15:01:21 #startmeeting third-party 15:01:22 Meeting started Mon Sep 14 15:01:21 2015 UTC and is due to finish in 60 minutes. The chair is anteaya. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:24 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:27 The meeting name has been set to 'third_party' 15:01:30 hello 15:01:45 o/ 15:01:52 hi anteaya 15:02:15 hey asselin_ 15:02:27 asselin_: how are you this morning? 15:02:45 anteaya, I'm fine :) 15:03:05 glad to hear it 15:03:17 anything third party wise you would like to discuss today? 15:03:54 nothing in particular 15:04:47 okay 15:05:01 any developments on the third party dashboard tool? 15:05:38 I don't think there's anything new. I'm planning on looking into them in more depth today 15:06:04 hopefully we can make a go-forward decision tomorrow 15:06:06 awesome 15:06:08 woooooo 15:06:22 but I still haven't heard back from Sweston regarding radar 15:07:53 okay 15:08:03 a data point to keep in mind 15:08:08 anteaya, have you seen the tool options? do you have any opinions? 15:08:20 I have seen a few 15:08:35 I am willing to look at your assessment of the tools 15:08:46 and I can offer an opinion if you like 15:09:31 ok I need to find the etherpad link 15:09:39 thanks 15:12:54 I can't find it. it's in the #openstack-third-party-ci room which isn't logged.... 15:13:00 ah okay 15:13:06 well when you can find it 15:13:09 do share 15:13:33 sure I'll ping you later today when I find it 15:13:34 should I leave the meeting open do you think? 15:13:38 sure thanks 15:14:52 well..the other thing I'd like to discuss is https://review.openstack.org/#/c/189762/ 15:15:09 but that is more infra related. 15:15:24 #link https://review.openstack.org/#/c/189762/ 15:15:40 do you want to share your thoughts for the log? 15:15:43 but it does impact 3rd party, especially towards migrating to a common-ci solution 15:16:14 the patch introduces a backwards compatibility issue 15:16:43 so anyone deploying nodepool and not using the latest puppet scripts would need to create the new secure.conf file 15:17:17 my ci-s & infra are not affected b/c we're using the latest puppet-nodepool scripts. 15:17:55 clarkb mentioned last week since nodepool version < 1.0 we don't need to bump the version number per semver 15:18:02 #link http://semver.org/ 15:19:32 there's also some question (Joshua Hesketh) as to where to put the nodepool test-job. 15:20:28 awesome 15:20:38 jhesketh: is sitting to my right 15:20:38 so it'd be great if we can iron out those issues and get the patch merged.... 15:21:01 in about 5 more minutes after we finish going round the room I will get jhesketh to join us 15:21:32 o/ 15:22:03 it would be great to have yolanda pabelanger and jeblair as well 15:23:10 asselin_: hi there 15:23:15 asselin_: ah right, that one (just catching up) 15:23:35 asselin_: so I have an opinion about moving the test-job param, but it's just an opinion and not blocking on it at all 15:23:36 yay 15:23:51 so asselin_ has some thoughts as he has outlined above 15:24:10 anyone else care to share their perspective on asselin_'s goal and the patch linked? 15:24:52 what's the question? :) 15:25:09 asselin_: do you want to clearly form a question? 15:25:48 the goal is to get the patch merged, but there are some questions regarding the test job and backwards incompatibility 15:25:58 it's definitely backwards incompatible. 15:26:33 pabelanger, hi 15:26:39 but we make backwards incompatible changes a lot. and we'll keep doing it. but we can try to make sure people know about them. 15:26:40 o/ 15:27:32 and of course since we use puppet to continuously deploy, there will be a sequence of changes that stepwise implement the change without breakage 15:28:13 jeblair, right, for -infra & my 3rd party cis those changes have already merged (puppet-nodepool) I can look it up 15:28:38 #link https://review.openstack.org/#/c/188325/ 15:29:56 hi asselin, i'll be around for short time, but i'm available now 15:30:22 yolanda, we're discussing your nodepool secure.conf change 15:30:36 the test-job should not go in the secure.conf file 15:31:10 if we move it, it should probably be to associate it with a label. 15:31:16 ah yes, we decided to move to the nodepool yaml, with jenkins targets, i remember 15:32:07 so therefore, it ought to be a different change. i think it's a good idea, but we don't use it so it's not high on my list to do right now. 15:32:37 i think i said something to that effect in the comments already 15:32:48 but i can do it again if that is desired. 15:32:59 so as far as i know, i didn't include on secure.conf, but on yaml, is not like that? 15:33:14 jeblair, asselin_: yes, the discussion we had is here: https://review.openstack.org/#/c/189762/21/nodepool/nodepool.py 15:33:17 #link https://review.openstack.org/#/c/189762/21/nodepool/nodepool.py 15:34:21 Ya, I'd like to see if move forward. Breaking working nodepool configuration leaves some sour taste in my mouth. Any way to make secure.conf optional? 15:34:26 probably test-job would have been moved to label a long time ago if we actually used it. :) 15:35:23 my point about test-job was simply that rather than breaking compat twice, we do it all in one job 15:35:27 if not, I am fine with that. We just need to make sure everybody downstream knows of the change 15:35:34 but as jeblair points out, we don't promise non-breakage, so it's not overly important 15:35:38 pabelanger: nodepool isn't nearly well designed enough to carry around that kind of baggage :( 15:36:11 fair enough 15:36:19 this change is already complex enough to add more features to it 15:36:39 so i'd prefer to move test-job to an independent change 15:36:41 well, a version bump should be simple :) 15:37:11 i agree that we should bump version with that secure.conf addition, as we are breaking backward compatibility 15:37:45 Ya, don't get me wrong, I like the change. So I don't want to hold it up longer then already it has been. 15:38:41 i'm afraid of breaking downstream really. In our side, we consume a custom branch, and we manage syncs manually. But i'm not sure on all use cases, people consuming directly from upstream may have bad surprises 15:38:55 maybe sending an announcement to the mailing list, before approving, can work? 15:39:33 yeah, announce to -infra list would be prudent i thin 15:39:35 think 15:39:51 also, in zuul, we keep a NEWS file with notes on breaking changes 15:39:55 Ya, anything to give a heads up helps. For me, I have old clients that are running nodepool. So, if I can point them to a simple ML post, that helps 15:39:59 and -dev [third-party] 15:40:25 asselin_: let's focus these messages on -infra 15:40:27 that's what it's fore 15:40:45 ok, so this needs to be coordinated, maybe sending the announce on mailing list, with an exact day of the change 15:40:50 to give time for people to be prepared 15:41:01 and also point to the changes done in puppet to get a general idea on the new config 15:41:20 jeblair, I'm not sure how my third party folks subscribe to that list, but maybe that's something we can note that they do up-front. 15:41:49 sending that to opentack-dev, with the right topic, cannot hurt... 15:42:01 asselin_: i think that people using our tools should subscribe to it. it's not necessary information for third-party ci folks not using our tools. 15:42:23 yolanda: i think we should avoid sending every message to every list. let's try to keep them on topic. 15:42:29 jeblair, ok I can send the note to -dev and reference the -infra post 15:42:32 this is not related to openstack development 15:42:40 this is related to nodepool development 15:44:10 ok 15:44:42 so how do we move forward? 15:44:53 jeblair, ok, then we should ask third party folks to subscribe to -infra. I'll take that action. 15:45:02 asselin_: the ones using our tooling, yes :) 15:45:18 #action asselin_ ask third party operators using infra tooling to subscribe to -infra list 15:46:01 send announcement to -infra; merge change; (optional: move test-job); (optional: start a NEWS file in nodepool repo); (optional: release with version bump) 15:46:04 yolanda: ^ 15:46:55 jeblair, how do you want to schedule that? maybe giving a week between the announcement and the deploy? 15:48:11 yolanda: no, i think immediate is fine. we don't promise that we won't break CD systems running master. if someone wants to run our tools CD, they should probably be using puppet and keeping on top of things. 15:49:36 are we getting to a way forward? 15:49:47 jeblair, you will take care? or do you want me to send the announcement to the ml? 15:49:49 anyone what to propose a plan of action? 15:50:05 anteaya: i thought i did. 15:50:11 great 15:50:14 yolanda: would you please send it? 15:50:27 so jeblair has proposed a plan of action, can we all agree to it? 15:50:42 jeblair, sure 15:50:47 will it be landing today? 15:55:19 jeblair, asselin, so i need to step out for a while. I'll come back later and write some announcement, first i'll send you for review. Are you ok with that? 15:55:29 yolanda: sounds good, thanks 15:55:29 pabelanger as well ^ 15:55:32 ok 15:55:37 yolanda, agree 15:55:43 ack 15:56:44 I do believe I have to end the meeting 15:56:46 anteaya, I think that's it. we can take the rest offline. We have a plan forward now. 15:56:53 thanks everyone 15:56:53 have we reached a happy place? 15:56:56 thank you 15:57:05 sorry I was multitasking with the qa sprint 15:57:06 thanks all, nice to have it moving forward 15:57:13 thanks all for contributing to the conversation 15:57:15 I appreciate it 15:57:19 #endmeeting