05:30:56 #startmeeting senlin 05:30:57 Meeting started Fri Oct 12 05:30:56 2018 UTC and is due to finish in 60 minutes. The chair is dtruong. Information about MeetBot at http://wiki.debian.org/MeetBot. 05:30:58 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 05:31:01 The meeting name has been set to 'senlin' 05:31:12 it's time for senlin meeting 05:31:16 anyone here to join? 05:34:58 hi 05:35:19 hi qiming 05:36:19 ok, let's get started then. 05:36:52 #topic autoscaling forum in berlin 05:37:24 #link http://lists.openstack.org/pipermail/openstack-dev/2018-September/134770.html 05:37:40 Qiming: I saw your replies to the ML 05:37:54 do you have anything else to add on this topic? 05:39:07 If you mean something I want to stress again, I would say ... users first 05:39:42 yes, i agree with that 100% 05:40:29 If there are people who just want to find something to work on for a particular project or play games ... we can just ignore them 05:41:22 i think the important thing to recognize is that there are different types of users as well 05:41:37 yes 05:42:25 there are users of heat ASG or senlin. but then there is k8s users who want autoscaling but have different use cases 05:43:06 yep. I would vote for supporting the k8s use cases 05:44:33 a k8s cluster can live and scale happily on google cloud or aws, it would be desirable to have it better supported on openstack 05:46:07 again, at the end of the day, the community agenda should be driven by use cases 05:46:10 yes. but i think it is important to support the use cases of clustering or autoscaling with only openstack 05:46:26 that is the traditional scaling use case 05:46:39 if no one seems interested in running k8s on openstack, the above assumption won't be true 05:46:53 right 05:47:15 back to the proposed forum topic 05:48:02 IIUC, the key pain point from Heat team's perspective is that "there are existing users already doing autoscaling using ASG..." 05:49:03 a valid concern but they won't reinvent a new thing seriously, i guess 05:49:04 yes, so i proposed that you could potentially switch the Heat ASG implementation to use Senlin 05:49:26 and then slowly deprecate Heat ASG in favor of Senlin resource in Heat 05:49:29 exactly 05:50:07 long live Heat ASG resource type, but .... just kill the implementation 05:50:21 to me this follows what other openstack projects have done. 05:50:31 e.g. cinder split from nova-volume 05:50:45 if there are feature requirement to autoscaling, the requirements should be redirected to Senlin team 05:50:58 yes 05:51:32 Heat should only cares about orchestration, period 05:51:32 i think we are in agreement. so i'll discuss that proposal with the Heat team at the forum 05:52:11 ok, moving on to the next topic then 05:52:26 #topic blueprint status 05:53:04 fail-fast blueprint is implemented. waiting for more reviews. 05:53:19 multiple detection type implementation has been merged. 05:53:28 i need to add documentation and release notes. 05:53:41 cool 05:54:18 about the documentation, i thinking we should add a sphinx extension so that we can automatically generate docs for policy and profiles 05:54:37 could you please double check if the gate failures are real for #602460? 05:54:38 using the schema defined in the policy and profile 05:54:46 https://review.openstack.org/#/c/602460/ 05:55:07 that is a great idea 05:55:15 about the gate failure, the health policy tempest are broken 05:55:21 seems to be a timing issue 05:55:30 so i merged a patch to disable them for now 05:55:38 i see. that is the reason I didn't w+1 it 05:55:44 https://review.openstack.org/#/c/609563/ 05:55:49 understood 05:56:47 let me recheck #602460 to make sure that the failures are gone with my patchset to disable health policy tempest test 05:58:10 looks like you did already =) 05:58:58 going back to the sphinx extension, i want to extract the description in the schema for each field 05:59:01 impatient me 05:59:26 and build it into the docs 05:59:41 so that we don't have to change the docs every time the policy or profile changes 05:59:58 I have no idea how to do that, although I have the same requirement in a different project 06:00:11 ya, that would be awesome 06:00:45 my requirement is to generate web UIs from the class definition 06:00:56 it's new for me too. so i'm looking into it right now. mainly because i rather write code than documentation =P 06:01:28 great, looking forward to your findings 06:02:04 That's great 06:02:13 hi xuefeng 06:02:34 hi dtruong Qiming 06:02:49 you scared me XueFeng 06:02:58 sphinx has the ability to create your own extension so hopefully it won't be too difficult 06:03:15 suddenly showed up from nowhere 06:03:17 Why? 06:03:33 em 06:03:33 HAHA 06:04:07 so it is basically walking through the class field definitions and to extract the docstrings for those fields? 06:04:12 yes 06:04:40 that was my guess, but it looked to me not so straightforward how to get the doc string for a class field 06:04:47 actually, extract the description in the schema field 06:05:16 https://github.com/openstack/senlin/blob/master/senlin/policies/health_policy.py#L119 06:05:18 okay ... that is a much easier job 06:05:35 sphinx can already get the docstring 06:06:21 that have something called autodoc built-in that extract docstring for classes and their attributes and members 06:07:00 em ... 06:07:01 but in our case that is not that useful because we keep the usage information inside the description string 06:07:04 like this line? http://git.openstack.org/cgit/openstack/openstacksdk/tree/openstack/clustering/v1/profile.py#n36 06:07:47 yes, that would get extracted by sphinx 06:08:13 okay, maybe I should try dig into it again 06:08:28 but it's too generic for profile spec and policy spec 06:09:10 the policy spec is a dictionary with lots of parameters and the description string has the most useful information 06:09:28 indeed 06:10:12 the so called "doc" can be extracted from python variable/field values, that is making things a lot easier 06:10:36 yes 06:11:24 I'm looking for something that can help extract data from comments such as: 06:11:53 # max=1200,optional=true 06:12:10 repeatable 06:12:42 sorry, repeat_interval is more appropriate here 06:13:25 then by reading the source code I can generate a jsonschema for this particular field and even web form input for validation 06:13:49 or else I will have to do the schema definition again ... boring 06:15:05 if the comment is formatted with the #: then sphinx can extract using autoattribute 06:15:06 http://www.sphinx-doc.org/en/master/usage/extensions/autodoc.html#directive-autoattribute 06:15:52 I see, there is the trick 06:16:00 so maybe you can reuse the autodoc implementation to generate a web form instead of a doc 06:16:12 em, worth a try 06:18:36 ok, let's move on 06:18:45 #topic stein community goals 06:19:10 no updates for python 3 goal or upgrade checkers goal 06:20:10 #topic reviews 06:20:34 i added functional tests for senlin in openstacksdk 06:20:37 #link https://review.openstack.org/#/c/607061/ 06:20:48 please review it when you have time 06:21:56 wow ... I still have w+1 privilege there ... 06:21:57 i only added the basic tests for senlin but that should be enough to detect any breaking change in openstacksdk that could affect senlin 06:22:35 I see 06:22:43 a good time to use that privilege =) 06:22:51 :D 06:23:15 #topic open discussion 06:23:23 that's all i had an the agenda 06:23:33 anybody have anything else to discuss> 06:23:37 ? 06:23:59 no from me 06:24:49 XueFeng has vapored into thin air again 06:25:01 lol 06:25:13 ok, let's end the meeting then 06:25:19 thanks everyone for attending 06:25:33 thx 06:25:37 #endmeeting