20:02:49 #startmeeting trove 20:02:50 Meeting started Wed Jul 3 20:02:49 2013 UTC. The chair is hub_cap. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:02:51 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:02:53 The meeting name has been set to 'trove' 20:02:55 i almost did startmeeting reddwarf 20:03:05 o/ 20:03:07 #link https://wiki.openstack.org/wiki/Meetings/TroveMeeting 20:03:13 \o 20:03:17 \o/ 20:03:21 hands across america 20:03:22 o7 20:03:28 \o 20:03:41 does that actually take role for us? 20:03:53 don't think so 20:03:53 o/ 20:03:58 i think just saying something in the channel logs you 20:03:59 nope 20:04:01 its just cool 20:04:12 oh it's roll call anyway 20:04:18 so lets get started seems like we have enough people 20:04:22 whers grapex_? 20:04:29 hes the only one from core i dont see here 20:04:34 hub_cap: i thought he said he would be late 20:04:40 he did 20:04:43 ok 20:04:49 #topic action items 20:04:53 #link http://eavesdrop.openstack.org/meetings/trove/2013/trove.2013-06-25-21.04.html 20:05:12 esmute SlickNik vm-gate integration status? 20:05:49 tap tap tap... is this thing on 20:05:50 Damn.. feel bad.. been dragging this for a while 20:05:54 its ok 20:05:56 still work in progress... 20:06:00 no update is a status 20:06:11 re-action it plz 20:06:14 wait 20:06:22 just add it ot the meeting as a thing to talk about each wk 20:06:28 then it will annoy u enough to do it 20:06:30 :D 20:06:35 Okay, I will do that. :) 20:06:42 juice: vipul whats this about packaging in heat? 20:06:46 i spoke briefly with clarkb and he gave me some points where to start 20:06:58 heat doesnt define the "how" for packaging 20:07:02 hub_cap: didn't get a chance to look into this yet 20:07:10 vipul: can u add to agenda 20:07:12 this was supposed to be used as an example for how we'd split guestagent 20:07:14 so i can get caught up on your conerns 20:07:21 ok 20:07:21 Sure 20:07:24 lets talk about it 20:07:28 hub_cap: I think it's less about heat and more about coming up with better packages for trove 20:07:32 yes? 20:07:36 vipul? 20:07:39 yep 20:07:43 sure, dont u mean tarballs :P 20:07:47 using heat's example pretty much 20:07:59 no zip files 20:08:02 :) 20:08:06 HA 20:08:10 ok we talked about replication 20:08:17 and ive added a "lets talk about it again" to the agenda 20:08:24 we should pick another time to meet 20:08:37 oh and this time, lets do it _in_ a room w/ meetbot 20:08:46 so we can have minutes etc for it 20:09:07 that sounds good. 20:09:12 tehre was some concern on our end about /instances and /clusters not overlapping but we can talk about that during the next meeting 20:09:18 sure 20:09:25 vipul: i updated the wiki article already ;) 20:09:35 noticed that :) thanks hub_cap 20:09:41 we got a bunch of broken links though.. 20:09:53 #action vipul to fix broken links hanging off wiki.openstack.org/trove 20:10:10 #action hub_cap to help out w/ the changes 20:10:21 ive done, id say, 30% of the changes to random pages from a search i had done 20:10:41 SlickNik: looks like the last one is u 20:10:49 im assuming the rename+jenkins is ok 20:10:54 so lets talk about those dev dox 20:11:00 Yeah, I looked into it a bit. 20:11:04 o/ 20:11:13 hello grapex_! 20:11:22 These are built by the CI infra team as part of a different "documentation" job. 20:11:25 you havent missed much still in action items territory 20:11:31 hub_cap: I read. 20:11:40 looks like you've inherited an _ isntead of a character :) 20:11:42 ok so we need to spend some cycles on the CI infra/docs stuff then eh? 20:11:49 lol vipul 20:12:03 vipul: I hopped over another user on freenode. It's like getting crowned in checkers. 20:12:09 Yes. Building it is fairly straightforward. 20:12:23 ok good deal 20:12:24 We need to figure out what we want to publish in these docs. 20:12:27 grapex_: ghost cmd 20:12:39 grapex_: register your nick 20:12:53 ok seems a bit low prio to me, but we are getting more and more questions about deving in trove 20:13:17 Perhaps some of the trove-integration docs would be a good start and we can build on that. 20:13:32 probably so SlickNik 20:13:37 lets do this 20:13:46 next n00b in trove writes docs ;) 20:13:50 I don't know if people trying to develop with/on Trove should be "low prio" 20:13:51 lol 20:13:57 ha ha 20:14:03 that's a new one :) 20:14:03 datsun180b: tahts not a low prio docs are 20:14:10 I'll get a job up and running so that there will be a place to write them too. 20:14:13 but as soon as more people ask, theyu become a high prio 20:14:20 right 20:14:26 one thing that would really be good is a install guide 20:14:27 basically the ammt of uptick defines the priority 20:14:29 now its low :) 20:14:32 s/too/to 20:14:33 like what all the configuration params mean 20:14:40 kinda like Nova has an op's guide 20:14:41 vipul: yes thats a good idea 20:14:57 I'm trying to leapfrog install guides with vagrant with moderate success 20:15:00 the CERN guys are asking how to install this in a non 'dev env' way 20:15:21 #action SlickNik to look into getting a documentation ci-infra job up. 20:15:24 SlickNik: where would such guides be committed? 20:15:34 deployment guides would be really nice 20:15:50 #agreed 20:15:55 lets move on we have a big agenda 20:16:07 vipul: get w/ SlickNik to see where they go, im interested as well 20:16:14 coo 20:16:14 #topic h2 blueprints 20:16:21 vipul: I think a different repo, although I've seen some variations in what other projects do. 20:16:37 fwiw, since we are incubated, we need to keep better track of the blueprints and when they are integrated 20:16:40 err implemented 20:16:52 ive seen a lot of code go thru the last few wks but no blueprints actually marked as implemented 20:17:01 there is no automated process for this 20:17:07 its your mouse and clicking 20:17:08 :D 20:17:13 and hudson doesn't actually close the bps does he 20:17:14 vipul: so will talk to clarkb to see what the intended location is. 20:17:15 #link https://launchpad.net/trove/+milestone/havana-2 20:17:20 datsun180b: correct 20:17:36 it might help if we approved the bps ahead of merges too 20:17:39 hub_cap: I don't use a mouse. 20:17:44 plz go over the blueprints you have implemented in havana and make sure they are there 20:17:51 is this list ok to change? meaning can i add stuff to H2? 20:17:53 SlickNik: s/mouse/trackpad 20:18:03 vipul: if u think it can be implemented in 2wks yes 20:18:07 or if its alrady done 20:18:07 adn what is the date for the h2 milestone 20:18:10 then hell yes 20:18:17 its on that page vipul 20:18:22 2013-07-18 20:18:23 oh 20:18:25 got it thanks 20:18:50 #action hub_cap vipul to make sure all blueprints that are actually in master are in h2 + implemented status 20:18:55 lets split it up vipul and do it 20:19:03 hub_cap: I see you've already marked a bunch of them as completed. 20:19:04 Thanks 20:19:07 Ok sounds good.. i know half of those may have already been complete 20:19:07 cp16net: you have a BLOCKED bp in h2 20:19:09 SlickNik: correct 20:19:14 vipul: im sure they are 20:19:20 oh 20:19:21 we just ahvent been diligent about it (yet!) 20:19:28 should that be in h2? 20:19:32 is it completely necessary? 20:20:07 and juice i assume yours will be done by then eh? i think grapex_ +2'd u today 20:20:40 does this mean we tag on 7/18? 20:20:41 hmm i forgot about that 20:20:43 im going to look @ it after i eat some lunch today juice as well 20:20:48 vipul: on or around yes 20:20:57 and we dont, ttx does 20:21:03 he has some automation for all of this 20:21:08 sweet 20:21:16 nice 20:21:20 so we will have official h2 pkgs 20:21:24 its a good bp but i havnt had time to go back and fix it up with other things i have been working on 20:21:33 ok cp16net plz move it to h3 20:21:45 core team 20:21:50 we need to sit down and look @ all the bps 20:21:59 and decide whats 100% necessary for havana 20:22:14 #action hub_cap to grab core and discuss blueprints sometime next wk 20:22:20 ill take a pass 20:22:25 and then we can discuss the rest 20:22:30 ive taken a few passes already too 20:22:39 Sounds good, hub_cap. 20:22:44 aight horse beaten. anything else on the subject? 20:22:56 how many milestones are there in havana 20:23:01 and how long is each 20:23:02 vipul: 3 in all releases 20:23:07 2mo ea 20:23:22 k 20:23:22 hub_cap: moved to h3 20:23:26 #link https://launchpad.net/trove/havana 20:23:32 ttx deleted our h1 since we didnt push a release 20:23:42 and moved all the h1 completed stuff to h2 20:23:48 hub_cap: There are a few bps that are still marked as grizzly. Probably want to reset them as you're taking a pass. 20:23:58 SlickNik: ya i saw that yest 20:24:01 if they arent in already def 20:24:23 moving on, eh? 20:24:32 cool, sounds good. 20:24:43 #topic Cluster+Heat findings 20:24:53 #link https://wiki.openstack.org/wiki/Trove/HeatIntegration 20:25:07 i had some misconceptions from the #heat channel a few days ag 20:25:08 *ago 20:25:23 heat will fit into clustering quite nicely 20:25:51 but it dropped on my prio list a bit 20:26:03 so im going to finish rhel integration and then move to heat 20:26:44 Those last allcaps notes make it sound like Mr. Hyde took over while you were editing 20:27:07 datsun180b: he did 20:27:11 hes my sanity 20:27:28 ok so does anyone have anything to add here? 20:27:43 I guess some background on how you had originally planned to use it 20:27:44 itll be pretty straightforward. ill push things in small reviews 20:27:52 and how do you see things now 20:28:11 sure vipul there is a email that explains things well but ill summarize (more context in email if needed) 20:28:19 datsun180b: can u find the pipermail link to that email chain? 20:28:39 so 1) launch configuration defines the flavor for a grouping of instances 20:28:41 I don't know all of those words 20:29:02 datsun180b: like http://lists.openstack.org/pipermail/openstack-dev/2013-May/009666.html 20:29:09 but for our email 20:29:12 the one that i sent to the list while im typing this stuff 20:29:24 #link http://lists.openstack.org/pipermail/openstack-dev/2013-July/011289.html 20:29:35 got it. 20:29:37 so we dont need to specify a launch configuration and one of the core devs doesnt even like them 20:29:41 with the heat reasons? 20:29:43 originally i thought it had to be homogeneous 20:29:48 datsun180b: SlickNik linked it 20:29:50 there you go 20:29:56 i'm helping! 20:30:07 but thats not the case, resizes on individual instances in a stack are a-ok 20:30:17 ok so that allows us to have differnt flavors for instances in a single template? 20:30:19 which is necessary 20:30:23 correct vipul 20:30:27 so it meets that criteria 20:30:41 also you can add/remove pretty easily, just by specifying changes to your yaml config 20:30:48 s/config/template 20:31:01 so that meets the criteria of scaling up/down 20:31:23 and the nodes can be different, as in u can have a primary and a secondary in the same template 20:31:53 I think the last point is key for us. 20:31:54 so they can all be easily grouped together (i was originally told that the secondarys/replicas/etc could be one stack but the primary had to be its own) 20:31:56 Ok makes sense 20:32:46 i.e. the fact that we can have multiple heterogenous nodes in the same template. 20:32:59 but so some work is needed in Heat to support all this? 20:33:33 vipul: not really 20:33:40 lol heat is talking about trove in #openstack-meeting 20:33:46 HAAHAHHA circular dependency 20:33:50 hey ... 20:34:20 (probably other heaters here if you have questions) 20:34:22 I think we need to come up with the right templates that suit our needs. 20:34:36 heaters gonna heat 20:34:36 SlickNik: ive got a template already 20:34:40 LOL 20:34:56 but thats about as far as i got SlickNik, was testing and getting a template working 20:35:14 lol@amytron 20:35:17 BUT ill only be impl'ing it for single instances for now, obvi since we dont have clusters yet 20:35:25 but ill help guide the integration for clusters 20:36:40 ok this is hard doing both 20:36:42 lets move on 20:36:46 sounds like a plan, hub_cap 20:36:47 heh 20:36:58 #topic next replication meeting 20:37:08 someone spit out some times 20:37:21 Thursday @1pst 20:37:34 That works for me. 20:37:51 I can't use doodle again, my free trial expired 20:38:11 at least i don't get the full feature set or something 20:38:14 heh, hopefully we won't have to resort to doodle again.. 20:39:01 yes i dont care about a doodle, lets say sometime tomorrow 20:39:08 LOL JK 20:39:10 monday 20:39:11 what time 20:39:12 lol 20:39:19 when was the last oen? lets do it again at the same time next wk 20:39:23 day/time 20:39:32 thursday 1pst? 20:39:36 :) 20:40:08 Last one was on Thursday 1PST/ 3CST 20:40:19 Ah, I see what you did there, vipul :) 20:40:41 :P 20:40:46 do we want to hold the next one tomorrow 20:40:49 HEHE 20:40:52 next thr? 20:41:05 Yeah, I think that's what vipul was suggesting. 20:41:20 (Or at least I hope that's what he was suggesting) 20:41:31 oh yea i meant next week lol 20:41:38 phew! 20:41:45 ok good 20:41:50 someone #action it and we can move on 20:41:57 i have averted my double meeting crisis 20:41:57 tomorrow 8:00 AM CST is good for me 20:42:15 kevinconway: im glad you dont care about your 4th of july holiday 20:42:16 but the rest of us do 20:42:19 :P 20:42:20 kevinconway always suggesting those gawd awful times 20:42:40 #action Next meeting to discuss replication: Thursday, 11th July 1 p.m. PST/3 p.m. CST 20:42:44 kevinconway must not understand >1 TZ for trove hehe 20:43:00 i think he exactly understands 20:43:07 i know ;) 20:43:09 moving on? 20:43:18 o7 20:43:23 #topic guest-agent packaging 20:43:39 i was literally JUST talking about this in #openstack-meeting w/ heat 20:43:56 we came to the conclusion of no pypi lol 20:44:00 this kinda force our hand in splitting the repo? 20:44:23 what's the suggestion instead 20:44:29 no suggestion instead 20:44:37 just dont use pypi 20:44:47 id like to help yall w/ this 20:45:01 since itll affect the heat-trove integration as well 20:45:15 i dont think i have any good ideas as of now tho 20:45:21 well we have the ability to bake into images 20:45:30 NO 20:45:41 then we need to roll a new image to test any guest changes 20:45:46 yup 20:45:47 thats the opposite of what we want :D 20:46:06 so are we talking in devstack context or how heat would consume the agent? 20:46:08 itll be interesting, cuz it has to be different 20:46:15 vipul: i dont know 20:46:22 what are we talking about lol 20:46:35 hmm 20:46:37 im starting (mentally) from local development 20:46:39 :| 20:46:55 if what yall were talking about last wk was different then plz correct me!! 20:47:18 im like a child that wanders into the middle 20:47:18 Aren't there really 2 different issues here. PyPi only talks about the vector of delivery. The work to split the guest agent to live in its own separate repo is tangential... 20:47:22 So we talked about splitting out guest-agent inot own repo.. 20:47:38 SlickNik: it actualy solves the packaging issue 20:47:49 yes im ok w/ splitting repos 20:47:51 that makes sense 20:47:55 Since we could just have a setup.py in the new repo that pushed to pypi 20:48:12 and you'd be able to release a version of guest agent in the same cycle as Trove 20:48:16 the heat guys seem to dislike pypi cuz of its speed / reliability 20:48:21 Well, you could have the setup.py and still not use pypi. 20:48:34 correct.. 20:48:39 (but use something like tarballs instead) 20:48:47 But we can cross that bridge when we get to it. 20:49:03 Basically we have no way to independently version the guest agent currently 20:49:07 so we need to solve 1) local dev, 2) devstack, 3) production-esque installations 20:49:08 so how do we solve that 20:49:43 tripleo can help us solve #2 and #3 20:50:08 #link https://github.com/stackforge/tripleo-image-elements/blob/master/elements/heat-cfntools/install.d/05-heat-cfntools 20:50:20 this is how heat-cfntools is installed in an element 20:50:26 we can install the guest in the same way 20:50:59 yea that would be fine.. so a requirement for that is to split the repo 20:51:07 correct 20:51:09 i mean its not honestly 20:51:14 but its a nicety 20:51:21 so really the _only_ reason we _have_ to split 20:51:24 is for pypi 20:51:54 hub_cap: there's a couple of other advantages to splitting the repo as well. 20:51:58 sure 20:51:59 when trove is tagged with havana-2.. what's the deliverable of that? 20:52:08 i think it just means that it goes to pypi 20:52:16 server/guest are deliverables 20:52:23 We'd be able to use endpoints for the guestagent. 20:52:26 so it makes sense to have independent ones 20:52:31 SlickNik: endpoints? 20:52:51 Enabling us to use stevedore, for instance. 20:52:55 and are you solving 1, 2, or 3? 20:52:58 entrypoints 20:52:58 sorry entry_points 20:53:00 our personal little packaging of trove that annashen did supports separate versions of guest and trove-api 20:53:11 using the current layout 20:53:17 for sure juice 20:53:32 mordred: says pypi hates > 1 deliverable in 1 project ( i believe) 20:53:47 but regardless, im not arguing against splitting 20:53:58 im just trying to figure out what the higest prios are 20:54:02 well, the other reason you have to split is for OpenStack CI 20:54:07 pypi doesn't seem to support much of anything AND in any case it's a bit ugly but it does seem to work fine 20:54:09 does splitting guest agent to another repo imply that we will have a common-trove package? 20:54:33 mordred: sure 20:54:34 we're not going to write custom it-works-different-for-this-proejct automation. so, the rule is one tarball per repo 20:54:49 esp: i don't think it has to introduce another repo 20:54:54 mordred: we arent unique snowflakes?!?! 20:55:05 * mordred hands hub_cap a bunny rabbit 20:55:10 vipul: gotcha, juice set me straight :) 20:55:15 * hub_cap shreds mordred with a sensational knife 20:55:29 so srsly im not arguing against splitting, im just trying to figure priorities here 20:55:33 lets take it offline 20:55:38 i think a split is necesary 20:55:41 just not sure when we shoudl do it 20:55:50 im sure itll be done before icehouse 20:56:01 i just want to see whats necesary first, if anything 20:56:10 is that the name for hte next release? 20:56:15 hub_cap: I'm +1 for splitting; what are we prioritizing against? 20:56:15 vipul: aye 20:56:26 i think heat + integration 20:56:45 and making sure the 1, 2, 3 i mentioned above all work smoothly when we cut the rope 20:56:46 +1 for splitting too 20:56:58 i guess im saying we need a plan :D 20:57:09 and how we will integrate once we cut it 20:57:10 i think we need 16 repos 20:57:17 i think we need to figure out if we need to do anythigin speical for heat 20:57:19 kevinconway: you arelady have that many 20:57:28 kevinconway: https://github.com/openstack 20:57:37 acutally 45! 20:57:46 anyhoo lets move on and talk about it in #openstack-trove 20:57:54 we have 3 min and i wanted to briefly talk about the next item 20:58:01 go for it 20:58:06 #topic Mission stmt 20:58:20 so if you follow the TC meetings / list, u know now that projects are now called programs 20:58:27 and that each program needs a mission statement 20:58:42 nothing persay is different in our incubation, but we are now a program 20:58:50 Yes, heard about that. 20:58:50 and Nova, Glance, Infra, etc... are too 20:58:54 Trove Mission: Needs more Heat! 20:59:09 whats cookin on the heat? 20:59:14 kevinconway: The heat is on. 20:59:18 It's on the street. 20:59:23 kevinconway: remind me to never let u or cp16net or grapex_ write a mission stmt 20:59:27 #link https://wiki.openstack.org/wiki/Trove#Mission_Statement 20:59:32