20:00:31 <sbalukoff> #startmeeting Octavia 20:00:31 <openstack> Meeting started Wed Sep 17 20:00:31 2014 UTC and is due to finish in 60 minutes. The chair is sbalukoff. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:00:32 <blogan> 100 is fine for that as well 20:00:32 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:00:34 <openstack> The meeting name has been set to 'octavia' 20:00:34 <TrevorV_> o/ 20:00:37 <dougwig> blogan: 100 does that with widescreen monitors. 20:00:41 <sbalukoff> Ok, folks! 20:00:42 <dlundquist> o/ 20:00:47 <sbalukoff> Potentially short agenda for today. 20:00:48 <blogan> well what if i want 3 editors side by side? 20:00:55 <dougwig> 2 monitors? 20:00:59 <johnsom_> Hello 20:01:00 <blogan> then i could have 6! 20:01:00 <dlundquist> blogan: 4k display? 20:01:11 <sbalukoff> As usual, agenda is here: 20:01:13 <blogan> now i can have a lot of editors! 20:01:14 <sbalukoff> #link https://wiki.openstack.org/wiki/Octavia/Weekly_Meeting_Agenda#Agenda 20:01:14 <xgerman> o/ 20:01:17 <ajmiller_> o/ 20:01:23 <blogan> 10 char line limit! 20:01:27 <blogan> oh hi 20:01:28 <dougwig> alright,let's settle down for our chair here. :) 20:01:29 <blogan> meeting started 20:01:35 <TrevorV_> #action TrevorV write up 2 weeks worth of meeting notes. 20:01:43 <TrevorV_> o_0 20:01:46 <sbalukoff> #topic Review progress on gerrit reviews and blueprints 20:02:23 <sbalukoff> I feel like we're getting good progress on gerrit reviews, but that only a handful of us are doing said reviews presently. 20:02:47 <sbalukoff> Also, I apologize: I was too distracted with other priorities to actually update any of the blueprints in launchpad this last week. 20:02:55 <sbalukoff> I will be doing so this week. 20:03:20 <davidlenwell> o/ 20:03:47 <davidlenwell> I will start in on helping with the reviews also 20:03:52 <sbalukoff> Question I have for you, especially those looking to get involved: Is there something we can do to help you get started in particular? 20:04:00 <sbalukoff> Thanks, david 20:04:01 <TrevorV_> On this topic, I'd like to draw attention to this review: https://review.openstack.org/#/c/116718/ 20:04:21 <sbalukoff> (And by "we" I mean "those of us who have been working on LBaaS and Neutron LBaaS for months.) 20:04:22 <TrevorV_> It was dependent on the migrations, and since that's merged it would seem prudent to review this one next. 20:04:27 <davidlenwell> sbalukoff feel free to tag me in gerrit on reviews 20:04:31 <xgerman> we should probably start an etherpad with links to what we like eyeballs on 20:04:36 <sbalukoff> davidlenwell: Will do! 20:04:43 <xgerman> that helped me a lot when we did LBaaS v2 20:04:52 <sbalukoff> #action sbalukoff to assign all review work to davidlenwell 20:04:55 <sbalukoff> #undo 20:04:56 <openstack> Removing item from minutes: <ircmeeting.items.Action object at 0x1f4e310> 20:04:57 <johnsom_> +1 on etherpad for reviews 20:05:15 <blogan> can we have an etherpad listing out all the etherpads as well? 20:05:18 <blogan> j/k 20:05:28 <xgerman> that would be the wiki 20:05:32 <xgerman> :-) 20:05:33 <blogan> oh snap 20:05:38 <TrevorV_> actually blogan that might be helpful, since they don't explicitly show an organizational structure 20:05:48 <dougwig> i can setup an etherpad again. 20:05:52 <sbalukoff> xgerman: Sounds good. Question for you, as well: Is this link helpful for knowing what is in the review queue? 20:05:53 <sbalukoff> #link https://review.openstack.org/#/q/stackforge/octavia+status:open,n,z 20:05:57 <blogan> well 20:06:04 <dougwig> #action dougwig octavia review etherpad 20:06:06 <blogan> yeah sbalukoff 20:06:10 <blogan> thats what iw as going to link 20:06:11 <dougwig> i use that link, but it doesn't prioritize. 20:06:12 <davidlenwell> sbalukoff: maybe make that the irc chanels topic 20:06:14 <blogan> all the reviews are right there 20:06:38 <sbalukoff> davidlenwell: Good idea! 20:06:39 <xgerman> well, I still need to figure out what is WIP 20:06:49 <blogan> that has teh WIP status 20:07:23 <johnsom_> I will put that link on our wiki page for easy reference 20:07:27 <blogan> if there's a big X under W, that means its a WIP 20:07:36 <dougwig> johnsom_: it's already there 20:07:38 <dougwig> at the top 20:07:41 <dougwig> :) 20:07:49 <xgerman> make it bold and blinking 20:07:53 * Vorrtex__ power is fluctuating at random... might not be on here consistently 20:07:54 <johnsom_> So it is, cool, missed that 20:07:57 <sbalukoff> #action sbalukoff to try to update channel topic (even though we don't have ops here) 20:08:41 <dougwig> i can do that. what topic? 20:08:47 <blogan> also I thought a good reason to put WIPs in gerrit was so people could look at the direction the code is going and comment on it 20:08:49 <sbalukoff> dougwig: https://review.openstack.org/#/q/stackforge/octavia+status:open,n,z 20:08:54 <blogan> not wait for it to get out of WIP and ready for review 20:09:08 <dougwig> will set after meeting 20:09:15 <dougwig> #action dougwig fix lbaas channel topic 20:09:30 <sbalukoff> blogan: Yeah, I've sort of been doing the latter. I'll be more dilligent about reviewing WIP code. 20:09:41 <blogan> well thats what I thought, I could be wrong 20:09:43 <xgerman> blogan, you are right but I want to make sure I at least review aht's urgent first :-) 20:10:18 <dougwig> there are multiple kinds of reviews. the ones where you look for errors, omissions, or not being openstack-y, are kinda useless while WIP. the ones where you want to give design feedback, those are when you go into a WIP. IMO. 20:10:18 <blogan> xgerman: totally understand, I just wanted to make sure my understanding was correct, not saying any particular WIP should be reviewed right now 20:10:31 <blogan> dougwig +1 20:10:41 <sbalukoff> In any case, I can certainly put together an etherpad for people to update if they don't think the automatic listing is helpful (since it's not prioritized). We can see how that goes and decide whether it's worth maintaining long-term. 20:10:43 <xgerman> +1 20:10:44 <blogan> those are the most useful reviews in any phase of the review process 20:11:30 <Vorrtex__> Forgive me, did we get a page set up with links to reviews? etherpad or wiki? either? 20:12:17 <sbalukoff> Vorrtex__: There's this automated listing here: https://review.openstack.org/#/q/stackforge/octavia+status:open,n,z 20:12:24 <sbalukoff> But it is not prioritized by urgency 20:12:37 <Vorrtex__> Oh, ha, I've seen this page in passing, but never paid any attention to it. 20:12:39 <Vorrtex__> Thanks 20:12:42 <sbalukoff> Anyway, I'll go ahead and set up that etherpad. 20:12:57 <sbalukoff> #action sbalukoff to create etherpad listing reviews that need attention in order of urgency. 20:13:08 <blogan> we can set priorities in the launchpad blueprint page, but it's not easy to get to the reviews from there 20:13:18 <sbalukoff> blogan: I agree 20:13:25 <xgerman> +1 etherpad 20:13:54 <sbalukoff> I don't know about y'all but I do find this stand-up etherpad useful as well: https://etherpad.openstack.org/p/octavia-weekly-standup 20:14:02 <sbalukoff> But I notice not everyone updated it this week. 20:14:07 <blogan> ah crap I forgot to update taht 20:14:09 <blogan> sorry 20:14:12 <sbalukoff> Would it be useful for me to send a reminder to the mailing list? 20:14:35 <blogan> yeah 20:14:36 <sbalukoff> (I had been doing that prior to each meeting, but figured people might find it tiresome.) 20:14:38 <blogan> you should just utomate it 20:14:55 <xgerman> blogan, you can also use your won Outlook 20:14:59 <sbalukoff> blogan: Yeah, easily done. The question is: Do people mind? 20:15:19 <sbalukoff> dougwig: Should the bots have operator / voice status here? 20:15:19 <blogan> xgerman: yes i could do that, but that requires me to you know, do something 20:15:32 <xgerman> fair... 20:15:33 <sbalukoff> blogan: Want me to send you a calendar invite? 20:15:38 <dougwig> sbalukoff: they already have those rights, they just don't sit with them active. 20:15:40 <blogan> lol im being dumb 20:15:43 <sbalukoff> (I have one to remind me to update the agenda. XD ) 20:15:48 <blogan> no i can set my own, its fine 20:16:03 <blogan> im joking about me being helpless, i just wasn't thinking 20:16:21 <sbalukoff> Ok, so again, question: Is it useful for me to send something to the mailing list reminding people of the agenda, the meeting time and location, and the stand-up etherpad? 20:16:29 <ctracey_> hola folks 20:16:30 <sbalukoff> Because I absolutely could do that. :) 20:16:38 <sbalukoff> Howdy Craig! 20:16:40 <ctracey_> sbalukoff: yes 20:16:54 <blogan> i think it is useful, but once more and more people get involved it won't scale very well 20:17:00 <xgerman> yes, agenda to mailing list is good 20:17:05 <blogan> i mean the etherpad in general 20:17:05 <sbalukoff> Good enough! The rest of you will just have to put up with my spam. (as usual) 20:17:06 <blogan> not the ML 20:17:16 <ctracey_> well the agenda should be posted 20:17:18 <Vorrtex__> I did a general update for the weekly standup page, sorry about that sbalukoff. I'll draw my attention to that for before the future meetings 20:17:32 <blogan> yeah agenda should be posted, and the etherpad can be in that same post 20:17:44 <sbalukoff> #action sbalukoff to send weekly reminders to ML about agenda, meeting time + location, and stand-up etherpad 20:17:57 <sbalukoff> blogan: Yes, I intend to do this in a single e-mail. 20:17:59 <ctracey_> yes - a standing post in the meeting invite is fine. 20:18:36 <sbalukoff> Ok! 20:19:51 <sbalukoff> So, one thing I would like to see for next week is for everyone interested in contributing to have a look through the blueprints and either ask questions (here, in the ML, or in next week's IRC meeting) about what is unclear, what is to ambiguous, or anything else that's a subtle blocker for getting started helping. :) 20:20:16 <sbalukoff> I'm totally going to make an action item out of that. 20:20:22 <xgerman> I think once we hav more specs it might be easier for people to jump in 20:20:32 * dougwig thinks that sbalukoff just discovered the #action tag. 20:20:49 <xgerman> last week was vote-tag 20:20:58 <xgerman> now it's action - whoc knows what's next 20:20:59 <sbalukoff> #action everyone to look through blueprints, help flesh out and/or come to IRC, ML or meeting with questions. 20:21:11 <sbalukoff> Next week it'll all be about the #undo tag. 20:21:39 <sbalukoff> Ok, on this note, does anyone have anything else they'd like to ask about this topic before we switch to open discussion? 20:22:42 <johnsom_> crikets... 20:22:44 <sbalukoff> I'll take that as a 'no' 20:22:50 <sbalukoff> #topic Open Discussion 20:23:06 * Vorrtex__ likes these short and to-the-point meetings 20:23:25 <dougwig> question: are we going to move these meetings to the openstack meeting channels? 20:23:31 <sbalukoff> Anyone have anything they'd like to bring up before the group? (Otherwise, we might as well end early and let people get back to, you know, doing actual work.) 20:23:32 <blogan> good question 20:23:44 <blogan> we should move to the meeting channels 20:23:48 <blogan> i have no idea what that process is though 20:23:53 <sbalukoff> dougwig: I've yet to see whether there's an opening during this time slot. 20:23:59 <dougwig> i think you just reserve a slot in the wiki and go for it. 20:24:17 <sbalukoff> I'm happy to do that, so long as we can get this same slot (or something very near it). 20:24:53 <sbalukoff> (Mid week, and not forcing me to get up at 5:00am makes for a slightly less cranky sbalukoff) 20:25:11 <blogan> looks like openstack-meeting-3 would be available at this time 20:25:15 <sbalukoff> #action sbalukoff to look into / move Octavia meeting to a standard openstack meeting channel. 20:25:18 <blogan> just from a quick search 20:25:19 <dougwig> i'd suggest that we do so, unless there are time conflicts. but see as all of my other openstack meetings are at horrendous hours, the mid-day blocks should be free. :) 20:25:32 <Vorrtex__> dougwig: aint that the truth 20:25:40 <xgerman> or we start openstack-meeting4 20:25:42 <xgerman> ... 20:25:45 <sbalukoff> dougwig: That was my sardonic hope, as well. ;) 20:25:59 <sbalukoff> xgerman: Isn't that what #openstack-lbaas is? ;) 20:26:14 <sbalukoff> Ok, folks, anything else, or are we done for today? 20:26:26 <xgerman> then we need tolerate other projects doing meetings in our channel 20:26:33 <dougwig> i'm done. trevor, you got some -1 love. 20:26:38 <Vorrtex__> dougwig: thanks 20:26:40 <blogan> sbalukoff, xgerman: name of the controller driver interface to push teh configs along 20:26:50 <sbalukoff> blogan: Ooh! Good one. 20:26:55 <Vorrtex__> dougwig: if you reviewed the repository review then I know its broken 20:27:05 <blogan> is it a WIP? 20:27:20 <Vorrtex__> blogan: last I checked yeah 20:27:23 <sbalukoff> #topic "Discussion" about what to name class that is the controller<->driver interface. 20:27:36 <blogan> naming! yay! 20:27:45 <xgerman> driver means the driver which controls LBs on an Amphora 20:27:49 <dougwig> what's the name for a roman vomitorium? 20:27:51 <sbalukoff> I was really tempted to make the topic "Weekly holy war" 20:28:03 <dougwig> or a roman bottle opening? 20:28:08 <dougwig> jk. 20:28:13 <sbalukoff> dougwig: Haha! 20:28:29 <blogan> so the current name suggested is AmphoraeDriver, but that to me seounds like its responsible for spinning up and down Amphorae 20:28:49 <blogan> wherease SoftwareLoadBalancerDriver is more specific, but still generic enough (though the name is a bit dumb I know) 20:29:17 <dougwig> ControllerDriver. AmphoraConfigDriver. AmphoraMetaDriver. 20:29:22 <dougwig> (just throwing stuff out) 20:29:37 <sbalukoff> ControllerDriverInterface 20:29:46 <sbalukoff> That's the most literal term I can think of. 20:29:47 <Vorrtex__> while we're throwing stuff out, how about nuking the term amphora? 20:29:49 <Vorrtex__> lulz jp 20:29:51 <blogan> is there also going to be a driver interface that is responsible for amphora lifecycle mangement? 20:30:16 <sbalukoff> Vorrtex__: Don't make me send the phone spiders after you. 20:30:33 <sbalukoff> blogan: There needs to be something like that, yes. 20:30:43 <blogan> and that will live in the controller as well then? 20:31:00 <sbalukoff> Well, we've talked (briefly) about having an abstract interface to Nova 20:31:02 <blogan> so doesn't AmphoraeDriver seem more appropirate for that 20:31:09 <sbalukoff> It seems to me it would be there, and probably not in the thing German is working on. 20:31:26 <sbalukoff> blogan: Maybe AmphoraeManager 20:31:27 <xgerman> yeah, ironically blogan suggested that name 20:31:40 <blogan> xgerman: shhh 20:31:44 <blogan> no one needs to know that 20:31:49 <xgerman> lol 20:32:31 <dougwig> Manager gets overloaded a lot, between python context managers and openstack in general. 20:32:33 <blogan> AmphoraeManager is still the same problem 20:32:42 <blogan> noo dougwig, no more overloaded complaints 20:32:49 <xgerman> I guess we should pick a few and vote 20:32:49 <xgerman> exercise the #vote tag 20:33:05 <sbalukoff> I think some of the trepidation here is that there's probably some confusion about the responsibilities of each component. 20:33:20 <blogan> ill be fine with AmphoraDriver, i don't want to get into a long drawn out discussion and vote 20:33:29 <sbalukoff> blogan: +1 20:33:32 <sbalukoff> Ok, so! 20:33:39 <Vorrtex__> but driver is overloaded. 20:33:56 * sbalukoff sends the phone spiders after Vorrtex__ 20:34:08 * Vorrtex__ laughs as his friends return to his side 20:34:15 <sbalukoff> Ok, so! 20:34:19 <dlundquist> I think it would be easier if someone put forward a high level architecture with their best names and then we reviewed it, otherwise we can't decide if the name fits better somewhere else. 20:34:20 <xgerman> then lets name it chauffeur 20:34:21 <sbalukoff> Any other suggestions for a name here? 20:34:29 <sbalukoff> I'm about to compile a list and call a vote. 20:34:30 <Vorrtex__> xgerman: how about alfred, or jarvis. 20:34:34 <blogan> AmphoraLoadBalancerDriver 20:34:35 <sbalukoff> So we don't have to spend too much time on this. 20:34:48 <sballe_> blogan, +1 20:34:53 <xgerman> +1 20:35:34 <xgerman> AmphoraLoadBalancersDrivers to illustrate the M to N 20:35:57 <Vorrtex__> xgerman: I thought we went 1:M LB:amphora? 20:35:59 <blogan> wouldn't it be Amphorae? 20:36:23 <xgerman> correct and Vortex_ sadly that didn't get ratified 20:36:25 <blogan> anyway, sorry for brining up yet another naming issue 20:36:41 <Vorrtex__> So I should undo that change in the models then? 20:36:45 <blogan> i thought we agreed to go wtih 1:M LB:amphora at first 20:36:50 <Vorrtex__> Yeah, same 20:37:06 <xgerman> I thought dougwig threw a rench 20:37:09 <xgerman> wrench 20:37:25 <xgerman> and we left it M to N? 20:37:33 <blogan> well he thinks it should be left up to the drivers 20:37:36 <Vorrtex__> I didn't see that xgerman, but I could have missed it 20:37:59 <sbalukoff> #vote What should we call the class that is the controller-driver interface? AmphoraLoadBalancerDriver AmphoraDriver ControllerDriver AmphoraConfigDriver AmphoraMetaDriver ControllerDriverInterface AmphoraeManager 20:38:12 <sbalukoff> Let's see if the voting system barfs over that. 20:38:19 <blogan> is it active now? 20:38:28 <blogan> #vote AmphoraLoadBalancerDriver 20:38:31 <sbalukoff> Dammit no.. 20:38:34 <johnsom_> It doesn't look like it fired off the vote 20:38:35 <Vorrtex__> I thought it was "#start-vote" or something like that 20:38:36 <sbalukoff> Sorry... just a sec. 20:38:37 <blogan> start-vote 20:38:46 <sbalukoff> #startvote What should we call the class that is the controller-driver interface? AmphoraLoadBalancerDriver AmphoraDriver ControllerDriver AmphoraConfigDriver AmphoraMetaDriver ControllerDriverInterface AmphoraeManager 20:38:47 <openstack> Begin voting on: What should we call the class that is the controller-driver interface? Valid vote options are AmphoraLoadBalancerDriver, AmphoraDriver, ControllerDriver, AmphoraConfigDriver, AmphoraMetaDriver, ControllerDriverInterface, AmphoraeManager. 20:38:48 <openstack> Vote using '#vote OPTION'. Only your last vote counts. 20:38:51 <blogan> #vote AmphoraLoadBalancerDriver 20:38:53 <sbalukoff> Ok, NOW vote 20:39:00 <xgerman> #vote AmphoraLoadBalancerDriver 20:39:02 <sballe_> #vote AmphoraLoadBalancerDriver 20:39:02 <sbalukoff> #vote AmphoraLoadBalancerDriver 20:39:10 <sbalukoff> I sense a trend. 20:39:11 <blogan> lol didn't need a vote 20:39:22 <Vorrtex__> #vote AmphoraeManager 20:39:22 <sbalukoff> blogan: But now it'll be official. 20:39:23 <johnsom_> #vote AmphoraLoadBalancerDriver 20:39:26 <blogan> i think we just like the vote sript 20:39:38 * Vorrtex__ really likes using overloaded terms that still make sense 20:39:53 <jwarendt> #vote AmphoraLoadBalancerDriver 20:40:01 <ajmiller_> #vote AmphoraLoadBalancerDriver 20:40:09 <sbalukoff> 60 seconds until voting is ended... 20:41:09 <sbalukoff> #endvote 20:41:11 <openstack> Voted on "What should we call the class that is the controller-driver interface?" Results are 20:41:12 <openstack> AmphoraeManager (1): Vorrtex__ 20:41:13 <openstack> AmphoraLoadBalancerDriver (7): xgerman, jwarendt, sbalukoff, ajmiller_, johnsom_, blogan, sballe_ 20:41:18 <sbalukoff> Ok! 20:41:19 <sbalukoff> handy. 20:41:23 <xgerman> runaway victory 20:41:27 <sbalukoff> Anyone have anything else to bring before the group? 20:41:34 <dougwig> Lost network. :) 20:41:44 <Vorrtex__> dougwig: blogan spoke for you a few times 20:41:46 <Vorrtex__> just sayin 20:41:47 <blogan> sbalukoff: keeping the M:N table structure for LB:Amphora 20:41:58 <Vorrtex__> yeah, good call blogan 20:42:02 <xgerman> +1 20:42:25 <sbalukoff> #topic voting on M:N table structure for LB:Amphora 20:42:28 <blogan> if it is indeed left to the driver to decide, then we should just keep the M:N, but w can't put in unique constraints either to enforce 1:M 20:42:36 <sbalukoff> Anyone want to summarize the arguments for either side? 20:43:12 <sbalukoff> blogan: And therefore the code will have to deal with M:N, even if the driver uses 1:N 20:43:20 <blogan> yes 20:43:22 <johnsom_> sbalukoff you are good with definitions, should we review what an "LB" is and how it is different than a "Amphora" just so we are all on the same page? 20:43:24 <xgerman> sbalukoff and I agreed on 1:N 20:43:53 <sbalukoff> johnsom_: Ok, so "LB" is load balancer as it came to be understood in the Neutron LBaaS project... 20:43:57 <sbalukoff> Speaking of which... 20:43:57 <Vorrtex__> xgerman: blogan and I also agree on that 20:44:01 <blogan> xgerman: we agreed to do that at first, but we can still have the table structure set up to be M:N to not paint oursevles in a corner to not allow it in the future or other drivers to do it 20:44:29 <sbalukoff> #action sbalukoff to start dictionary / glossary of terms for Octavia project. 20:44:36 <xgerman> +1 20:44:37 <sbalukoff> I keep forgetting to do that. :P 20:44:42 <johnsom_> +1 20:44:44 <blogan> that way its up to the driver to decide whether its M:N or 1:M 20:44:52 <masteinhauser> sbalukoff: You defined like 20 on-the-fly when we were on the phone... 20:45:20 <sbalukoff> That's because I talk too quickly and interminably. :) 20:46:46 <blogan> anyone hve a strong opinion on whether we allow M:N LB:Amphora table structure (but still aim at 1:M LB:Amphora for the driver we actually implement)? 20:46:48 <sbalukoff> To further clarify what "LB" means: It's essentially the same thing as a "VIP" in other load balancing terminology (ie. everywhere outside of Neutron LBaaS), with the exception that a load balancer *might* have more than on IP address associated with it in the future. 20:47:46 <sbalukoff> blogan: So this question, to me, is more about whether we ever want to allow more than one load balancer per amphora. (ie. whether there are practical, technical, or business needs to allow for this.) 20:48:10 <sbalukoff> If we intend to allow 3rd party vendors to have more freedom in how they implement their solutions, we need M:N, IMO. 20:48:15 <blogan> sbalukoff: true, and if someone really needs it then we shouldn't not allow it 20:48:17 <masteinhauser> M:N seems to make sense in the case when you may be using LVS or other Direct Routing style load balancers. 20:48:24 <sbalukoff> But... I dunno. Maybe we don't want Octavia to allow that. :/ 20:48:45 <blogan> xgerman: thoughts? 20:49:16 <xgerman> well, my thought was for a software LB we can always adjust the size of the vm -- so if you need to LB's just spin up tow tiny vms 20:49:49 <xgerman> so one lb per amphora is sufficient and you tune with nova 20:50:29 <blogan> is there another case for M:N other than trying to save space/resources by putting many LBs/Listeners on an amphora? 20:50:42 <sbalukoff> I could imagine a 3rd party solution where a vendor makes a "big" load balancer appliance and allows "virtual load balancers" to be created on it in some fashion. This model doesn't actually break with 1:N, per se... 20:50:50 <xgerman> also since we have migrations how diffiuclt is it to geo from 1:N to M:N? 20:50:58 <sbalukoff> It's also a question of "do we really need to allow for colocation? 20:51:07 <sbalukoff> Apolocation is necessary to fulfill HA requirements, in any case. 20:51:18 <sbalukoff> But I'm having a hard time coming up with a solid case for LB colocation. 20:51:31 <Vorrtex__> sbalukoff: colocation I thought was being handled inside neutron or did I miss a conversation there as well o_0 20:51:52 <blogan> xgerman: it really shouldn't be too difficult, code will have to changed as well probably 20:51:59 <sbalukoff> xgerman: It's not just migrations, per se... it's also a bunch of places in the code where people, by that time, might have assumed 1:M and aren't prepared to deal with M:N 20:52:36 <xgerman> yeah, so the reason would need to be really compelling by then :-) 20:52:46 <sbalukoff> Vorrtex__: There was some talk of it being handled in Nova, IIRC... but we'll still need a logical representation in Octavia in any case, I think. 20:53:08 <xgerman> yeah, so the lifecycle driver can tell nova what to do 20:53:10 <sbalukoff> xgerman: Yes. And again, I'm having trouble coming up with a compelling justification. 20:53:13 <Vorrtex__> I see, thanks sbalukoff, I thought we had talked about it at some point there 20:53:14 <blogan> dougwig: do you have any thoughts on this? 20:53:19 <sbalukoff> Can anyone here think of a good reason why we would need colocation? 20:53:44 <xgerman> colocation of two LBs on the same vm -- NOT colocation of two vms conatining LBs on the same host 20:54:04 <sbalukoff> xgerman: Yes, exactly. Thanks for the clarification. 20:54:15 <blogan> sbalukoff: define colocation and apolocation in yoru gloassary too 20:54:26 <xgerman> +1 20:54:40 <blogan> i'm colocated with everyone right now, on earth 20:54:48 <sbalukoff> I can think of one potentially compelling reason not to allow colocation: It makes our system less flexible. 20:55:21 <sbalukoff> (Because users would then, effectively, be able to dictate where certain cloud resources get placed.) 20:55:31 <xgerman> it's always more difficult to take the right thing away then to add things 20:55:50 <sbalukoff> xgerman: Another compelling reason. 20:55:58 <sbalukoff> dougwig: Are you still here? 20:56:48 <sbalukoff> dougwig: I would like to get your perspective on this because I think you're probably the person most in favor of M:N here. 20:57:03 <sbalukoff> So, if dougwig has lost connectivity, I will forego the vote until next week. 20:57:11 <xgerman> but we need to know 20:57:41 <xgerman> well, I can assume the 1:N case in the interface 20:58:02 <sbalukoff> xgerman: Let's assume 1:N for now, then, unless dougwig can give us a compelling reason to do M:N that outweighs the two reasons we've come up with for not to allow colocation. 20:58:16 <sbalukoff> Ok! 20:58:22 <sbalukoff> We have about 2 minutes left. 20:58:22 <xgerman> Deal! 20:58:28 <sbalukoff> Anything else? 20:58:30 <xgerman> blogan? 20:58:44 <blogan> i have nothing else 20:58:52 <blogan> but yeah thats fine by me 20:59:03 <sbalukoff> Thanks for coming y'all! 20:59:16 <sballe> bye 20:59:18 <xgerman> bye 20:59:24 <sbalukoff> #endmeeting