15:00:20 <jimbaker> #startmeeting craton 15:00:20 <openstack> Meeting started Mon Jul 25 15:00:20 2016 UTC and is due to finish in 60 minutes. The chair is jimbaker. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:21 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:24 <openstack> The meeting name has been set to 'craton' 15:00:33 <jimbaker> #topic roll call 15:02:06 <sigmavirus> o/ 15:02:08 <cmspence> o/ 15:02:19 <jimbaker> sigmavirus, cmspence o/ 15:02:56 <mrhillsman> o/ 15:03:02 <jimbaker> mrhillsman, o/ 15:03:09 <sulo> o/ 15:03:20 <Mitschke> o/ 15:03:32 <jimbaker> sulo, Mitschke o/ 15:04:01 <jimbaker> today's agenda can be found here: https://etherpad.openstack.org/p/craton-meetings 15:04:12 <izaakk> o/ 15:04:28 <jimbaker> also feel free to take any notes above and beyond the minutes in that etherpad 15:04:58 <jimbaker> let's get started 15:05:11 <jimbaker> #topic action items from last week 15:06:06 <jimbaker> perhaps most importantly, the action item to move to openstack infra, which sarob helped to drive, is now completed 15:06:52 <jimbaker> at least in terms of gerrit; gates; reviewers 15:07:05 <sulo> still one to merge 15:07:12 <jimbaker> sulo, ? 15:07:13 <sulo> for tests 15:07:29 <jimbaker> ahhh, ok, i am head of this then 15:07:42 <jimbaker> ahead 15:07:56 <sulo> It already has a +2 so just a matter of time really 15:08:26 <sulo> I'll poke around to get it going today 15:08:34 <jimbaker> sulo, thanks 15:09:11 <rainya_> o/ 15:09:22 <jimbaker> so initial core reviewers for #craton are jimbaker, sarob, sigmavirus, sulo 15:09:44 <jimbaker> we hope to see more join this team over time 15:09:59 <jimbaker> izaakk, rainya_ o/ 15:10:10 <sulo> hi rainya_ 15:10:52 <jimbaker> another interesting aspect of the gate is that we have dropped python 2.7 and 3.4 testing of craton core 15:11:09 <jimbaker> the implication: we are only supporting 3.5 going forward 15:11:58 <jimbaker> this should simplify our code base and deployment, while allowing us to take full advantage of the most recent aspects of python 15:12:09 <cmspence> jimbaker: we are still planning on supporting 2.7+ for client, correct? 15:12:15 <jimbaker> cmspence, in fact we must 15:12:35 <cmspence> I know we had talked about that last week, just wanted to make sure nothing changed 15:12:39 <rainya> o/ (for the record) mostly just lurking again today all 15:12:52 <jimbaker> that was a crucial distinction we got in operator feedback 15:13:27 <jimbaker> this also applies presumably to horizon plugins 15:13:48 <jimbaker> anything that is client side is 2.7 or higher; but craton core is 3.5 15:14:48 <jimbaker> next question i have related to openstack infra - presumably we should move from #craton to #openstack-craton 15:15:38 <jimbaker> cmspence, sulo - any thoughts here? 15:16:03 <jimbaker> rainya, btw, we are trying to capture all of our meetings here https://etherpad.openstack.org/p/craton-meetings 15:16:14 <cmspence> jimbaker: I think that sounds good 15:16:29 <rainya> jimbaker, looks good 15:17:40 <jimbaker> sulo, do you know any registration details we need to go through for #openstack-craton ? 15:17:42 <sulo> unless it's a problem for openstack, I don't mind keeping just #craton 15:18:13 <sulo> but cool if we need to move 15:18:13 <jimbaker> sulo, it does have the advantage of being shorter... 15:18:41 <jimbaker> sulo, ;et 15:18:52 <jimbaker> let's research details 15:19:31 <jimbaker> but i would think it's a net positive, despite minor disruption 15:20:06 <jimbaker> sarob is not here today, so we will have to table the oneops discussion 15:20:23 <jimbaker> i will follow up with him on email re getting that work kicked of 15:20:49 <jimbaker> #action jimbaker discuss oneops integration kickoff with sarob 15:21:40 <jimbaker> cmspence, one last action item from last week re horizon storyboards 15:21:54 <jimbaker> any progress? do you need help? 15:23:08 <cmspence> no, I've had a couple meetings with the horizon folks and we have a meeting setup with sulo to see how the current rackspace UI looks to get an idea of functionality and flow. The Horizon team with tamayo will be digging in after that to create us some mockups 15:23:35 <jimbaker> cmspence, sounds like good progress to me, thanks! 15:24:22 <jimbaker> #action development highlights 15:24:30 <jimbaker> sorry, wrong command 15:24:44 <jimbaker> #topic development highlights 15:25:59 <jimbaker> one topic we discussed this past week in our dev meeting, based on discussion with product owners around the fleet management user stories, is the focus on FLT005 15:26:12 <jimbaker> (https://specs.openstack.org/openstack/openstack-user-stories/user-stories/proposed/fleet-management.html) 15:27:05 <jimbaker> i believe we made good progress on how to use FLT005 to better define the craton product 15:27:26 <jimbaker> specifically in outlining the need to focus in doing so with respect to the craton client 15:28:02 <jimbaker> eg an operator should be able to use the craton client to define/configure an audit workflow, such as against the OSA security role 15:28:28 <jimbaker> an operator should be able to use the craton client to apply a workflow against some queried inventory 15:29:04 <jimbaker> this in turn can be done equivalently in the horizon dashboard that we are building out for craton 15:30:06 <jimbaker> cmspence, sigmavirus, sulo - we should start defining the specifics of this craton client api around workflow support 15:30:35 <sulo> yeah that would help me to get the api side going quicker 15:30:38 <sulo> so +1 15:30:48 <jimbaker> so that this can drive dev there, much as we used the REST API (more directly) earlier against inventory 15:31:16 <cmspence> +1 15:31:56 <sigmavirus> Sounds good 15:32:29 <jimbaker> cool, i'm hopeful we can start working on some gists/etherpads/etc to prototype this api this week 15:32:43 <jimbaker> and discuss in thurs dev meeting 15:33:45 <jimbaker> we can then work out the corresponding CLI aspects, which i expect to be an important focus of the barcelona demo (assuming talk acceptance etc) 15:33:59 <jimbaker> perhaps the most important 15:34:59 <jimbaker> sigmavirus, you mentioned github3 as one possible source of api inspiration - i think it looks like a good approach 15:35:30 <sulo> api inspiration ? 15:35:41 <jimbaker> sulo, what the client api would look like 15:36:07 <jimbaker> https://github3.readthedocs.io/en/develop/ 15:36:25 <jimbaker> obviously github is a much larger api surface 15:37:05 <jimbaker> but if we focus on making our resources behave in a similar fashion with respect to that api, this seems like a nice approach 15:37:14 <cmspence> maybe this is a topic we can discuss in #craton, but what aspects are we expecting to be similar? 15:38:01 <sulo> so we want to model the client like github3 ? 15:38:27 <jimbaker> cmspence, sulo - i think in terms of session mgmt; simple usage 15:38:40 <jimbaker> we don't have repos, of course 15:39:03 <jimbaker> but https://github3.readthedocs.io/en/develop/repos.html represents a simple way of interacting with this type of resource 15:39:23 <jimbaker> we do have workflows and inventory. what can we do to get similar simple usage? 15:39:30 <jimbaker> that's just the question i'm posing here 15:39:43 <jimbaker> cmspence, and yes, it's a perfect topic for more discussion on #craton 15:39:56 <jimbaker> hence my attempt to kick off everyone's thought processes 15:41:04 <jimbaker> before i go to far off in this topic: consider what it means to define a multiphase workflow, such as seen in actual live migrations 15:41:10 <sigmavirus> jimbaker: yeah, I mean people like github3.py because it lets people follow their noses 15:41:19 <jimbaker> sigmavirus, exactly 15:41:24 <sigmavirus> (I have a github client, I can get a repo. now I can get a pull request, etc.) 15:42:02 <jimbaker> so that connectivity also makes sense. what workflows did i apply to this host? 15:42:09 <jimbaker> what variables did this workflow capture? 15:42:10 <jimbaker> etc 15:42:45 <sigmavirus> jimbaker: sure, and cratonclient at present kind of allows for that already 15:42:57 <sigmavirus> we'll just end up doing things differently 15:43:01 <jimbaker> the craton client is the most important thing for our operators 15:43:14 <sigmavirus> instead of craton = client.Client(...); client.hosts.list() we'd end up with client.list_hosts() 15:43:45 <sigmavirus> er, 'craton.hosts.list()', 'craton.list_hosts()' or even just 'craton.hosts()' 15:43:47 <jimbaker> sigmavirus, more direct access makes a lot of sense 15:43:53 <jimbaker> craton.hosts() 15:43:58 <jimbaker> etc 15:44:04 <jimbaker> or so i would think 15:44:33 <jimbaker> anyway, if we can address FLT005 from this question, and come up with the corresponding code usage of the cratonclient api 15:44:40 <jimbaker> i think we will be in great shape 15:46:11 <cmspence> I've already started on the CLI commands we may want to use for FLT005: I would anticipate the simliar python command as well as eventually how we achieve this in horizon: Feel free to add some, remove some, etc: https://etherpad.openstack.org/p/craton_rest_api_basics 15:46:38 <jimbaker> can we agree then on an action item like so: begin operator-focused api definition against FLT005 ? 15:46:53 <cmspence> jimbaker: +1 15:47:18 <jimbaker> i can add in some thoughts based on sulo's script that we use for live migration at rackspace 15:47:39 <jimbaker> ok, need to wrap this up 15:48:02 <jimbaker> #action begin operator-focused api definition against FLT005 jimbaker, cmspence, sigmavirus, sulo 15:48:58 <jimbaker> #topic taskflow container discussion 15:49:01 <sulo> lets move this to #craton maybe, there is few things to sort out for workflow though cli 15:49:19 <sulo> especially .. understanding flow -> workflow relationship etc 15:49:22 <jimbaker> sulo, yeah, we must do that given the lack of time 15:49:31 <jimbaker> in this meeting room 15:49:35 <sulo> ok 15:50:02 <jimbaker> sulo, jimbaker, and sputnik13 met up last friday to discuss container support 15:50:08 <jimbaker> w/ respect to taskflow 15:50:34 <jimbaker> harlowja could not attend due to travel problems/nightmare 15:51:02 <jimbaker> we are going to follow up with them, starting in the forthcoming #openstack-meeting-alt meeting in 9 min 15:51:33 <jimbaker> sulo, do you want to briefly summarize where we stand? 15:51:39 <sulo> sure 15:51:49 <sulo> so the idea from craton side is: 15:52:26 <sulo> allow running tasks inside a container such that we have a continer based engine in taskflow 15:52:47 <sulo> this will allow to use multiple languages support including bash scipts and binaries 15:52:51 <sulo> which is what operators do 15:53:09 <sulo> this is not possible currently in taskflow model .. so we'd like to bake this into either taskflow project 15:53:14 <sulo> or in oslo 15:53:24 <sulo> so we discussed this. 15:53:49 <jimbaker> with the possible outcome: a separate library focused on container support, but in the taskflow namespace 15:54:05 <jimbaker> given general applicability to not just craton 15:55:10 <jimbaker> i think this is a pretty exciting discussion, given the possibility of radically simplifying how we package up tasks to run on workers 15:55:24 <jimbaker> by moving to a docker packaging problem 15:55:52 <jimbaker> we just have 4 minutes left 15:55:55 <jimbaker> any questions? 15:56:02 <jimbaker> #topic questions 15:57:23 <jimbaker> we can always follow up in #craton (or perhaps soon #openstack-craton...) 15:58:54 <jimbaker> if you are interested in the taskflow discussion, we may have a chance to followup in the oslo meeting starting momentarily, #openstack-meeting-alt 15:59:09 <jimbaker> sulo, ^^^ feel free to attend, but i will scope it out 15:59:18 <jimbaker> #endmeeting