15:04:17 <carl_baldwin> #startmeeting distributed_virtual_router 15:04:18 <openstack> Meeting started Wed Jul 2 15:04:17 2014 UTC and is due to finish in 60 minutes. The chair is carl_baldwin. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:04:19 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:04:21 <openstack> The meeting name has been set to 'distributed_virtual_router' 15:04:52 <carl_baldwin> #topic agenda 15:05:12 <carl_baldwin> What agenda items do y’all have for today? 15:05:45 <armax> I don’t have anything specific besides a usual status update 15:05:50 <mrsmith> +1 15:05:58 <carl_baldwin> Sounds good. 15:06:19 <carl_baldwin> This morning, all of the patches are passing Jenkins tests. That is great. 15:06:21 <Sam-I-Am> i'm here for the status updates 15:06:28 <Sam-I-Am> <- docs 15:07:24 <carl_baldwin> Sam-I-Am: thanks. You have status to give? 15:07:49 <Sam-I-Am> nope. but i am curious when/where i might be able to find some hints on how to deploy dvr so we can start on the docs 15:08:22 <carl_baldwin> Sam-I-Am: I see. The best we’ve got now in the howto page. 15:08:24 <carl_baldwin> #link https://wiki.openstack.org/wiki/Neutron/DVR/HowTo 15:08:41 <carl_baldwin> It is a bit raw at the moment but I’ve been working on it this week. 15:09:04 <carl_baldwin> #topic status 15:09:05 <armax> Sam-I-Am: we’ll be adding more info 15:09:10 <Sam-I-Am> carl_baldwin: cool. its the doc team's job to make it look pretty for the official docs 15:09:13 <phil_h> Sam-I-Am and I can help on that 15:09:17 <armax> so watch it closely :) 15:09:20 <Sam-I-Am> this stuff will probably go into the admin guide or a networking book 15:09:38 <armax> Sam-I-Am: also patches #link: https://review.openstack.org/#/q/status:open+project:openstack/neutron+branch:master+topic:bp/neutron-ovs-dvr,n,z 15:09:49 <armax> have DocImpact in their commit sections 15:10:06 <armax> to help a little more 15:10:07 <Sam-I-Am> glad to see #docimpact being used 15:10:15 <carl_baldwin> Sam-I-Am: We are planning to get together at the mid-cycle to discuss docs with emagana. 15:10:20 <Sam-I-Am> cool 15:10:25 <carl_baldwin> Will others be attending the mid-cycle next week? 15:10:59 <pcm_> yes 15:12:16 <mrsmith> carl_baldwin: I won't be able to attend 15:13:25 <carl_baldwin> Speaking of deploying the code. I’ve been making an attempt of my own this week. 15:14:02 <carl_baldwin> It resulting in discovering some missing code which armax recovered. Looks like the rebase to master left some behind. 15:14:10 <carl_baldwin> *resulted 15:14:11 <armax> armax: so have I 15:14:43 <mrsmith> carl_baldwin: there is the rpc out of sync issue as well as the config param change between the l3-agent and l3-scheduler 15:14:55 <armax> as we’re getting more and more into integration testing, more issues will arise and get addressed over time 15:15:18 <mrsmith> for rpc there is the missing handlers but from a review request, I bumped up the RPC version to 1.1 on the agent 15:15:24 <carl_baldwin> I attempted to convert an existing legacy router to a DVR router. I uncovered a few bugs that I’ll be noting this morning in the reviews. 15:15:56 <armax> mrsmith: is this reflected on the patch under review? 15:15:57 <carl_baldwin> mrsmith: I believe that we’ve restored the missing RPC handlers. I did not have any RPC problems with the latest code. 15:16:17 <mrsmith> ok 15:16:26 <carl_baldwin> There are many code paths that I’ve not touched yet though. 15:16:50 <mrsmith> we done alot of testing on icehouse 15:16:58 <mrsmith> juno has been new 15:17:06 <armax> mrsmith: as for the config param change between l3-agent and l3-sched what do you mean exactly? 15:17:27 <mrsmith> centralized -> centralized_router 15:17:41 <mrsmith> so the scheduler needs to change to follow 15:17:51 <armax> mrsmith: got it 15:18:04 <armax> mrsmith: we were also thinking of consolidating some of the config options if possible 15:18:09 <mrsmith> agreed 15:18:11 <armax> mrsmith: would you have time to look into that? 15:18:14 <mrsmith> but this is just a diff now 15:18:45 <mrsmith> consolidating the config options? we can discuss.... not sure what you had in mind 15:19:19 <armax> mrsmith: I posted my comments on the review 15:19:28 <mrsmith> k - I'll look 15:19:37 <carl_baldwin> mrsmith: Could you note in the review of the scheduler patch where the config option problem is? 15:19:48 <mrsmith> sure 15:19:52 <armax> as for centralized vs centralized_router, not sure I follow…maybe you can point out the discrepancies on the reviews in question? 15:19:54 <carl_baldwin> Thanks. 15:20:11 <armax> L3-sched and L3-agent look to be in sync, but I could see double :) 15:20:34 <mrsmith> fyi - I hope to push a new patch today for l3-agent 15:21:34 <carl_baldwin> Some of you may not be aware that I’ve started to maintain a branch on github that merges together all of the patches. 15:21:37 <carl_baldwin> #link https://github.com/ecbaldwin/neutron-dvr.git 15:22:27 <armax> carl_baldwin: this is used in this devstack patch #link https://review.openstack.org/#/c/103180/ 15:23:16 <carl_baldwin> armax: What is the status of this patch? Is it usable yet? 15:23:23 <armax> locally it should 15:23:34 <armax> I am still trying to figure out why the gate does not pick up the changes 15:23:40 <armax> more to follow 15:24:28 <carl_baldwin> I look forward to using it. It will be very useful. 15:25:05 <carl_baldwin> Is there any other status? 15:25:25 <armax> none from me 15:25:43 <mrsmith> none from me 15:25:57 <carl_baldwin> #topic mid-cycle meeting 15:26:29 <armax> the L3 HA effort seems to have lost momentum 15:26:59 <armax> carl_baldwin: I suggest that the dvr picks it up as soon as the bulk of the dvr stuff merges 15:27:00 <carl_baldwin> #undo 15:27:01 <openstack> Removing item from minutes: <ircmeeting.items.Topic object at 0x2aa5390> 15:27:12 <carl_baldwin> #topic HA 15:27:16 <armax> carl_baldwin: sorry :) 15:27:18 <carl_baldwin> armax: good point. 15:27:22 <carl_baldwin> No worries. 15:27:31 <armax> it just came up to my mind 15:27:49 <carl_baldwin> armax: That is a good suggestion. 15:27:53 <armax> as I was saying, I’d suggest that the dvr team picks it up as soon as the bulk of the dvr stuff merges 15:28:03 <armax> I spoke with mestery about this 15:28:12 <armax> and we agreed on a plan going forward 15:28:33 <Sam-I-Am> any plans to make l3 redundancy work with linuxbridge rather than ovs? 15:29:15 <armax> in the meantime I’ll be working on the current code being proposed 15:29:28 <armax> so that more people can work on it at the same itme 15:29:34 <armax> *time 15:29:54 <carl_baldwin> Sam-I-Am: Is the current HA code limited to ovs? 15:30:14 <phil_h> Is says so on the how to page 15:30:21 <Sam-I-Am> i was just reading through the howto and noticed it was specific to ovs and vxlan 15:30:31 <armax> rebase etc... 15:30:42 <armax> carl_baldwin: basically, reviving the code 15:31:01 <carl_baldwin> Sam-I-Am: Ah, I had shifted my thinking toward the L3 HA code. 15:31:16 <carl_baldwin> Yes, DVR is currently limited to ovs and vxlan. 15:31:21 <armax> is safchain around? 15:32:16 <Sam-I-Am> carl_baldwin: i guess my mind was also wondering about the diff between the l3 ha and dvr stuff 15:32:58 <armax> Sam-I-Am: I believe this is covered in the respective blueprints 15:34:03 <carl_baldwin> Sam-I-Am: They are fundamentally different. DVR is about distributing routers so that routing is done on the compute node. It does not add redundancy but rather decouples routing from the network host. 15:34:32 <safchain> armax, hi 15:34:35 <carl_baldwin> HA routing adds redundancy by running any centralized routing redundantly on multiple network hosts. 15:34:44 <Sam-I-Am> thx 15:34:52 <armax> hi safchain: we were talking about L3 HA 15:35:15 <carl_baldwin> Sam-I-Am: There is some opportunity to use to two together in the case of default SNAT which DVR does not handle in a distributed way. 15:35:28 <Sam-I-Am> ok 15:36:15 <armax> safchain: maybe we could sync up offline? 15:36:34 <safchain> armax, yes sure 15:37:57 <carl_baldwin> armax: safchain: I hope to see this revived and merged. I think it is long overdue. 15:38:26 <armax> carl_baldwin: will try to make this happen 15:38:36 <carl_baldwin> armax: great. 15:39:01 <carl_baldwin> #topic mid-cycle sprint 15:39:29 <carl_baldwin> I’d like to get some ideas for what we could be working on during the three day sprint coming up next week. 15:39:41 <carl_baldwin> We’ve mentioned a docs discussion. 15:40:50 <carl_baldwin> Other ideas? 15:41:54 <carl_baldwin> I think that I’ll use whatever time I can get to continue trying to get my deployment off the ground and rooting out problems with the basic use cases. 15:43:06 <carl_baldwin> What I might do in preparation is build a couple more VM-based two-node devstack clusters with the DVR code ready to test. 15:43:38 <carl_baldwin> Hopefully using the devstack patch armax is working on. 15:44:50 <carl_baldwin> If the basic use cases are working by then maybe I can flesh out the use case to convert a legacy router to distributed. 15:44:55 <carl_baldwin> Any other ideas? 15:45:17 <Sam-I-Am> that would be a good feature to help adoption 15:46:07 <mrsmith> armax: have you had success with vivek's example of multi-node devstack? 15:46:13 <mrsmith> we use it locally 15:46:15 <carl_baldwin> Sam-I-Am: Agreed. The feature is in the merged blueprint. Just needs some work. 15:46:21 <armax> carl_baldwin: not yet… 15:46:31 <armax> carl_baldwin: ironing out all the issues I found on a single node first 15:47:17 <carl_baldwin> armax: Understood. I’m just hoping. I’ll probably get to building a couple more clusters next week before flying out to Minn. 15:47:43 <armax> carl_baldwin: if there are guys from QA/infra 15:48:03 <armax> carl_baldwin: you could discuss steps to get this a non-voting experimental job 15:48:10 <armax> carl_baldwin: even though that shouldn’t be extra difficult 15:48:41 <carl_baldwin> armax: Good idea. Let me look at the list of attendees. 15:48:50 <carl_baldwin> armax: Do you know off-hand if any will be attending? 15:49:43 <armax> carl_baldwin: not sure 15:49:46 <armax> #link: https://etherpad.openstack.org/p/neutron-juno-mid-cycle-meeting 15:49:52 <armax> lists the attendes 15:50:07 <armax> but I am not sure about affiliations nor responsibilities of some of those names 15:50:43 <carl_baldwin> armax: I’m sure that Mark or Kyle will be able to help find the right people even if they’re not in attendance. 15:51:10 <carl_baldwin> mlavelle may be able to help as well. 15:51:13 <armax> carl_baldwin: ok 15:51:33 <carl_baldwin> I’ll be sure to start a discussion on it. 15:52:29 <carl_baldwin> #topic Open Discussion 15:52:44 <yisun> Carl, this is Yi from FWaaS 15:52:49 <Sam-I-Am> whats a good place to keep up with the progress here? 15:53:03 <carl_baldwin> yisun: Hi 15:53:15 <yisun> Have you get a chance to talk to Vivek? 15:53:21 <yisun> About the detail of his idea? 15:53:54 <viveknarasimhan> hi yisun 15:53:56 <carl_baldwin> Sam-I-Am: Not sure. I guess the meeting status reports here and in the Neutron meeting. What could we do to help? 15:54:10 <carl_baldwin> yisun: No, I did not get a chance. Thanks for the reminder. 15:54:36 <carl_baldwin> viveknarasimhan: I was going to ask you about your ideas for integrating FWaaS. 15:54:55 <viveknarasimhan> yes carl 15:55:02 <Sam-I-Am> carl_baldwin: i guess that works. what i'd like to see at some point is a recommended/tested configuration that we could put into official docs 15:55:08 <viveknarasimhan> i was proposing current behaviour where 15:55:33 <viveknarasimhan> in integration bridge we know whihc is DVR routed, which is traditionally routed and which is switched packet 15:55:50 <viveknarasimhan> if we could put a hook in there for DVR routed packets 15:56:27 <carl_baldwin> Sam-I-Am: Thanks. I will document what I test. It will take some time. 15:56:27 <viveknarasimhan> that is where FwAAs would handle the routed packet before forwarding it out/into of the DVR interface 15:57:34 <carl_baldwin> viveknarasimhan: I don’t see yet how that solves the problem of path asymmetry. 15:58:10 <viveknarasimhan> proposing a way to handle access control 15:58:22 <viveknarasimhan> i wasn't mentioning this will give a stateful firewall 15:58:32 <carl_baldwin> Ah, I see. 15:58:50 <carl_baldwin> I think there may have been some confusion there. 15:59:26 <carl_baldwin> I’m glad we got that cleared up. 15:59:48 <yisun> Vivek, got you 15:59:51 <carl_baldwin> So, the problem of running a stateful firewall with DVR E/W traffic is still an open issue. 16:00:00 <viveknarasimhan> yes , 16:00:02 <yisun> Carl, yes 16:00:15 <yisun> Vivek:Carl: FYI—-Currently, I’m thinking the FW zone concept may not be applicable in the distribute environment . The Zone was used to group interfaces and be used ask context for policy, with the DVR, there are no interface really. So, we will could provide another type of construct to be the context of policy objects. This way we could simplify our issue 16:00:29 <carl_baldwin> I just noticed that we hit the top of the hour. 16:00:34 <viveknarasimhan> for DVR we have only bindings 16:00:35 <yisun> s/ask/as/ 16:00:42 <carl_baldwin> We can take this discussion in to the neutron room. 16:00:50 <carl_baldwin> #endmeeting.