic one
21:49:41 <SumitNaiksatam> for other resources policy is not separate from their own state
21:50:02 <SumitNaiksatam> this is kind of different for the firewall model
21:50:22 <markmcclain> Ok looks like there is more to discuss, so let's look at the code: https://review.openstack.org/#/c/41353/
21:50:32 <nati_ueno> gotcha
21:50:35 <SumitNaiksatam> markmcclain: sure
21:50:40 <markmcclain> and then follow-up on the ML for concerns about the big picture
21:50:54 <markmcclain> and in the review if there are code specific items
21:51:01 <dflorea> I'm also not clear why all resources would have a commit mode. I think only firewalls require a commit to allow policies to be applied to firewalls one firewall at a time.
21:51:30 <SumitNaiksatam> dflorea: thats correct
21:51:47 <nati_ueno> SumitNaiksatam: dflorea: I'll post my comment in the mailing list, so let's continue
21:51:48 <SumitNaiksatam> we have a model where one policy can be associated with multiple firewalls
21:52:01 <nati_ueno> SumitNaiksatam: dflorea: I'll post my comment in the mailing list, so let's continue on there
21:52:09 <SumitNaiksatam> current model of implicit commit does not lend well to that
21:52:17 <markmcclain> yeah.. we're running short on time
21:52:27 <markmcclain> Any other FWaaS items/
21:52:28 <markmcclain> ?
21:52:31 <SumitNaiksatam> yeah, sure i am done with the update on the fwwas
21:52:39 <SumitNaiksatam> thanks for all the reviews
21:52:39 <markmcclain> #topic ML2
21:52:44 <markmcclain> rkukura: hi
21:52:49 <rkukura> hi
21:52:54 <rkukura> nothing major to report
21:53:05 <rkukura> devstack support for ml2 has been improved
21:53:12 <markmcclain> was going to ask about it
21:53:32 <rkukura> various mechanism drivers and portbinding BPs under development
21:53:54 <rkukura> once the gating is stable, should we think about adding ml2 to the gating?
21:54:41 <markmcclain> right now it is the default config which is OVS
21:55:02 <markmcclain> so I think the next logical step for deprecation is making the ml2+ovs the default in devstack
21:55:20 <rkukura> should we start gating on it before switching default?
21:55:54 <markmcclain> I think that would make sense to be the least disruptive
21:56:08 <markmcclain> we can talk offline the CI folks for their feedback on that plan
21:56:11 <nati_ueno> rkukura: That's possible, so may be we need to discuss with it with clarkb or jeblair
21:56:18 <rkukura> OK
21:56:23 <markmcclain> Anything else?
21:56:34 <rkukura> not that I can think of - anyone else?
21:56:54 <markmcclain> #topic Horizon
21:57:25 <markmcclain> amotoki is traveling this week, but has updated Horizon report on the agenda
21:57:33 <markmcclain> #topic Open Discussion
21:57:42 <pcm_> I have a quick question... For VPNaaS API doc review, should I just do std gerrit review (diffs are hard to review), or also create a PDF for early review rounds?
21:57:44 <ivar-lazzaro> I've updated the embrane's plugin patch, it'll appreciate some love :) : https://review.openstack.org/#/c/38222/
21:58:20 <nati_ueno> pcm_: review is enough :)
21:58:26 <hajay> I wanted to schedule a discussion around the policy and ipam blueprint we submitted at juniper-plugin-with-extensions
21:58:32 <jackmccann> for open discussion: would be nice to see this in H-3: https://blueprints.launchpad.net/neutron/+spec/multi-workers-for-api-server
21:58:33 <salv-orlando> pcm_: we can build the pdfs from the patch under review
21:58:35 <markmcclain> pcm_: standard gerrit review is fine
21:58:38 <pcm_> nati_ueno: Just wondering what's easies to read.
21:58:42 <pcm_> ok
21:58:47 <hajay> would a discussion in ML be better or do we need an IRC meeting to discuss?
21:58:58 <nati_ueno> pcm_: yeah, generating pdf is quite easy if it is on review
21:59:13 <nati_ueno> pcm_: but commenting is hard in the pdf
21:59:28 <pcm_> nati_ueno: roger that. Will Gerrit
21:59:47 <salv-orlando> jakemccann: I would be happy to assist with that review, but I would like to hear from somebody else from the community whether it's ok to have this kind of change in the last milestone of the release cycle
21:59:55 <jackmccann> multi-workers-for-api-server has code up for review
22:00:31 <salv-orlando> jackmccann: see above (sorry I mistype your nick)
22:00:48 <jackmccann> salv-orlando: thanks, would be nice to have thoughts from core on which direction to take
22:00:54 <markmcclain> I see the benefit.. my biggest concern at this point is stability
22:01:06 <salv-orlando> I think the direction is fine. My concern is stability.
22:01:07 <markmcclain> and the amount of time left to test
22:01:09 <salv-orlando> Sorry mark
22:01:20 <jackmccann> https://review.openstack.org/#/c/37131 based on glance (we've been running since May)
22:01:35 <nati_ueno> jackmccann: is that func can be optional function by conf
22:01:36 <jackmccann> https://review.openstack.org/#/c/36487/ using openstack/common/service.py
22:02:01 <jackmccann> yes, it is optional, set number of workers to zero in conf (would be default, no change in current behavior)
22:02:43 <jackmccann> salv-orlando will follow-up with you folks offline
22:02:54 <markmcclain> jackmccann: sounds good
22:03:17 <nati_ueno> salv-orlando: I'm +1 for this bp in H3, I'm also happy to assist the review
22:03:29 <markmcclain> Any other items for discussion?
22:05:36 <markmcclain> Ok.. We've got 11 days left before the freeze and other projects have deadlines around the same time next week
22:06:10 <markmcclain> so expect that the time the CI systems to evaluate could be longer if there are lots of pending reviews (which always happens during H3)
22:06:26 <markmcclain> Have a good week
22:06:29 <markmcclain> #endmeeting