14:00:33 <jed56> #startmeeting watcher
14:00:33 <openstack> Meeting started Wed Apr 13 14:00:33 2016 UTC and is due to finish in 60 minutes.  The chair is jed56. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:34 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:37 <dtardivel> hi
14:00:37 <openstack> The meeting name has been set to 'watcher'
14:00:43 <jed56> hi everyone !
14:00:45 <tkaczynski> hi
14:00:46 <gzhai1> hi
14:00:49 <vtech> hi
14:00:51 <jinquan> hi
14:00:52 <edleafe> \o
14:00:59 <tpeoples> o/, on another call right now so may be slow to respond
14:01:01 <jed56> Agenda for today #link https://wiki.openstack.org/wiki/Watcher_Meeting_Agenda#13.2F06.2F2016
14:01:04 <sballe_> o/
14:01:20 <alexchadin> hi
14:01:31 <jed56> #topic Announcements
14:01:48 <jed56> #info We could run tempest tests with multiple nodes. The multiple nodes testing allows us to use tempest to cover the features which require multiple nodes in gate job.
14:01:54 <jed56> thanks vincentfrancoise
14:02:21 <jed56> This is an important feature for to improve test cases of watcher
14:02:30 <jed56> In fact, most of the strategies of watcher proposes the action “live-migration or cold migration” of a VM.
14:02:34 <vmahe1> hi
14:02:36 <vtech> jed56, is it already operational?
14:02:42 <jed56> For the moment, this gate is experimental.
14:02:46 <vincentfrancoise> vtech: yes
14:02:50 <jed56> We plan to change the state of this gate to “no-voting” after the Openstack Summit.  If you want to trigger the gate, you should leave a comment on the review  “check experimental” in order to run/add the job in the pipeline.
14:03:04 <vtech> cool
14:03:10 <jinquan> great
14:03:13 <vincentfrancoise> see https://review.openstack.org/#/c/296633/ (gate-watcher-dsvm-multinode-nv)
14:03:16 <alexchadin> cool
14:03:54 <jed56> everybody can push  code ! go go go
14:04:03 <jed56> :-)
14:04:03 <vincentfrancoise> :p
14:04:11 <jed56> any other announcement ?
14:04:45 <jinquan> BP:https://blueprints.launchpad.net/watcher/+spec/select-destinations-filter code completed, i will it push tomorrow.
14:05:03 <jinquan> push it
14:05:06 <jed56> jinquan: okay awesome
14:05:16 <jed56> #topic Review Action Items
14:05:53 <jed56> there are many specifications which are waiting a second core reviewer agreement for merging
14:06:09 <jed56> Like last week, I would like that these specifications are merged as soon as possible.
14:06:19 <jed56> Scoring Module  #link https://review.openstack.org/#/c/289880/
14:06:19 <tkaczynski> +1
14:06:27 <jed56> Provide efficacy indicators #link https://review.openstack.org/#/c/283449/
14:06:33 <sballe_> jed56: I +2 an intel spec whihc I know is not a good practice. But I waiting for b-com to +1 it
14:06:33 <jed56> Add Overload standard deviation #link https://review.openstack.org/#/c/286153/
14:06:49 <seanmurphy> apols - i said i would review one or two but i have been overloaded
14:07:15 <jed56> Who could take some time to look at these reviews in priority, please ?
14:07:25 <jed56> seanmurphy: do you need a load balancer ?
14:07:28 <jed56> :)
14:07:38 <seanmurphy> haha ;-)
14:07:42 <seanmurphy> i can do something this week
14:07:44 <sballe_> jed56: give ma an actions to review stuff
14:07:49 <jed56> sballe_: acabot is in holidays for the moment
14:07:56 <seanmurphy> i will recheck what i was supposed to review now
14:08:20 <jed56> #action acabot, jwcropper sballe review #link https://review.openstack.org/#/c/283449/  #link https://review.openstack.org/#/c/286153/ https://review.openstack.org/#/c/275474/
14:08:54 <jed56> junjie: sorry i didn't take the time
14:08:57 <jed56> to review https://review.openstack.org/#/c/275474/
14:09:11 <jed56> but it is in my pipeline
14:09:23 <jed56> #action jed56 review https://review.openstack.org/#/c/275474/
14:09:31 <tkaczynski> what about scoring module?
14:09:55 <jinquan> hi, jed56  we have a question about junjie's BP: #link https://review.openstack.org/#/c/292188/ , we need some helps
14:09:56 <jed56> tkaczynski: we are waiting a +1 from acabot or joe
14:10:12 <jed56> jinquan: okay, can you wait open discussion please
14:10:13 <tkaczynski> ok, thanks
14:10:20 <jinquan> ok
14:10:26 <jed56> jinquan: thanks :)
14:10:27 <seanmurphy> prevoiusly i said i would review https://review.openstack.org/#/c/283449/ but if this is assigned to others i can take something else
14:10:45 <jed56> tpeoples: do you plan to update https://review.openstack.org/#/c/287019/
14:10:51 <tpeoples> jwcroppe has been on a long vacation for the past few weeks. he'll be back next week
14:11:08 <jed56> tpeoples: great!
14:11:12 <sballe_> jed56: they are both on vacation but it would be nice to get a +1 from tpeoples  and some mroe +1 from b-com
14:11:17 <jed56> seanmurphy: okay
14:11:27 <tpeoples> yes jed56. should i abandon and put up for review under newton?
14:11:33 <sballe_> jed56: I meant on the scoring engine
14:11:41 <jed56> you should postpone to newton
14:11:49 <tpeoples> will do
14:11:55 <sballe_> tpeoples: can you review the scroing engine and +1 it assuming you are ok with it?
14:12:07 <jed56> #action tpeoples review  https://review.openstack.org/#/c/289880/
14:12:22 <tpeoples> sballe_: yes, i can do that
14:12:28 <jed56> okay great!
14:12:29 <sballe_> tpeoples: thx
14:13:11 <sballe_> jed56: you already +1 it. Can we get vincentfrancoise or dtardivel to review it too?
14:13:41 <sballe_> this will make life easier for acabot and jwcroppe when they het back and have to +2 it
14:13:49 <tkaczynski> sballe_: I believe they already had that on their action list
14:13:50 <vincentfrancoise> sballe_: I'm a bit overloaded right now with my BP
14:14:01 <jed56> dtardivel is very busy
14:14:15 <jed56> it plan to review https://review.openstack.org/#/q/status:open++branch:master+topic:bp/get-goal-from-strategy
14:14:21 <sballe_> fair enough. I was just seeing how we could speed up the +2
14:14:23 <vincentfrancoise> so that would be EOW if I get some spare time
14:14:40 <jed56> #action review https://review.openstack.org/#/c/289880/
14:15:02 <jed56> vmahe : do you think you could review the scoring spec ?
14:15:16 <jed56> and the others : )
14:15:24 <sballe_> yes of course ;-)
14:15:24 <vmahe1> yes, I can do that
14:15:26 <seanmurphy> i can do it also
14:15:30 <sballe_> vmahe1: thx!
14:15:36 <sballe_> seanmurphy: thx!
14:15:49 <seanmurphy> sballe_: np!
14:15:54 <sballe_> :)
14:15:56 <jed56> #action vmahe review   https://review.openstack.org/#/c/289880/   https://review.openstack.org/#/c/286153/
14:16:08 <jed56> #action seanmurphy review  https://review.openstack.org/#/c/286153/
14:16:12 <tkaczynski> thanks guys :)
14:16:14 <jed56> thanks you
14:16:26 <sballe_> +1
14:16:40 <jed56> who wants to review https://review.openstack.org/#/q/status:open++branch:master+topic:bp/get-goal-from-strategy ?
14:16:43 <jed56> 11 patchsets :)
14:17:10 <jed56> this is important to us because
14:17:23 <jed56> many blueprints are depending of this one
14:17:32 <gzhai1> I can have a look
14:17:32 <jed56> this is the critical path :)
14:17:35 <dtardivel> jed56: I will do it
14:17:44 <tkaczynski> I can review that. it will be a good lesson for my implementation :)
14:17:46 <sballe_> ditto
14:18:10 <jed56> #action gzhai1 sballe_ dtardivel tkaczynski  review https://review.openstack.org/#/q/status:open++branch:master+topic:bp/get-goal-from-strategy
14:18:21 <vincentfrancoise> FYI, I've put some WIP but they'll be removed soon
14:18:26 <gzhai1> one question. should we address it one by one?
14:18:49 <jed56> you should address them by order
14:18:57 <vincentfrancoise> well yes
14:19:08 <jed56> these patchets are linked
14:19:41 <vincentfrancoise> I tried to make them as readable as possible but you may not understand some bits in the beginning
14:20:05 <jed56> don't hesitate to ask vincentfrancoise on #openstack-watcher
14:20:11 <gzhai1> vincentfrancoise: do you want reorg them?
14:20:24 <jed56> if you  have any questions
14:20:32 <vincentfrancoise> so feel free to ask me on #openstack-watcher if you wish to
14:20:41 <jed56> vincentfrancoise: :)
14:21:07 <tpeoples> vincentfrancoise:  i will take another pass at all of those this week
14:21:11 <vincentfrancoise> gzhai1: not really I already changed the order so it shouldn't move much from now on
14:21:20 <gzhai1> ok
14:21:29 <jed56> #action tpeoples  review https://review.openstack.org/#/q/status:open++branch:master+topic:bp/get-goal-from-strategy
14:21:36 <vincentfrancoise> I'm currently fixing some minor bugs as I am making some testing
14:21:50 <vincentfrancoise> hence the -1 WIPs
14:22:10 <jed56> any other questions ?
14:22:29 <jed56> okay let's continue
14:22:31 <jed56> #topic Blueprint/Bug
14:22:42 <jed56> #info the specifications for Newton are open
14:22:50 <jed56> https://review.openstack.org/#/c/301753/ has been merged
14:22:57 <jed56> #info everybody can retarget their own specification to newton
14:23:22 <jed56> who have a specification open to mitaka ?
14:23:47 <jed56> i can take a look to lauchpad :)
14:25:02 <alexchadin> there are almost all opened to newton except of two.
14:25:10 <jed56> #action acabot,sballe, jwcropper , jed56 set comment on specs on mitika
14:25:32 <alexchadin> https://blueprints.launchpad.net/watcher/+spec/optimization-threshold
14:25:41 <alexchadin> one already:)
14:25:59 <jinquan> retarget just a moment ago :)
14:26:08 <jed56> okay great!
14:26:34 <jed56> #topic open discussion
14:26:47 <jed56> There are many specifications which have been reviewed several times but they need a core reviewer agreement for merging
14:26:50 <jed56> Do we need to add a new core reviewer to watcher-specs ?
14:27:15 <alexchadin> i may try this role
14:27:42 <jed56> alexchadin: we should have an election
14:27:46 <sballe_> I think we need to wait for acabot since he is the PTL
14:27:51 <jed56> we need to wait acabot
14:27:55 <sballe_> lol
14:27:59 <vincentfrancoise> sballe_: +1
14:28:14 <jed56> sballe +1 :)
14:28:28 <jed56> #action acabot :  Do we need to add a new core reviewer to watcher-specs ?
14:28:55 <alexchadin> one more question: how may i join watcher driver team at launchpad? through election?
14:29:20 <dtardivel> alexchadin: no, I can add you directly
14:30:30 <jed56> #action dtardivel add alexchadin  in watcher-drivers
14:31:09 <alexchadin> i'm in pending state currently
14:31:09 <jed56> tkaczynski: is working  on the implementation of the scoring engine
14:31:32 <jed56> we have question on the implementation on this use case
14:31:40 <jed56> As a developer.
14:31:40 <jed56> I want to be able to list the available Scoring Engines. So that I can quickly
14:31:40 <jed56> identify them and reuse the available predicted results in my strategy, e.g.
14:31:40 <jed56> prediction energy consumption of the VMs, predicted CPU of the VMs, etc.”
14:31:43 <edleafe> jed56: is there a candidate for core reviewer you have in mind?
14:31:55 <jed56> edleafe: no
14:32:26 <edleafe> jed56: ah. Usually cores are selected based on mastery of the topic, not just to get a higher number
14:32:39 <jed56> but we have problem to merge specs :)
14:32:53 <dtardivel> alexchadin: done. you should receive a confirmation email
14:32:54 <jed56> it is maybe due to two core are in holidays
14:33:16 <jed56> edleafe: okay
14:33:18 <tkaczynski> jed56: actually, this is the use case:
14:33:20 <edleafe> jed56: understood, but the idea is that you don't make someone core without them first demonstrating an appropriate knowledge level
14:33:27 <tkaczynski> As a developer.
14:33:27 <tkaczynski> I want to be able to provide a dynamic list of Scoring Engines in a single
14:33:27 <tkaczynski> plug-in. So that I can register/unregister similar types of Scoring Engines
14:33:27 <tkaczynski> without restarting any Watcher service.
14:33:30 <alexchadin> Great!
14:33:40 <jed56> This use case involve a way to discover the available scoring engines.
14:33:46 <jed56> It seems that we have two options:
14:33:49 <jed56> option 1: extend the RPC API on decision engine (so API calls DE directly)
14:33:54 <jed56> option 2: use watcher database so decision engine keeps database up to date and the watcher API get the data from there.
14:33:58 <edleafe> jed56: I would first find someone whose opinions you trust, and then propose them for core
14:34:30 <jed56> edleafe: we will discuss that when acabot will return
14:34:40 <jed56> Does somebody want to give his/her opinion ?
14:34:44 <edleafe> jed56: ok, great
14:35:16 <tkaczynski> there are important implications for each use case, not sure if we have time to lay them all here, but I'm proposing option 1
14:35:49 <jed56> tpeoples, edleafe: do you have an opinion ?
14:36:25 <edleafe> jed56: the DB option seems cleaner
14:36:30 <jed56> vtech: h
14:36:49 <jed56> edleafe: i also prefer the option2 to avoid strong coupling
14:37:00 <jed56> between th api and de
14:37:05 <vmahe1> vmahe: the same for me
14:37:07 <tkaczynski> the biggest problems with option 2: need to keep DB in sync and complicates the implementation of the scoring modules (every developer will have to deal with DB I think)
14:37:44 <jed56> tkaczynski: whay every developer will have to deal with the DB ?
14:37:51 <jed56> *why
14:38:07 <tkaczynski> option 1: data is always in sync, much more simpler implementation of scoring module without dealing with internals of Watcher
14:39:37 <tkaczynski> jed56: because every scoring engine plugin will have to handle registration/deregistration of the scoring engines. the list will be dynamic, so some sort of event handling will have to be in place and will be part of the interface / abstract class for each scoring engine
14:39:38 <vincentfrancoise> I'd rather go for option 2 since I already use this approach to sync goals and strategies into the DB
14:40:01 <jinquan> With the  function 's increase, maybe  it is difficult to avoid deal with DB ?
14:40:36 <tkaczynski> actually it's much easier to not use DB I think
14:40:37 <jed56> tkaczynski: we should add a class in charge of this taks
14:40:54 <jed56> this is not the responsability of the plugins
14:41:22 <edleafe> The main determinant is how often this information changes. Making repeated RPC calls only to get the same information is poor design. That's what DBs are for.
14:41:35 <vtech> very quickly looking at the options I would go with option 1 as well but I am afraid I am not able to see all the consequences.
14:42:15 <jed56> We can continue the debate on the openstack-dev mailing list with  the watcher tag.
14:42:21 <tkaczynski> DBs are very bad for integration. the smallest change and you have to update all the clients / parties using a given table
14:42:29 <jed56> to let everbody the time to think about it
14:42:55 <jed56> everbody agree ?
14:42:56 <tkaczynski> I can prepare an initial email and try to explain the context a bit more. what is the email I should use for that?
14:43:32 <vincentfrancoise> tkaczynski: openstack-dev@lists.openstack.org
14:43:40 <vtech> tkaczynski +1
14:43:57 <tkaczynski> ok. and subject with "[watcher]" prefix ?
14:44:05 <jed56> #action tkaczynski  send a mail to openstack-dev@lists.openstack.org with watcher prefix
14:44:09 <jed56> tkaczynski: yes
14:44:25 <jed56> any other open discussions ?
14:44:28 <tkaczynski> will I get this email too, or I need to register somewhere?
14:44:33 <jinquan> i have question:
14:44:33 <jinquan> migration with target host will broken many instance's important attributes like affinity, anti-affinity.
14:44:33 <jinquan> So, There are concerns about this bug: #link https://bugs.launchpad.net/nova/+bug/1427772 ?
14:44:35 <openstack> Launchpad bug 1427772 in OpenStack Compute (nova) "Instance that uses force-host still needs to run some filters" [Low,Confirmed] - Assigned to Anant Kaushik (anantkaushik-nsit)
14:44:36 <vincentfrancoise> you can take this email as an example: http://lists.openstack.org/pipermail/openstack-dev/2016-March/089244.html
14:44:40 <jed56> #link http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
14:44:48 <tkaczynski> jed56: thanks
14:45:38 <jed56> jinquan: i looking
14:46:36 <vincentfrancoise> If you look at https://bugs.launchpad.net/nova/+bug/1427772/comments/3
14:46:37 <openstack> Launchpad bug 1427772 in OpenStack Compute (nova) "Instance that uses force-host still needs to run some filters" [Low,Confirmed] - Assigned to Anant Kaushik (anantkaushik-nsit)
14:46:50 <vincentfrancoise> It's not meant to be fixed anytime soon
14:47:01 <jinquan> This bug a long time without treatment, need i try fix?
14:47:12 <jed56> it seems that this is not a priority
14:47:26 <vincentfrancoise> jinquan: not really because the cores do not agree on how to fix it
14:47:38 <sballe_> lol
14:47:50 <jed56> we can maybe discuss that wwith sbauza
14:48:12 <edleafe> The proposal to run scheduler on migrations is also slowed down
14:48:22 <jed56> jinquan: IMHO, this is very neer with select_destinations
14:48:24 <edleafe> More complicated interactions than first imagined
14:48:26 <jed56> near
14:48:45 <jed56> #action acabot take a look to https://bugs.launchpad.net/nova/+bug/1427772
14:48:47 <openstack> Launchpad bug 1427772 in OpenStack Compute (nova) "Instance that uses force-host still needs to run some filters" [Low,Confirmed] - Assigned to Anant Kaushik (anantkaushik-nsit)
14:48:52 <sballe_> I feel this whole nova scheduler topic is a problem
14:48:59 <sballe_> and has always been
14:49:07 <sballe_> and now it is hard to fix
14:49:10 <edleafe> sballe_: Heh, tell me about it
14:49:11 <jed56> sballe +1
14:49:17 <jed56> huhu
14:49:19 <vincentfrancoise> +1
14:49:33 <jinquan> oh  i see
14:49:36 <edleafe> I have been ranting about it since I first wrote parts of it 6 years ago
14:50:02 <jed56> edleafe: thanks !
14:50:09 <sballe_> :) I told Sandy thta they were doing it wrong at the Essex summit :)
14:50:30 <edleafe> in fact, I'm working on yet another ranty blog post about the scheduler direction
14:50:38 <sballe_> lol
14:50:48 <jed56> What is Essex summit ?
14:50:49 <sballe_> let me know if you want any input
14:51:02 <sballe_> Essex summit was on 2010
14:51:09 <jed56> ah !
14:51:22 <jed56> woah long time !
14:51:25 <sballe_> or maybe 2011..it was in Boston
14:51:36 <jinquan> I have a processing method about this bug
14:51:41 <jinquan> and  i will try
14:51:47 <jinquan> to fix it
14:52:00 <edleafe> Essex summit was spring 2012
14:52:11 <jed56> jinquan: you can try. it is good to be brave
14:52:12 <sballe_> ok so since then :)
14:52:36 <edleafe> sballe_: Sandy and I clashed on that more than once. He ended up convincing more people than I did
14:53:04 <jinquan> jed56 thks, just try :)
14:53:09 <sballe_> yeah he wasn't convinced about my apporach either so I decided to do something else. i hate politics
14:53:30 <jed56> any other  discussions or I can close the meeting ?
14:53:35 <edleafe> sballe_: there is little joy in "I told you so"s
14:53:45 <jed56> lol
14:53:48 <sballe_> I agree
14:54:04 <sballe_> because we are still stuck with the bad scheduler
14:54:36 <jed56> bye
14:54:45 <sballe_> already?
14:54:45 <jinquan> bye
14:54:56 <sballe_> talk to you later... ttyl
14:54:59 <sballe_> bye
14:55:03 <jed56> sballe_: do you wants yo add something ?
14:55:07 <vtech> bye
14:55:08 <sballe_> nope
14:55:09 <jed56> *want
14:55:14 <jed56> #endmeeting