14:01:23 <markmcclain> #startmeeting lbaas
14:01:24 <openstack> Meeting started Thu May 16 14:01:23 2013 UTC.  The chair is markmcclain. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:01:26 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:01:28 <openstack> The meeting name has been set to 'lbaas'
14:02:13 <enikanorov-w> does everyone agree with agendai've just sent?
14:02:31 <ogelbukh> link it, enikanorov-w
14:02:47 <markmcclain> I got a copy but I liked to cover the items in a slightly different order
14:02:48 <enikanorov-w> ogelbukh: that was in email to primary audience, sorry :)
14:02:55 <enikanorov-w> markmcclain: sure
14:03:04 <SamuelB> Sure
14:03:18 <SamuelB> 1) LBaaS development process  2) multi-vendor-support-for-lbaas blueprint implementation: https://review.openstack.org/#/c/28245/  and https://review.openstack.org/#/c/28289/  3) LBaaS plan for Havana.
14:03:45 <enikanorov-w> hi Youcef
14:03:45 <markmcclain> Can we cover 1) LBaaS Dev Process, 2) LBaaS plan for Havana 3) conflicting reviews 4) Code reorganization
14:03:54 <enikanorov-w> of course
14:03:55 <Youcef> Hi eugene
14:04:02 <markmcclain> is everyone ok with that order?
14:04:23 <ilyashakhat> +
14:04:27 <avishayb> I am OK with the order
14:04:33 <SamuelB> Sure
14:04:46 <Youcef> yep
14:05:07 <markmcclain> ok
14:05:17 <markmcclain> #topic LBaaS Development Process
14:05:25 <enikanorov-w> ok, markmcclain you have something on (1)  or can I stert?
14:05:29 <enikanorov-w> *start
14:06:13 <markmcclain> I wanted highlight that we've got a BP
14:06:17 <markmcclain> that contains this doc: https://docs.google.com/a/dreamhost.com/document/d/1OT9m3bWl4yimvXLXTh_REQqONSS_f8jwplm7Y1iBxC8/edit
14:06:44 <markmcclain> enikanorov-w: you can start
14:07:43 <enikanorov-w> ok, so in fact we need to discuss the process as we got in the case where we have overlapping patches for the same blueprint
14:08:24 <enikanorov-w> and I wanded to propose that everyone who is willingtoimplement a major feature post it's brief design on ML first
14:08:31 <enikanorov-w> so others could take a look and discuss
14:08:47 <enikanorov-w> that was done on most major features
14:08:48 <markmcclain> enikanorov-w: that is generally the purpose of blueprints
14:09:37 <avishayb> I have posted this BP to the ML
14:09:44 <enikanorov-w> I think having approved blueprint doesn't automatically mean everyone agrees on the design or implementation
14:09:51 <enikanorov-w> it's rather acceptance into project plan
14:10:34 <enikanorov-w> that leads to a case where major design questions start to be discussed in gerrit, whish is not productive
14:10:48 <enikanorov-w> *which
14:11:36 <enikanorov-w> So our team would like to see implementer participating in ML discussions and corresponding meetings (we already had one)
14:11:41 <SamuelB> I actually disagree. For this specifci case. 1. we had this discussion prior to the summit 2. we have discussed this in the summit and agreed that this should be done in baby steps
14:12:16 <SamuelB> 3. based on that Avishay has published the BP
14:13:20 <SamuelB> And last the change was planned as small to make sure it is done and ready for H1 so that vendors could start to implement
14:13:20 <enikanorov-w> I think we'll get there when we discuss patches
14:13:57 <SamuelB> This as far as I understnad the correct way to go.
14:14:02 <enikanorov-w> right now the question is generic: would you agree that it's more productive to discuss on the meeting and ML than on gerrit?
14:14:27 <SamuelB> markmcclain: did I miss anything?
14:15:11 <markmcclain> SamuelB: you're correct on the flow
14:15:50 <enikanorov-w> so guys, you think, ML and meetings are needless? :)
14:16:02 <markmcclain> enikanorov-w: no both are useful
14:16:44 <enikanorov-w> ok, then our team would expect that everyone interested in lbaas would participate in public discussion
14:16:51 <Youcef> If there are any design decisions to be discussed, I prefer them discussed on ML than on gerrit, as I for one don't follow the gerrit reviews closely.
14:17:15 <SamuelB> What was not well understood is how a BL that was approved and got broken down to small bits to make sure we complete it, got a parallel implementation done by  enikanorov-w which tries to achive the first two steps with the addition of Service Types that do not have an approved design and BL as one patch
14:17:58 <enikanorov-w> Sam, i guess that relates to (3), lets discuss it a bit later
14:18:09 <markmcclain> So in terms of process
14:18:11 <ralekseenkov> guys, it's Roman. I think the issue Eugene is having is very simple
14:18:16 <ralekseenkov> he would make LBaaS meeting to discuss the code, what should be done and how, and noone would come
14:18:28 <SamuelB> As process goes this does not comply yo any process that I am aware of
14:18:32 <ralekseenkov> the last one was 11 days ago on Monday
14:18:52 <enikanorov-w> right
14:18:55 <ralekseenkov> so instead of having a productive live chat everyone fights on gerrit
14:19:24 <markmcclain> does every agree on the basic flow?
14:19:47 <ralekseenkov> I think we should agree that there is a standing meeting and all involved parties attend it
14:19:57 <avishayb> I did not submit even 1 line of code before I had an approved BP...
14:19:59 <markmcclain> we're getting ahead
14:20:15 <markmcclain> meetings are next :)
14:20:39 <markmcclain> general work flow is blueprints are accepted into the project plan
14:21:07 <markmcclain> and the assignee is responsible for shepherding the blueprint to completion
14:21:24 <Youcef> is the project plan published somewhere?
14:21:30 <enikanorov-w> Youcef: yes
14:21:43 <enikanorov-w> Youcef: thaths how we see it https://wiki.openstack.org/wiki/Quantum/LBaaS/HavanaPlan
14:22:22 <Youcef> enikanorov-w: thx
14:23:51 <enikanorov-w> I have posted this plan on ML 3 weeks ago btw
14:24:38 <markmcclain> that wiki is different from the blueprints
14:24:43 <markmcclain> that are in Havana
14:25:28 <enikanorov-w> so how we may get them accepted? I guess most of them have "proposed milestone" filled
14:26:45 <markmcclain> I'm the one who accepts them
14:26:57 <markmcclain> my understanding was this was the first one to be implemented:
14:26:57 <markmcclain> https://blueprints.launchpad.net/quantum/+spec/multi-vendor-support-for-lbaas
14:27:05 <enikanorov-w> right
14:27:41 <enikanorov-w> I was not aware of this BP filed until like 2 weeks ago
14:27:51 <enikanorov-w> and the plan was created earlier
14:28:01 <enikanorov-w> so I've added BP later
14:28:24 <markmcclain> that blueprint matched my notes from summit which is my I approved it
14:28:47 <enikanorov-w> the blueprint is usefull, no questions
14:29:16 <enikanorov-w> however splitting it and those baby steps are what needed to be discussed
14:30:12 <markmcclain> ok
14:30:27 <markmcclain> let's wrap up process and then we can jump to next topics
14:30:29 <enikanorov-w> i think the plan is just for everyone to consider and discuss
14:30:31 <enikanorov-w> right
14:30:54 <ralekseenkov> Eugene - the vision for lbaas in havana described on the wiki makes sense IMO. the action item is that you have to follow up with Mark over email and get the rest of the BPs justified, approved & scheduled.
14:31:15 <markmcclain> general flow is approved blueprints is what the core expects to review
14:31:28 <enikanorov-w> ok, btw, Mark, does launchpad allow you to subscribe for BPs?
14:31:49 <enikanorov-w> or you have to manually browse for new ones
14:32:33 <markmcclain> I'm subscribed to all of them
14:32:42 <enikanorov-w> ok
14:33:09 <markmcclain> so I get change deltas when items change
14:33:36 <SamuelB> markmcclain: can you please summarize the process?
14:33:37 <markmcclain> I also regularly review them to make sure the BP is on target for the linked milestone
14:34:44 <enikanorov-w> let's move on discussing patches maybe?
14:36:21 <markmcclain> process summary: the sub teams proposes blueprints that get evaluated and then approved for the cycle. The core team expects that the assigned person will be the one responsible for completing the task.
14:37:04 <Youcef> markmcclain: Is the project plan above (wiki), the approved plan of record for LBaaS (are all approved blueprints for H there)?
14:37:48 <markmcclain> Youcef: No.  It's a nice overview, but the blueprints are the authoritative record.
14:37:57 <SamuelB> I was actualy not aware on this wiki untill now
14:39:07 <enikanorov-w> ok, let's move to the patches
14:39:10 <SamuelB> I do follow the blueprints for H
14:39:33 <markmcclain> Blueprints that are approved when there's consensus.  If you disagree with the approval two options 1) reach out to the assignee to discuss concerns. 2) Reach out to me.
14:39:38 <markmcclain> So that's the process
14:39:45 <markmcclain> Let' move onto the current plan
14:39:54 <markmcclain> #topic Havana LBaaS Plan
14:40:21 <markmcclain> The one item that we need achieve is a working LBaaS implementation
14:40:41 <enikanorov-w> right.
14:42:39 <enikanorov-w> markmcclain: can you review that wiki page and send out your thoughts?
14:43:33 <markmcclain> my general thoughts are the same that I've been discussing with both the FW and VPN teams
14:43:37 <markmcclain> we need simple first
14:43:37 <SamuelB> enikanorov: please send the wiki on the ML. there were a couple of other items discussed at the summit that do not appear there
14:44:07 <enikanorov-w> SamuelB: the plan was posted on ML
14:44:11 <markmcclain> While service types and multi-vendor support is eventually where we want to go
14:44:22 <markmcclain> we need to get working single vendor implemetations first
14:44:29 <enikanorov-w> markmcclain: current reference impl is working
14:44:37 <enikanorov-w> i think it is a good starting point
14:44:52 <enikanorov-w> and can be treated as 'single vendor'
14:45:40 <enikanorov-w> so i think that goal is almost reached.
14:46:00 <markmcclain> enikanorov-w: it works, but we don't have any vendors who've written to that interface
14:46:07 <Pattabi> Can we start the vendor implementation integration  ?
14:46:47 <enikanorov-w> ok, let's discuss patches, because it seems to be quite related to what we're talking about
14:47:14 <Pattabi> markmcclain: we have been waiting for the framework to be available to integrate our vendor impl
14:47:39 <enikanorov-w> so, let me start still
14:47:50 <markmcclain> Pattabi: that's is part of what we are discussing
14:48:10 <markmcclain> Just wanted to give everyone warning that we've got about 12 mins until Ceilometer has the room
14:48:17 <enikanorov-w> we have two patches which are technically complimentary
14:49:12 <enikanorov-w> and we would definitely agree on the code if it would consist of abstract_Driver and noop_driver
14:49:50 <enikanorov-w> other parts which introduce additional plugin can be omitted.
14:50:07 <markmcclain> enikanorov-w: I'm not sure they are fully complimentary
14:50:22 <enikanorov-w> markmcclain: by intent they are not
14:50:41 <markmcclain> I'd like to see us focused on getting the driver done first
14:50:45 <enikanorov-w> because avishayb tries to address in new plugin what i've addressed in existing plugin
14:50:53 <markmcclain> that way Pattabi and others can begin work
14:51:10 <enikanorov-w> ok, abstract driver is good,
14:51:20 <enikanorov-w> driver for reference impl follows the interface
14:51:45 <markmcclain> enikanorov-w: it does, but the other proposal seeks to tweak that interface a bit
14:52:15 <markmcclain> I made a few design decisions that were HAProxy specific
14:52:20 <enikanorov-w> markmcclain: the driver for the reference impl will inherit from abstract, obviously
14:52:32 <markmcclain> right
14:52:33 <enikanorov-w> markmcclain: i tried to make them mor generic
14:52:46 <enikanorov-w> would be nice if you could review it
14:53:06 <enikanorov-w> also, we've tested it with devstack for different scenarious and got it working
14:53:20 <enikanorov-w> (most logic is just unchanged)
14:53:47 <Pattabi> which patch should be used :  https://review.openstack.org/28245 or https://review.openstack.org/28289
14:54:26 <markmcclain> I think enikanorov-w is referring to 28289
14:55:25 <markmcclain> I've read over both
14:55:45 <markmcclain> but withheld comments because I wanted to talk with everyone first
14:56:24 <markmcclain> for H1.. we really need to focus on the driver interface
14:56:29 <enikanorov-w> ok, so our opinion that we may just take abstract driver from avishay's patch, and derive the lbaas reference impl driver from that.
14:56:34 <enikanorov-w> markmcclain: no objections
14:56:46 <enikanorov-w> but i would object agains introducing different plugin
14:57:36 <enikanorov-w> also I think even both patches together are now so big and could make it into h-1
14:57:44 <avishayb> There is no diffrent plugin - after step1 we will end up with 1 plugin
14:58:04 <enikanorov-w> especially because you are the author of 99% of code of my patch
14:58:39 <avishayb> Guys - i do not see any reason to hold my work (steps 0 & 1) - do you?
14:58:43 <dhellmann> are you guys going to be wrapping up soon? we'll need the room soon for ceilometer
14:59:04 <markmcclain> dhellmann: yeah we'll wrap up
14:59:16 <dhellmann> thanks!
14:59:22 <enikanorov-w> avishayb: i think driver interface definitely has a value, we need to merge it
14:59:35 <Youcef> I think we all agree that we want one LBaaS plugin, and each vendor develop their own driver to the common driver interface.
15:00:01 <markmcclain> Ok.. let me look at reviews and follow up with everyone the ML
15:00:12 <enikanorov-w> markmcclain: thanks.
15:00:12 <markmcclain> we have to yield the room
15:00:15 <markmcclain> #endmeeting