16:00:47 <rakhmerov> #startmeeting Mistral 16:00:51 <openstack> Meeting started Mon Jun 29 16:00:47 2015 UTC and is due to finish in 60 minutes. The chair is rakhmerov. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:53 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:55 <openstack> The meeting name has been set to 'mistral' 16:03:05 <rakhmerov> anyone here? 16:03:10 <rakhmerov> m4dcoder? 16:03:11 <devkulkarni> hi rakhmerov 16:03:17 <rakhmerov> hi 16:03:19 <rakhmerov> how are you? 16:03:21 <ruhe> o/ 16:03:28 <devkulkarni> doing fine 16:03:29 <xylan_kong> rakhmerov: hi 16:03:38 <rakhmerov> xylan_kong, hi 16:04:29 <rakhmerov> ok, let's start the meeting 16:04:37 <xylan_kong> ok 16:04:37 <rakhmerov> the agenda is at: https://wiki.openstack.org/wiki/Meetings/MistralAgenda 16:04:48 <rakhmerov> #topic Review action items 16:05:09 <rakhmerov> there was just one AI 16:05:11 <rakhmerov> NikolayM: discuss yaql 1.0 with Murano team 16:05:20 <rakhmerov> NikilayM, you're here? 16:05:55 <xylan_kong> didn't met him today 16:06:23 <rakhmerov> ruhe, do you happen to know where we are going to have YAQL 1.0 in global requirements? 16:06:37 <rakhmerov> ...where -> when ... 16:06:40 <ruhe> rakhmerov: JFYI the main driver of yaql 1.0 adoption in Murano, Stan Lagun, is currently on a long sick leave, so there was not much progress in this area recently in Murano 16:06:55 <rakhmerov> ooh, ok 16:07:07 <rakhmerov> so we'll keep this AI active 16:07:27 <ruhe> yeah. Murano team keeps this item as a must have in Liberty 16:07:32 <rakhmerov> #action NikolayM: talk to Stan Lagun about YAQL 1.0 inclusion into global requirements 16:07:45 <rakhmerov> ruhe, ok, got it 16:07:48 <rakhmerov> that's good 16:08:14 <rakhmerov> we just wanted to switch to it asap because it's kind of a part of Mistral DSL from a user perspective 16:08:19 <rakhmerov> but that's ok 16:08:40 <rakhmerov> #topic Current status (progress, issues, roadblocks, further plans) 16:08:51 <rakhmerov> let's quickly report our statuses, as usually 16:09:13 <xylan_kong> my status 16:09:21 <xylan_kong> * get rid of openstack/common package, to comply with the whole upstream behavior 16:09:31 <xylan_kong> * fix some bugs and review work 16:09:37 <xylan_kong> * write a proposal for removing utils.wf_trace module 16:10:02 <rakhmerov> xylan_kong, is that openstack/common task considered completed? 16:10:09 <xylan_kong> yes, sure 16:10:17 <rakhmerov> ok, perfect 16:10:45 <rakhmerov> as far as that wf_trace module, what's wrong with it? 16:10:54 <rakhmerov> why do you want to replace it? 16:11:08 <rakhmerov> we might have discussed it with you a while ago 16:11:20 <xylan_kong> it's ok :-) just a little redundant IMO 16:11:20 <rakhmerov> I don't remember though 16:11:43 <rakhmerov> ok, so when are you planning to come up with your proposal? 16:11:52 <xylan_kong> just a little weird we add a module to implement a feature that oslo.log has 16:12:16 <rakhmerov> #action xylan_kong: proposal to replace utils.wf_trace with oslo.log 16:12:28 <xylan_kong> ok 16:12:45 <rakhmerov> ok, I think it's more or less clear, just want to see some details on that 16:12:57 <rakhmerov> my status: implemented "Workflow variables" and released Liberty 1 last week, also discussed and reviewed a bunch of changes with the team 16:13:25 <rakhmerov> I made a mistake with client versioning on Friday so had to fix it today 16:13:35 <rakhmerov> spent a lot of time on that 16:13:43 <devkulkarni> +1 rakhmerov 16:13:54 <xylan_kong> yeah, i notice that, nice work 16:14:25 <rakhmerov> nice work?? you kidding? :) I actually deserve a serious punishment :) 16:14:31 <xylan_kong> we correct it anyway 16:14:43 <rakhmerov> anyway, it's ok now 16:14:47 <xylan_kong> :-) 16:15:01 <rakhmerov> devkulkarni: btw, sorry that I can't catch you in our IRC channel 16:15:02 <devkulkarni> rakhmerov: :) is the new client released to global-requirements? 16:15:11 <rakhmerov> you write usually when I'm asleep ) 16:15:40 <rakhmerov> devkulkarni: we need to fix and merge this patch: https://review.openstack.org/#/c/195035/ 16:15:52 <rakhmerov> hoping it'll be done very soon 16:15:53 <devkulkarni> rakhmerov: no worries.. I am in US Central 16:15:55 <rakhmerov> 1-2 days 16:16:09 <rakhmerov> yeah, I guess I'm 12 hours ahead of you ) 16:16:17 <devkulkarni> rakhmerov: ok, will keep an eye out on it 16:16:44 <rakhmerov> devkulkarni: so do you have a problem because of a Mistral client versioning as well? 16:16:56 <rakhmerov> you're working on solum, right? 16:17:04 <devkulkarni> rakhmerov: I guess so.. I had added an open dicussion item 16:17:09 <devkulkarni> yes, I am part of the solum team 16:17:13 <rakhmerov> ok 16:17:33 <devkulkarni> #link http://logs.openstack.org/17/194817/4/check/gate-solum-devstack-dsvm/309c35e/logs/devstacklog.txt.gz#_2015-06-26_13_33_36_869 16:17:44 <devkulkarni> is this issue similar to what you were seeing? 16:17:51 <rakhmerov> give me a sec.. 16:17:52 <rakhmerov> I'll check 16:17:54 <devkulkarni> sure 16:18:06 <xylan_kong> i saw it 16:18:11 <rakhmerov> yeah, exact same issue 16:18:17 <rakhmerov> it should be now ok 16:18:32 <NikolayM> hi everyone 16:18:38 <xylan_kong> NikolayM: hi 16:18:38 <rakhmerov> hey, NikolayM 16:18:42 <devkulkarni> rakhmerov: you mean it should not happen after the above mentioned patch is merged, right? 16:19:03 <rakhmerov> I kept that action item for you about figuring our YAQL 1.0 perspectives 16:19:24 <rakhmerov> devkulkarni: nope, it's not related with that patch 16:19:31 <devkulkarni> oh okay 16:19:35 <NikolayM> my status: Implementing new RPC layer, have an idea to not use pika but use kombu instead (because it is already in global-requirements) 16:19:56 <rakhmerov> it was happenning because of wrong tags in git history 16:20:20 <rakhmerov> devkulkarni: so basically I pushed 1.0.0.0b1 by mistake 16:20:48 <devkulkarni> rakhmerov: ok, got it. and you mentioned you fixed that this morning. got it. 16:20:57 <rakhmerov> then I pushed the right one, 0.3.0 but PBR can't build the project for 0.3.0 if there's a greater version (1.0.0.0b1) 16:21:02 <rakhmerov> yes 16:21:22 <rakhmerov> so, devkulkarni: keep in mind that the right version for the client now is 1.0.0 16:21:22 <devkulkarni> rakhmerov: ok, makes sense. thanks for addressing this issue. 16:21:27 <devkulkarni> yes 16:21:31 <rakhmerov> I sent out the info about that in ML 16:21:37 <rakhmerov> ok 16:21:45 <m4dcoder> hey, Renat. sorry i'm late. 16:21:45 <rakhmerov> devkulkarni: btw, I have a question for you 16:21:50 <devkulkarni> rakhmerov: sure 16:21:54 <rakhmerov> m4dcoder: hey! no problem 16:22:15 <rakhmerov> devkulkarni: does Solum still use Mistral API v1.0? 16:22:25 <rakhmerov> or switched to v2? 16:22:29 <rakhmerov> v2 is the latest 16:22:32 <devkulkarni> I think we are still on v1 16:22:36 <devkulkarni> when was v2 released? 16:22:45 <rakhmerov> pretty long time ago actually 16:22:59 <rakhmerov> the thing is that stable/kilo doesn't support v1 anymore 16:23:11 <rakhmerov> the corresponding code has been removed 16:23:16 <devkulkarni> I see.. I will have to go back and check. 16:23:21 <devkulkarni> oh okay. thanks for the heads up 16:23:35 <rakhmerov> so you may want to check that and let us know, we can help you switch to v2 16:23:40 <devkulkarni> I will get with the solum team to check where we are with it 16:23:43 <devkulkarni> sounds good 16:23:46 <rakhmerov> yeah 16:23:54 <rakhmerov> feel free to ask our help 16:24:00 <devkulkarni> sure thing. 16:24:03 <rakhmerov> deal 16:24:08 <devkulkarni> you guys have been pretty helpful :) 16:24:22 <rakhmerov> you're very welcome 16:24:33 <rakhmerov> m4dcoder: can you print your status quickly? 16:25:00 <m4dcoder> sure. working on the "resume" bp. 16:25:31 <m4dcoder> and then right now just waiting for team to help me with the mysql/psql patch (mysql version and deadlock in scheduler) 16:25:52 <rakhmerov> I see 16:26:07 <rakhmerov> sorry, I didn't read your last email carefully 16:26:12 <rakhmerov> about mysql 16:26:34 <rakhmerov> what I understood from it is that MySQL 5.5 doesn't support even milliseconds 16:26:51 <rakhmerov> but I don't know if we even need them 16:26:57 <rakhmerov> what do you think? 16:27:03 <m4dcoder> looks like it unless i'm mistaken. 16:27:07 <NikolayM> can we use different variants with 5.5 and 5.6? 16:27:30 <NikolayM> say, use fractional secs in 5.6 and do not use if 5.5 16:27:32 <rakhmerov> my whole point is: I'd like to be compatible with other OpenStack projects in that regard 16:27:47 <rakhmerov> NikolayM, yes, something like this 16:28:03 <rakhmerov> I just don't want to lose the ability to work on mysql 5.5 16:28:17 <m4dcoder> i think we do need msec at the minimum unless we have no choice. we can have actions/tasks launched msec about from each other. 16:28:46 <rakhmerov> otherwise, imagine that we installed OpenStack where all projects are tested agains 5.5 but we'll have to have another mysql 5.6 specifically for Mistral 16:29:20 <m4dcoder> if we are sticking with 5.5, then we will have to change how timestamp is compared in the models. because right now, the unit tests fail at self.assertEqual(created, fetched) where created has fractional seconds in timestamps. 16:29:51 <rakhmerov> I think it's just a wrong test case 16:31:27 <rakhmerov> ok, m4dcoder, in any case we still owe you a review for your last patch 16:31:53 <rakhmerov> as far as the deadlock, it's one of the things I'm going to tackle this week 16:32:07 <rakhmerov> myself or NikolayM 16:32:17 <rakhmerov> he also knows all the pitfalls 16:32:53 <rakhmerov> #action rakhmerov: check Winson's info about deadlocks occurring in scheduler 16:33:13 <rakhmerov> m4dcoder: btw, does that deadlock only happens with mysql? 16:33:44 <m4dcoder> it happends with psql 16:33:54 <rakhmerov> postgres? 16:34:00 <m4dcoder> postgresql 16:34:01 <m4dcoder> yes 16:34:04 <rakhmerov> ooh 16:34:16 <rakhmerov> msql is fine? 16:34:20 <rakhmerov> mysql 16:34:38 <m4dcoder> i don't recall. i was testing that mostly on postgresql. but i can try and let you know later today. 16:34:43 <rakhmerov> it's a usefull info actually 16:34:48 <m4dcoder> ok 16:35:00 <rakhmerov> yeah, the thing is that we run Mistral with mysql all the time and scheduler works 16:35:02 <rakhmerov> ok 16:35:18 <rakhmerov> then can you pls send us your configuration for both Mistral and Postgres? 16:35:30 <m4dcoder> sure i'll sent that to you later today 16:35:50 <rakhmerov> so that we could see all TX properties applied to your operations 16:35:51 <rakhmerov> ok, thanks 16:36:06 <NikolayM> I think it is by the reason of using engine directly in tests 16:36:28 <NikolayM> in tests, we even don't use the engine client but use engine 16:36:54 <NikolayM> and in end of each test we delete all stuff from the DB, right? 16:37:10 <rakhmerov> #action m4dcoder: send conf files for Mistral and Postgres to investigate deadlocks in scheduler 16:37:12 <NikolayM> isn't it a block for DB? 16:38:06 <NikolayM> what do you think, guys? 16:38:10 <rakhmerov> NikolayM: it's supposed to be but if it really happens only for Postgres then I believe Winson's Postgres DB is not just propertly configured 16:38:15 <rakhmerov> as we require in scheduler 16:38:44 <NikolayM> ooh, I guess it also happens for mysql 16:38:54 <rakhmerov> may be 16:38:59 <m4dcoder> If that's the case, then you'll need to provide some instruction on how to properly configure it. 16:39:12 <rakhmerov> m4dcoder: yes, our fault 16:39:33 <rakhmerov> NikolayM: your assumption about not using engine client also looks interesting 16:39:40 <m4dcoder> no one's fault. just need some clarification and i'm reporting what i'm observing. 16:40:12 <rakhmerov> #info Do we need to use engine client in unit tests instead of engine directly? 16:40:21 <rakhmerov> m4dcoder: yes, sure 16:40:24 <rakhmerov> thanks 16:40:33 <rakhmerov> anyway, I'm sure we'll get it done soon 16:40:41 <devkulkarni> my topic in open-discussion has been already covered. need to run.. signing-off. thanks rakhmerov for the help. 16:40:56 <rakhmerov> it's the interesting and important issue but the fix should be simple 16:41:06 <rakhmerov> devkulkarni: thanks for coming! 16:41:39 <rakhmerov> ok, what else do we have on a plate... 16:41:50 <rakhmerov> #topic Duplicating messages on executors 16:42:17 <rakhmerov> unfortunately, we don't have the author of this topic today with us 16:42:24 <xylan_kong> rakhmerov: what's that mean? 16:42:32 <xylan_kong> do you know that? 16:42:34 <rakhmerov> but I'd like to ask you wether you faced it or not? 16:42:40 <rakhmerov> yeah, xylan_kong, let me explain 16:42:45 <xylan_kong> ok 16:43:27 <rakhmerov> so folks from alcatel lucent (Limor, Moshe, Noa) say that when Mistral engine puts a message into MQ (to run action) then more than one executor can poll it from MQ 16:43:38 <rakhmerov> e.g. 2 16:43:54 <rakhmerov> I don't really know how the heck that may happen 16:44:05 <rakhmerov> but maybe you came across this issue 16:44:09 <rakhmerov> m4dcoder: did you? 16:44:49 <xylan_kong> rakhmerov: i can test it tomorrow with the latest code 16:45:01 <rakhmerov> oooh, wait a second 16:45:16 <rakhmerov> I actually got an email from them with their configuration and topology 16:45:23 <rakhmerov> it says they use QPID 16:45:26 <rakhmerov> not RAbbit 16:45:46 <xylan_kong> i heard from Limor about this issue, but i don't meet in fact 16:46:20 <rakhmerov> ook, looks like it's qpid 16:46:23 <rakhmerov> not rabbit 16:46:23 <xylan_kong> yes, they use qpid, i guess rabbitmq doesn't have this problem 16:46:37 <rakhmerov> yep, then it should be tested with qpid 16:46:54 <rakhmerov> maybe qpid should be tuned somehow, dunno 16:47:24 <rakhmerov> alright, anyway, we could spend some time investigating it 16:47:37 <xylan_kong> i'm not familiar with qpid, are you? 16:47:45 <rakhmerov> I'm not either 16:47:52 <rakhmerov> just heard about it briefly 16:48:14 <rakhmerov> I think it should be rather a BP than a bug though since we've never even thought about using qpid 16:48:17 <xylan_kong> anyway, i will test it again with rabbitmq to ensure it's ok with the latest code 16:48:35 <xylan_kong> maybe 16:49:01 <rakhmerov> well, as far as rabbit we would have noticed that because we run it mostly every day 16:49:19 <xylan_kong> yes, actually 16:49:32 <rakhmerov> what I mean is that we need to aim to make it work with qpid intentionally 16:49:34 <rakhmerov> it's a BP 16:49:59 <rakhmerov> #action rakhmerov: create a BP to make Mistral work with qpid behind oslo.messaging 16:50:34 <rakhmerov> ok, don't know what else to say on this topic 16:50:41 <rakhmerov> needs to be tested and investigated 16:51:10 <rakhmerov> I guess eventually we need to create a new gate that would run functional tests with qpid 16:51:33 <rakhmerov> #topic Liberty-2 planning 16:52:05 <rakhmerov> guys, also I was going to talk about Liberty 2 plans but looks like we're running out of time today 16:52:36 <rakhmerov> so my suggestion is to send a list of BPs and bugs that are most important for you 16:52:53 <rakhmerov> either to myself or even better to openstack-dev 16:53:13 <xylan_kong> ok, no problem 16:53:19 <rakhmerov> so that I could collect them, assign them etc. etc. 16:53:51 <rakhmerov> I, of cource, have a high-level picture but also want to hear from you 16:54:37 <rakhmerov> #action xylan_kong, m4dcoder, NikolayM, LimorStotland: send a list of desired BPs and bugs for Liberty-2 so that we could plan 16:55:03 <rakhmerov> #topic Open Discussion 16:55:16 <rakhmerov> m4dcoder: how is that pause/resume thing going? 16:55:32 <m4dcoder> rakhmerov: i have not seem the problem with multiple executors getting same action from rabbitmq. 16:55:57 <rakhmerov> ok, yes 16:56:29 <m4dcoder> good progress with pause/resume. 16:56:45 <rakhmerov> does that patch on review completes it? 16:56:57 <rakhmerov> you're planning to do something else? 16:57:33 <m4dcoder> however, my colleagues and i are debating (just now) whether we should let users change input data on resume. the patch was WIP to let dmitri see where i'm at. no need to review yet. 16:58:01 <m4dcoder> patch is WIP not complete yet. 16:58:21 <rakhmerov> yeah, I know 16:58:28 <rakhmerov> ok, keep us posted please 16:58:47 <rakhmerov> ok, guys, I think it's time to end our meeting 16:58:56 <rakhmerov> thanks a lot for coming! 16:59:00 <rakhmerov> see you next time 16:59:03 <xylan_kong> ok, see you 16:59:05 <rakhmerov> bye 16:59:08 <NikolayM> bye! 16:59:15 <rakhmerov> #endmeeting