16:04:14 <akuznetsova> #startmeeting Mistral 16:04:15 <openstack> Meeting started Mon Nov 23 16:04:14 2015 UTC and is due to finish in 60 minutes. The chair is akuznetsova. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:04:16 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:04:17 <nmakhotkin> hi everyone! 16:04:19 <openstack> The meeting name has been set to 'mistral' 16:04:24 <bhaskarduvvuri> hi 16:04:27 <_gryf> hi 16:04:31 <xylan_kong> hi, guys 16:04:31 <akuznetsova> hi everyone 16:04:31 <^Gal^> hi 16:04:41 <LimorStotland> hi 16:05:20 <akuznetsova> give me a sec, I'll find last meeting minutes 16:05:20 <ddeja> hello 16:05:28 <nmakhotkin> :) 16:05:54 <xylan_kong> seems we have some new faces today 16:06:44 <akuznetsova> yeah, it's cool 16:06:54 <akuznetsova> let's start 16:07:17 <akuznetsova> #action items review 16:07:58 <akuznetsova> #topic 16:08:10 <akuznetsova> #topic action items review 16:08:29 <xylan_kong> :-) 16:08:36 <akuznetsova> 1. rakhmerov: create Mitaka series and milestone in 16:08:36 <akuznetsova> python-mistralclient LP 16:09:06 <akuznetsova> nmakhotkin, do you know, Renat did it or not? 16:09:48 <nmakhotkin> I see mistralclient ver 1.2.0 16:09:57 <xylan_kong> yeah, me too 16:10:14 <akuznetsova> ok, it means that this ai is done 16:10:23 <nmakhotkin> AFAIK client versions differ from version of main project 16:10:41 <akuznetsova> 2. rakhmerov, melisha: design "Mistral HA" and "Multi-region support" during M-1 16:11:10 <nmakhotkin> is melisha with us? 16:11:13 <akuznetsova> I am not sure that I see any docs/etherpads 16:11:29 <LimorStotland> I can update on it : they didn't do it yet 16:11:46 <xylan_kong> someone asked me about this topic today 16:11:50 <akuznetsova> LimorStotland, thanks 16:11:51 <nmakhotkin> so, let's forward action to next meeting 16:11:59 <akuznetsova> yes 16:12:05 <nmakhotkin> #action rakhmerov, melisha: design "Mistral HA" and "Multi-region support" during M-1 16:12:06 <LimorStotland> ok 16:12:16 <akuznetsova> thanks, nmakhotkin 16:12:21 <nmakhotkin> np :) 16:12:45 <akuznetsova> that's all action items from previous meeting 16:12:58 <rakhmerov> hi guys 16:13:09 <akuznetsova> hi rakhmerov 16:13:11 <rakhmerov> I reviewed OpenStack HA meeting log btw 16:13:29 <akuznetsova> #topic Current status (progress, issues, roadblocks, further plans) 16:13:43 <rakhmerov> please keep going... 16:13:45 <rakhmerov> I'll catch up 16:14:14 <nmakhotkin> my status: did a lot of review last week 16:14:16 <^Gal^> I'm on mistraldashboard pagination in execution screens 16:14:41 <LimorStotland> my status: i am working on internal alcatel-lucent issues + mistral reviews 16:14:47 <nmakhotkin> btw, nice to see ceilometer and trove actions in mistral :) 16:15:02 <xylan_kong> - completed the implementation of mistral workflow sharing, https://review.openstack.org/244155 [READY FOR REVIEW] 16:15:02 <xylan_kong> - fixed some bugs found during workflow sharing feature testing. [MERGED] 16:15:02 <xylan_kong> - completed mistral-spec initialization, I think it's a big improvement for Mistral as an official project. [MERGED] 16:15:03 <xylan_kong> - had an experience with Reno, gave a sharing to the whole team via email. [DONE] 16:15:03 <xylan_kong> - review work [ALWAYS] 16:15:28 <xylan_kong> nmakhotkin: yeah, I'll do zaqar integration in next few days 16:15:30 <^Gal^> xylan_kong: nice! 16:15:43 <rakhmerov> my status: reviews, backporting (not finished) and some email exchanges 16:15:43 <nmakhotkin> xylan_kong: awesome! 16:15:44 <akuznetsova> my status: worked on improving our Jenkins ci, split one gate into two separate gates, I am planning to continue this work and make gates voting 16:16:08 <xylan_kong> nmakhotkin: thanks 16:16:13 <LimorStotland> xylan_kong nice !!! 16:16:31 <akuznetsova> xylan_kong, oh, you came prepared ) 16:16:50 <xylan_kong> akuznetsova: yes, of course 16:16:58 <akuznetsova> xylan_kong, :-) 16:17:01 <xylan_kong> a habit 16:17:26 <akuznetsova> does anybody have any roadblocks? 16:17:48 <xylan_kong> i have one, but can be discussed in open discussion 16:18:01 <xylan_kong> about how we identify resource 16:18:12 <akuznetsova> xylan_kong, ok, so let's move to open discussion ! 16:18:17 <xylan_kong> akuznetsova: ok 16:18:25 <akuznetsova> #topic Open Discussion 16:18:43 <akuznetsova> xylan_kong, please, go ahead 16:18:56 <rakhmerov> I also wanted to discuss status of backporting but not sure if we need to do it now 16:19:03 <xylan_kong> the problem was found during the implementation of workflow sharing feature 16:19:18 <rakhmerov> about a half of bugs that ALU wants got merged into stable/liberty 16:19:19 <xylan_kong> #link https://blueprints.launchpad.net/mistral/+spec/use-workflow-id-in-rest-api 16:19:47 <xylan_kong> I want to talk about the bp/bug, for fixing that, maybe we need to consider re-design our Mistral API 16:19:49 <rakhmerov> some of them can't be merged because we need to repair a sequence of commits in stable/liberty 16:19:58 <rakhmerov> I'm planning to work on it tomorrow first thing 16:21:12 <LimorStotland> xylan_kong : i am not sure i understand... can you please give example or something ? 16:21:26 <xylan_kong> Without uuid, we can hardly distinguish different workflows with same name, especially when we use 'public' workflows or we want to get benifits from workflow sharing feature I have been implementing these days. 16:21:32 <akuznetsova> xylan_kong, what do you mean by re-design ? api v3 ? 16:21:40 <xylan_kong> .akuznetsova maybe 16:22:19 <xylan_kong> since it seems a big change for mistral, especially related to API 16:22:35 <xylan_kong> I have already met some bugs related to that, please see https://bugs.launchpad.net/mistral/+bug/1518655 and https://bugs.launchpad.net/python-mistralclient/+bug/1518276 16:22:35 <openstack> Launchpad bug 1518655 in Mistral "Cron-trigger running failed when workflow's scope is updated to 'private' from 'public'" [Undecided,New] 16:22:36 <openstack> Launchpad bug 1518276 in python-mistralclient "Can not distinguish private workflow and public workflow with the same name" [Medium,Fix committed] - Assigned to Lingxian Kong (kong) 16:22:37 <bhaskarduvvuri> that sounds like a nice idea xylan_kong 16:23:13 <xylan_kong> currently, different tanants can have workflows with same name 16:23:20 <rakhmerov> xylan_kong: yes, I fully understand you 16:23:31 <xylan_kong> rakhmerov :-) 16:23:32 <rakhmerov> I think it's a subject of API v3 design 16:23:43 <akuznetsova> I don't remember why we decided to use names instead on uuid 16:23:52 <rakhmerov> but also agree that we need to fix it in a backwards compatible way in v2 16:24:01 <xylan_kong> rakhmerov: yes, absolutely 16:24:13 <xylan_kong> we can both support name and uuid 16:24:17 <bhaskarduvvuri> or maybe have v2.1 16:24:17 <nmakhotkin> yes, we can use name or uuid 16:24:30 <bhaskarduvvuri> for backport 16:24:38 <rakhmerov> akuznetsova: because we wanted save API calls when we needed to do something with workflows 16:24:57 <rakhmerov> so that you don't need to lookup a workflow by name first in order to find its id 16:25:05 <rakhmerov> and then, for example, to execute it 16:25:32 <rakhmerov> assuming that names are unique you can run them immediately without having to know their ids 16:25:37 <rakhmerov> that was the idea 16:26:00 <akuznetsova> oh, I see 16:26:07 <nmakhotkin> names are human-readable :) 16:26:16 <nmakhotkin> and rememberable 16:26:16 <rakhmerov> nmakhotkin: indeed 16:26:34 <akuznetsova> but now in case of public wfs it became a problem 16:26:40 <xylan_kong> nmakhotkin: but it is also changable 16:27:00 <LimorStotland> xylan long do you have any design for it? 16:27:01 <xylan_kong> and it's part of ReST API 16:27:07 <nmakhotkin> xylan_kong: really? 16:27:12 <rakhmerov> actually, it can be solved also if we print out their "scope" field and "name + scope" is a unique constraint 16:27:16 <xylan_kong> LimorStotland: yes, i think it's not a hard thing todo 16:27:25 <nmakhotkin> I don't remember there is a possibility to change a name of workflow 16:27:47 <rakhmerov> xylan_kong: yes, I'm actually not sure we can change it 16:27:50 <rakhmerov> let me see.. 16:28:03 <xylan_kong> we can't change the name 16:28:42 <rakhmerov> we can't 16:28:45 <rakhmerov> that's right 16:29:04 <rakhmerov> because /workflows/name PUT accepts plain/text only 16:29:20 <rakhmerov> which is YAML 16:29:27 <LimorStotland> so can please publish the design... its a bit hard (at least for me to see the problem, for me it's sound like we only need to add workflow-id to the get ....) 16:29:52 <xylan_kong> rakhmerov: that works. but as a user, maybe I will be confused that i can't design a workflow with a name i wanted if you add the constraint 16:30:08 <rakhmerov> yes 16:30:14 <rakhmerov> well, after all I think it may have been a mistake to identify workflows by names 16:30:16 <xylan_kong> LimorStotland: yes, it deserves a spec for review 16:30:29 <LimorStotland> xylan_kong thanks 16:30:38 <xylan_kong> so, what's you suggestion about when we can start to do that? 16:30:49 <xylan_kong> in M? or N? 16:30:56 <rakhmerov> on one hand, it's a little bit more convenient in certain cases but on the other hands ids are much more flexible 16:31:06 <rakhmerov> xylan_kong: doing what? 16:31:15 <xylan_kong> rakhmerov: use uuid 16:31:26 <xylan_kong> or adding uuid support 16:31:30 <xylan_kong> in rest api 16:31:38 <rakhmerov> xylan_kong: whenever API v3 comes through, in my opinion 16:31:55 <rakhmerov> which I hope will happen in M 16:32:07 <xylan_kong> since i have to solve the problem in the resource sharing feature design 16:32:16 <rakhmerov> well, it's just my thought, I'm not sure about this suggestion 16:32:29 <akuznetsova> probably will be more changes in api v3, so I guess that we need to design it first 16:32:30 <rakhmerov> it really should be designed first and reviewed like LimorStotland said 16:32:48 <xylan_kong> rakhmerov: yes, agreed 16:32:51 <rakhmerov> akuznetsova: yes 16:32:54 <rakhmerov> xylan_kong: yes 16:33:23 <rakhmerov> xylan_kong: so you actually sent some patches with showing id, right? 16:33:23 <xylan_kong> so maybe i can have the first commit in mistral-spec :-) 16:33:30 <xylan_kong> rakhmerov: yes 16:33:32 <rakhmerov> I voted +2 for them 16:33:43 <xylan_kong> rakhmerov: but that's just some workaround for nwo 16:33:56 <xylan_kong> can not solve the whole problem totally 16:34:05 <rakhmerov> xylan_kong: ok, but do we need anything else for now? 16:34:11 <rakhmerov> I know that's a workaround 16:34:53 <xylan_kong> the problems still exist 16:35:13 <akuznetsova> yes 16:35:18 <xylan_kong> and we have to wait until we support uuid in rest api 16:35:36 <akuznetsova> let's create etherpad with new API design 16:35:58 <xylan_kong> akuznetsova: no need etherpad, just specification in mistral-spec 16:36:00 <rakhmerov> xylan_kong: I mean is it a roadblock for resource sharing implementation now? 16:36:08 <akuznetsova> xylan_kong, oh, yes 16:36:58 <xylan_kong> rakhmerov: yes, when i share a workflow to a tenant, he can't see the workflow details if he has a workflow with the same name 16:37:24 <rakhmerov> why? 16:37:26 <xylan_kong> in implementation, we don't know which workflow detail the tenant want to show 16:37:33 <rakhmerov> give me a sec.. thinking.. 16:37:39 <xylan_kong> with the only workflow name 16:38:00 <xylan_kong> given we have 2 tenant2, A and B. 16:38:05 <rakhmerov> yes 16:38:09 <xylan_kong> A creats a workflow, and shars with B 16:38:14 <rakhmerov> what do you do to share a workflow? 16:38:15 <rakhmerov> yes 16:38:30 <xylan_kong> blablabla.... 16:38:43 <xylan_kong> rakhmerov: you already know , right? 16:38:47 <rakhmerov> what's the sharing algorithm? 16:38:52 <rakhmerov> no I don't 16:39:08 <rakhmerov> more precisely, I remember what we discussed at the summit but didn't look at your impl yet.. 16:39:17 <xylan_kong> rakhmerov: https://blueprints.launchpad.net/mistral/+spec/mistral-resource-sharing 16:39:45 <xylan_kong> yes, you can take a look at the code first, including the sharing api design 16:40:00 <xylan_kong> especially the test cases 16:40:16 <LimorStotland> i think that what xylan_kong is saying is that it can create a situation where a user can run 2 different wf with the same name 16:40:28 <rakhmerov> xylan_kong: yes, I saw the BP, just please explain your algo in 2 words 16:40:36 <rakhmerov> you create a permissions table? 16:40:51 <xylan_kong> i added a new table to record the sharing relationship 16:41:12 <xylan_kong> the table can be used for workflow, workbook, and action 16:41:27 <xylan_kong> and any other resources in future that can be sharable 16:41:43 <rakhmerov> ooh, ok. I see now the problem 16:41:53 <rakhmerov> correct me if I'm wrong.. 16:42:01 <xylan_kong> ok 16:42:55 <rakhmerov> so even if we shared a workflow from A to B and users from B can see the WF in listings they still can't use it because they need to use unique names 16:43:10 <xylan_kong> rakhmerov: yes, exactly 16:43:20 <rakhmerov> that are not unique anymore because A and B can contain wfs with same names 16:43:25 <rakhmerov> yep... 16:43:29 <rakhmerov> f...k 16:43:39 <xylan_kong> calm down :-) 16:43:40 <rakhmerov> it's a serious issue :) 16:44:18 <rakhmerov> it definitely need to change but I just wonder if we can workaround it for now somehow 16:44:41 <xylan_kong> rakhmerov: yes, we can support name and uuid at the same time 16:44:50 <xylan_kong> name is just for backward compatible 16:45:00 <xylan_kong> and we recommend users to use uuid in future 16:45:05 <xylan_kong> maybe in M release 16:45:16 <rakhmerov> you mean we can change all that methods taking names to also support ids? 16:45:24 <xylan_kong> rakhmerov: yes 16:45:27 <rakhmerov> yeah, that should work 16:45:36 <akuznetsova> I guess that it is good variant 16:45:44 <rakhmerov> I actually I know some technologies that use that approach 16:45:52 <rakhmerov> for example, StackStorm does that 16:46:11 <rakhmerov> xylan_kong: so that is what you're going to work on? 16:46:22 <rakhmerov> create a spec for it? 16:46:26 <xylan_kong> and in release notes, we must be honest to tell users what's the restriction about using name 16:46:33 <xylan_kong> rakhmerov: sure, i can go with that 16:46:44 <rakhmerov> xylan_kong: yes, please do 16:46:50 <rakhmerov> because anyway it's related to your task 16:46:56 <xylan_kong> rakhmerov: yes 16:47:00 <rakhmerov> ok, deal 16:47:09 <xylan_kong> i can start with a specification 16:47:56 <akuznetsova> yes, please 16:48:16 <rakhmerov> ok 16:48:17 <xylan_kong> will call for review when i finish 16:48:21 <xylan_kong> :-) 16:48:29 <akuznetsova> anything else to discuss ? 16:48:47 <akuznetsova> rakhmerov, ? 16:48:50 <rakhmerov> I'm done 16:48:54 <ddeja> akuznetsova, rakhmerov: I have a short question, do you have any progress on https://bugs.launchpad.net/mistral/+bug/1513456 Today I was able to dummy fix it (using sleep), but have idea for better solution. I'm wondering if I can reassigne it to myself. 16:48:54 <openstack> Launchpad bug 1513456 in Mistral "task stuck in RUNNING state when all action executions are finished" [Critical,Triaged] - Assigned to Renat Akhmerov (rakhmerov) 16:49:04 <rakhmerov> and I would even say "I'm completely finished" :) 16:49:13 <nmakhotkin> AFAIK we don't 16:49:25 <nmakhotkin> rakhmerov? 16:49:25 <xylan_kong> rakhmerov :-) 16:49:26 <akuznetsova> yes, we don't fix it 16:49:28 <rakhmerov> ddeja: please do 16:49:32 <rakhmerov> feel free 16:49:45 <ddeja> ok, will do so :) 16:49:54 <rakhmerov> ddeja: just one question 16:50:17 <rakhmerov> were you just able to reproduce it or you were able to understand the cause of the issue? 16:50:30 <ddeja> both 16:50:34 <rakhmerov> I'd be rather interested to know where it comes from 16:50:35 <_gryf> rakhmerov, we were working on it last week 16:50:41 <rakhmerov> ddeja: ooh, awesome! 16:51:07 <rakhmerov> ddeja: can you please explain briefly what this issue is caused by? 16:51:16 <ddeja> It's a race condition in default engine 16:51:24 <rakhmerov> ok, where exactly? 16:51:30 <ddeja> in method 'on_action_complete' 16:51:33 <_gryf> rakhmerov, ddeja, it could be good, to make a comment under the bug on lp 16:51:44 <rakhmerov> yes, definitely 16:51:59 <rakhmerov> ok, let me put it this way: I'd like to know all the guts of this issue 16:52:01 <rakhmerov> :)) 16:52:12 <ddeja> when transaction is started, it sometimes (in my case - always) didn't finish 16:52:27 <ddeja> before transaction is started in second thread 16:52:27 <nmakhotkin> transaction hangs?? 16:52:46 <ddeja> so, both transactions didn't see that the other action has ended 16:53:10 <nmakhotkin> yep, I also investigated that a bit 16:53:47 <rakhmerov> hm.. 16:53:48 <rakhmerov> weird 16:53:55 <ddeja> it can be fixed by not checking in code if every action has ended, but leave it to database 16:53:58 <noa_koffman> I saw the same thing... 16:54:01 <rakhmerov> why didn't that TX finish? 16:54:24 <ddeja> it did finished 16:54:31 <rakhmerov> ddeja: as far as checks it makes sense, yes 16:54:32 <ddeja> but before it happend, second started 16:54:44 <rakhmerov> oooh! I think I got it 16:55:03 <rakhmerov> it's because an ORM model in memory differs from DB state.. 16:55:07 <rakhmerov> right? 16:55:09 <ddeja> yes 16:55:14 <rakhmerov> gotcha 16:55:17 <akuznetsova> folks, we have only 3 mins left 16:55:23 <ddeja> OK :) 16:55:25 <xylan_kong> ddeja: good job! welcome your contribution btw 16:55:29 <rakhmerov> then the fix should be pretty simple 16:55:43 <rakhmerov> ddeja: yeah, awesome job! Thanks 16:55:55 <akuznetsova> let's wrap our meeting 16:55:59 <rakhmerov> reassign it to yourself and ask any help if needed 16:56:00 <ddeja> xylan_kong, rakhmerov, thanks :-) 16:56:07 <ddeja> OK 16:56:11 <akuznetsova> we can continue in our usual channel 16:56:12 <rakhmerov> it's an extremely important issue 16:56:29 <rakhmerov> ok, now I'm completely finished 16:56:35 <rakhmerov> ok to end the meeting 16:56:38 <nmakhotkin> bye! 16:56:41 <akuznetsova> bye 16:56:42 <xylan_kong> ok, productive meeting 16:56:43 <rakhmerov> let'd meet in our channel 16:56:44 <rakhmerov> bye 16:56:44 <xylan_kong> bye, guys 16:56:49 <_gryf> bye 16:56:53 <noa_koffman> bye 16:56:54 <ddeja> bye 16:56:57 <akuznetsova> #endmeeting