14:02:19 #startmeeting airship 14:02:20 Meeting started Tue Dec 11 14:02:19 2018 UTC and is due to finish in 60 minutes. The chair is mark-burnett. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:02:21 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:02:23 The meeting name has been set to 'airship' 14:02:31 Hi all 14:02:31 o/ 14:02:37 GM mark-burnett 14:02:38 Here's the very empty etherpad for today: https://etherpad.openstack.org/p/airship-meeting-2018-12-11 14:02:39 o/ 14:05:23 dwalt I'm on mobile for few mins so no easy access to etherpad- could you share a status update on the airskiff gates, where they are, where they're going? 14:05:28 nice agenda 14:07:25 Absolutely! 14:11:26 Now that Airskiff has been integrated in the Treasuremap repository as a site, we have added several gates. Two of which lint each site (Airship-Seaworthy, Airskiff), and one that deploys a couple of components (i.e memcached) using the Airship components deployed during the Airskiff setup process. We’re going through the next phase right now of adding that Treasuremap gate as a non-voting job to each component’s 14:11:26 repository that is represented in Airskiff. 14:12:38 Additionally, we are trying to reduce the size of the Airskiff site. There are a lot of extra files that were inherited from the Seaworthy site that are ignored when deploying Airskiff 14:14:01 so the airskiff gate *in treasuremap* today -- what does it gate on? Any changes to treasuremap, or something more specific? 14:15:35 Great question. Because it deploys the whole site using the developer setup scripts, any changes to those scripts or the Airskiff site documents are gated 14:16:07 Additionally, we build all of the Airship components from source. So they are sort of verified on each run 14:16:42 gotcha. How long does the full gate run take in openstack infra? 14:18:16 It’s about an hour at the present time. While we are looking at ways to reduce the time (e.g. exploring minikube for K8s), the priority is still to represent this job in the other Airship components first 14:19:30 makes sense. If we're building the airship components {pegleg, armada, shipyard, deckhand} from source as part of the gate, would it speed things up to only build the project being tested, and pull the curent built images for the rest? 14:21:20 That’s another idea that’s also been tossed around a bit. That is another thing I am hoping to accomplish once the gate runs for the other Airship repositories 14:21:49 awesome 14:21:50 Right now, Treasuremap is sort of serving as an integration point 14:22:26 For those interested, we also have a roadmap on the Treasuremap storyboard 14:22:47 thanks for your work on that - it'll be really valuable getting some gating on every patch :) 14:22:59 link to storyboard? 14:23:35 https://storyboard.openstack.org/#!/project/openstack/airship-treasuremap 14:23:51 Definitely! 14:24:31 I think those are all my questions with which to pepper dwalt at this time mark-burnett :) 14:26:28 Great 14:26:41 Anyone else have any topics? 14:26:58 Airship-in-a-bottle ? 14:27:59 Ok 14:28:15 #topic Airship in a bottle 14:29:31 Is there an actual question or subject regard AIAB that needs discussion? 14:29:37 regarding* 14:30:07 I added what I know about efforts to move it forward to the etherpad. 14:30:56 thats great b-str 14:31:15 the question is more around what we do to try and get us in a situation where it works all the time 14:31:27 do we have capacity to put some cicd around it? 14:31:41 or a community goal to keep it healty 14:32:04 as its the entrypoint for most people into airship, its very important that their first experence is a good one. 14:32:50 +1 14:33:07 Seems the largest hurdle is 3rd party CI doesn't have a method for publishing logs 14:33:11 There is an AIAB pipeline that runs (internally) but it doesn't gate, and doesn't publish results publicly. 14:33:46 So all the results are silent due to OSF rules that 3rd party CI must publicly publish results 14:33:48 Substantial "doesn't work on my machine" feedback on ML and otherwise -- part of that is system requirements, part of that is occasionally something is broken 14:34:05 Was about to say - can we publish the results :) 14:34:24 I feel like we're onto rehashing old topics here 14:34:50 k, what was the outcome of that one mark-burnett 14:34:52 We've discussed our inability to publish 3rd party gate results a number of times in this forum 14:35:29 One ML request was to reduce the memory requirement, which we've known since June or so, but no substantial changes have happened since then on this topic. 14:35:30 The outcome is simply that the team responsible doesn't have the bandwidth to push that forward right now 14:36:13 Sure, bandwidth aside - I wouldn't phrase that inability, just opportunity 14:36:27 "when" not "if" 14:37:19 ok - I'll see if i can help on the publishing issue 14:37:29 i may have some levers i can pull 14:37:39 though would probably be after christmas 14:37:40 Did we ever get clarity on how Ironic or Triple-O does gating of the full platform? I've heard two stories around that - one is 3rd party gates and the other is they have worked zuul magic. 14:37:57 ty portdirect :) 14:38:06 ironic - just uses zuul and nodepool as per everyone else 14:38:41 triple-o uses zuul and nodepool, though with some deadicated hardware if i rememeber correctly 14:39:10 Okay 14:39:17 though as we are not 'openstack' we can use third party gates however we like as far as im aware 14:39:19 I was just looking at Ironic PS and it seems like they have vendor gates 14:39:26 so should have no limitations 14:39:29 Yeah, they have pretty significant 3rd party gates, iirc 14:39:40 ironic has third party gates, but do not rely upon them 14:39:58 the core implementation only requires vbmc to run 14:40:51 this is how we gate it in osh 14:41:13 so you may want to check that for somthing that would plug neatly into airskiff 14:42:40 Ok, sounds like we're basically done 14:42:45 Thanks for coming everyone! 14:43:02 #endmeeting