14:01:58 <enikanorov_> #startmeeting neutron lbaas 14:01:59 <openstack> Meeting started Thu Feb 20 14:01:58 2014 UTC and is due to finish in 60 minutes. The chair is enikanorov_. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:02:00 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:02:02 <openstack> The meeting name has been set to 'neutron_lbaas' 14:02:12 <xuhanp> hello 14:02:20 <enikanorov_> today we're going to finalize what has been discussed recently over ML 14:02:39 <obondarev> great 14:02:44 <sbalukoff> Yay! 14:02:48 <enikanorov_> i've just updated the wiki page on the discussion 14:02:50 <enikanorov_> #link https://wiki.openstack.org/wiki/Neutron/LBaaS/LoadbalancerInstance/Discussion 14:02:58 <enikanorov_> thanks sbalukoff for cool pictures 14:11:40 <enikanorov_> so far we agreed to go with model number 3 14:11:41 <enikanorov_> so I'll start with basic terminology to set things clear 14:11:42 <enikanorov_> main configuration objects in model #3 that are changed in comparison to existing model are VIP, Pool, Listener 14:11:44 <enikanorov_> I'm not talking right now about instance/loadbalancer, we'll discuss that later 14:11:44 <enikanorov_> so one of the concerns that was discussed the most is that we are mixing implementation detail with public API 14:11:54 <enikanorov_> and impl details are, in particular, a deployment scheme of logical config to a backend 14:11:54 <enikanorov_> so, VIP in new model #3 is the new root object 14:11:54 <enikanorov_> so any kind of bindings, and association which users does or doesn't controll are attached to VIP 14:11:54 <enikanorov_> Listener represent a listening socket on the same address but possibly with different port/protocol 14:11:55 <enikanorov_> Pool mostly remains the same, however it looses its 'root' role 14:11:55 <enikanorov_> To describe the essense of the change, let me provide you a workflow example 14:15:30 <enikanorov_> so tenant wants to create 2 pools and 2 listeners, 1listener will point to 1 pool (default) ,and another lsitener will point to 1st pol as a default and to another pool via l7 rules 14:15:30 <enikanorov_> the sequence canl be the following: 14:15:30 <enikanorov_> 1) create pool1 14:15:30 <enikanorov_> 2) craete pool2 14:15:30 <enikanorov_> pools creation doesn't effect in scheduling and backend consumption 14:15:30 <enikanorov_> 3) create VIP 14:15:30 <enikanorov_> 4) create listener 1 for VIP 14:15:30 <enikanorov_> 5) create listener 2 for VIP 14:15:30 <enikanorov_> 6) set pool1 as default pool for listener1 14:15:30 <enikanorov_> 7) set pool1 as default pool for listener2 14:15:30 <enikanorov_> 8) set pool2 as L7 pool for listener2 14:15:30 <enikanorov_> some of those steps can be combined in one 14:15:30 <sbalukoff> No problem! 14:15:49 <obondarev> sbalukoff: loadbalancer and cluster are HA related in your proposal, right? 14:15:51 <sbalukoff> Yes. 14:15:56 <obondarev> sbalukoff: I think HA may be added to the model at any moment and that can be done in a backward compatible way 14:15:58 <vjay> enikanorov: It might be trivial, but could we have quick recap, of why new object model. 14:15:58 <vjay> It started with having pools in the same device for L7 config. 14:16:00 <vjay> Moved to EndPoint IPs / Virtual IPs have to be hosted in same device and the drive was proposed to make IP the focal point for placement. 14:16:00 <sbalukoff> obondarev: By delegating HA to the driver? 14:16:01 <obondarev> sbalukoff: no 14:16:04 <obondarev> sbalukoff: not necessarily 14:16:04 <sbalukoff> obondarev: Could you please explain how? Would adding HA require a model change? 14:16:15 <obondarev> sbalukoff: it might require, but that would be backward compatible additions to the model 14:16:16 <obondarev> sbalukoff: so, #3 picture seems enough to me at the moment 14:16:16 <obondarev> sbalukoff: as we can add HA later 14:16:16 <edhall> what prevents us from adding it to the model now? 14:16:16 <obondarev> sbalukoff: as we can add HA laterI know you have concerns that #3 model doesn't handle HA 14:16:17 <iwamoto_> how do you maintain backward-compat with #3? 14:16:17 <obondarev> iwamoto_: it can be done 14:16:18 <enikanorov_> when you attach the pool to a listener, it starts to belong to a backend, and cant be shared by another backend (that's the API limitation we'd like to preserve) 14:16:18 <enikanorov_> so one of the discussion points was around the ability of sharing pools between VIPs (which may represent different backends) 14:16:18 <enikanorov_> vjay: right 14:16:18 <enikanorov_> let's not interleave threads; we can discuss HA later 14:16:18 <enikanorov_> so, are there questions on object model? 14:16:18 <enikanorov_> about vip-pool-listener relationships? 14:16:26 <enikanorov_> network is stormy today 14:16:36 <markmcclain> yeah.. wondering if we want to defer 14:16:43 <markmcclain> I'm only getting half of the conversation 14:16:47 <enikanorov_> iwamoto_: mostly the visible API change is around 'provider' attribute 14:16:55 <markmcclain> and the eavesdrop logs seem incomplete 14:16:59 <enikanorov_> markmcclain: you're getting in and out 14:17:36 <enikanorov_> honestly i'd be glad to defer, because i'm not feeling well. 14:18:04 <sbalukoff> I'm OK with deferring, eh. People haven't had much time to think of implications of new model anyway. 14:18:09 <sballe> enikanorov, sorry to hear that you are not doing well 14:18:16 <iwamoto_> enikanorov: backward-compat API doesn't have listener. how can listener be handled? 14:18:30 <markmcclain> enikanorov: sorry you're sick 14:18:31 <iwamoto_> +1 for deferring 14:18:50 <enikanorov_> thanks, no prob folks :) 14:18:57 <enikanorov_> ok, let me find the slot 14:19:10 <sbalukoff> Feel better, dude! 14:19:29 <markmcclain> we can just meeting next week at the scheduled time 14:19:34 <markmcclain> s/meeting/meet/ 14:19:44 <sbalukoff> I'm OK with that. 14:19:59 <enikanorov_> hm, ok then. it will give more time for everyone to dig through our large poems in ML 14:20:04 <vjay> take care enikanorov 14:20:07 <sballe> me too. I need exra time to catch-up. 14:20:28 <enikanorov_> ok. thanks every one for joining. lets end the meeting 14:20:34 <s3wong> Thanks 14:20:35 <sballe> bye 14:20:35 <iwamoto_> enikanorov: hope you get well soon 14:20:46 <enikanorov_> thanks! 14:20:46 <avishayb_> bye 14:20:46 <vjay> bye 14:20:47 <enikanorov_> #endmeeting