14:00:30 <acabot> #startmeeting watcher
14:00:31 <openstack> Meeting started Wed Jan 13 14:00:30 2016 UTC and is due to finish in 60 minutes.  The chair is acabot. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:32 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:35 <openstack> The meeting name has been set to 'watcher'
14:00:39 <acabot> hi
14:00:43 <jed56> o/
14:00:46 <dtardivel> Hi
14:00:55 <tpeoples> o/
14:00:56 <vincentfrancoise> o/
14:00:59 <acabot> agenda for today #link https://wiki.openstack.org/wiki/Watcher_Meeting_Agenda#01.2F13.2F2016
14:01:27 <fhermeni> hi
14:01:34 <edleafe> \o
14:02:01 <cdupont> hi
14:02:07 <acabot> #topic Annoucements
14:02:39 <acabot> #info 1st draft of mid-cycle agenda is available https://wiki.openstack.org/wiki/Watcher_mitaka_mid-cycle_meetup_agenda#Agenda
14:02:41 <sballe> o/
14:03:07 <acabot> feel free to change timings and subjects
14:03:58 <sballe> acabot: Do you know if we can do working lunches? I think an hour for lunch is a long time.
14:04:37 <sballe> maybe we do lunch in 30 minutes
14:04:41 <acabot> sballe: thats my french spirit ;-)
14:04:49 <sballe> lol
14:05:04 <jed56> Maybe we can a topic regarding ceilometer ( performance, cache, .. )
14:05:11 <sballe> +1
14:05:19 <acabot> I will ask jwcroppe
14:05:54 <acabot> #action acabot add a topic regarding ceilometer to the agenda
14:05:55 <sballe> jed56: I would add that topic to open discussions. Not sure we 2 hours for open discussions
14:06:12 <acabot> the Nova mid-cycle will take place in Bristol in 2 weeks and they plan to discuss scheduler improvements https://etherpad.openstack.org/p/mitaka-nova-priorities-tracking
14:06:39 <jed56> sballe : I am just wondering what we're doing to make us think Ceilometer won't pose a performance problem? Is there some other caching that's happening somewhere?
14:06:49 <acabot> I wont be able to attend but we should look at the etherpads between the 2 meetings
14:07:21 <acabot> any other announcement ?
14:07:25 <sballe> acabot: I cannot go either to the Nova meeting. I hope jwcroppe can go
14:07:41 <edleafe> I won't be at the Nova mid-cycle either
14:07:48 <sballe> jed56: I think Ceilomter is a problem so I would like for us to discss that
14:08:07 <acabot> #topic Review Action Items
14:08:21 <sballe> jed56 I think we said it wasn't a problem for the poc since we weren't doing produciton liek stuff
14:08:34 <tpeoples> i don't think jwcroppe was planning on making it to the nova midcycle
14:08:41 <acabot> ok
14:09:06 <acabot> sballe: maybe we should discuss also Monasca integration ;-)
14:09:23 <jed56> acabot: +1
14:09:24 <sballe> ok. acabot since you guys are kind of local could you send somebody just so our voice is heard
14:09:39 <sballe> Monasca and Ceilosca should be considered.
14:10:21 <acabot> sballe: unfortunately I cant but I had a discussion with bauzas who will attend about our requirements, I'm also in discussion with him to submit a talk for the summit
14:10:37 <sballe> ok cool!
14:10:55 <acabot> lets move to Watcher specs now
14:11:13 <acabot> #info Intel POC specs have been merged
14:12:07 <acabot> #info dynamic actions specs have been merged and there is a lot of code moving right now
14:12:14 <sballe> yeah +1 thx to everybody who contributed to make that happen
14:12:42 <acabot> #link https://review.openstack.org/#/q/status:open+project:openstack/watcher+branch:master+topic:bp/watcher-add-actions-via-conf
14:13:01 <acabot> all these reviews will have to be merged at the same time
14:13:28 <acabot> jed56: do you think we can do it by end of week ?
14:13:29 <sballe> ok
14:13:50 <jed56> yes I think a waiting for last reviews
14:14:42 <acabot> there are still 2 specs open regarding consolidation grammar and constraints management
14:14:54 <acabot> sballe: did you have a look at them ?
14:15:09 <dtardivel> tpeoples: we waiting for another +2 on these patchsets ;)
14:15:10 <acabot> cdupont: do you handle both of them ?
14:15:15 <cdupont> yes
14:15:26 <cdupont> they are in our hands
14:15:34 <tpeoples> looking dtardivel
14:15:50 <sballe> acabot: I will look at them today/ I have been traveling and it messes up my todos
14:15:53 <cdupont> it's a proof of concept for a strategy using BtrPlace
14:16:04 <acabot> cdupont: as I said in my last comment, I dont think we should have our own system dealing with constraints
14:16:12 <cdupont> OK
14:16:13 <sballe> +1
14:16:31 <cdupont> I'm not sure I understood your comment :)
14:16:39 <cdupont> Where would the constraints come from?
14:16:46 <acabot> cdupont : Nova
14:17:04 <cdupont> but the implementation/solver could be in Watcher?
14:17:24 <jed56> cdupont: yes
14:17:28 <cdupont> OK
14:17:36 <cdupont> so this work should be paused?
14:17:41 <acabot> cdupont : yes but the way you have written specs, you dont get constraints from Nova
14:17:53 <cdupont> I see
14:17:59 <jed56> IMHO, we have to pull the constraints from (congress, nova, ... )
14:18:20 <jed56> and then use it we any strategies or sovler
14:18:24 <acabot> +1 and then deal with consolidation through algorithms in Watcher
14:18:27 <cdupont> how the constraints are solved at the moment?
14:18:38 <sballe> +1 on what jed56 said
14:18:48 <jed56> cdupont : we don't :p
14:18:54 <cdupont> I mean, this is an NP hard problem, so only using filters is limited
14:19:02 <jed56> we have to call nova filters  for the P.OC
14:19:23 <jed56> https://blueprints.launchpad.net/watcher/+spec/helper-nova-select-destinations
14:19:35 <jed56> cdupont : +1
14:19:53 <acabot> cdupont : we are looking at how to get these constraints from Nova #link https://blueprints.launchpad.net/watcher/+spec/helper-nova-select-destinations
14:19:56 <cdupont> using a solver like CP solver / linear integer etc. can lead to better placements
14:19:56 <edleafe> jed56: remember, though, there is no public interface to Nova's filtering
14:20:07 <acabot> cdupont: but its a tricky question
14:20:08 <jed56> edlefe : yes sadly
14:20:08 <edleafe> jed56: and there isn't any planned in the future
14:20:49 <jed56> I think we can deal with contraints without nova filter but something it can be nice to use it
14:20:53 <acabot> edleafe: right, we need to discuss it at the mid-cycle
14:21:06 <edleafe> the best way to work with nova is to select a host and then call live-migrate. If the host doesn't match the filter requirement, catch that and retry
14:21:07 <cdupont> So I'll wait that this question is stabilised to make progress
14:21:10 <bzhou> remeber, scheduler select_destinations() only returns one host per instance
14:21:19 <sballe> I agree this is a good topic for the mid-ccyle
14:21:20 <jed56> bzhou : +1
14:21:31 <sballe> bzhou: +1
14:21:35 <edleafe> bzhou: remember too that there is no way to call select_destinations from outside of Nova.
14:21:51 <acabot> cdupont: I will keep your BPs in discussion state for now
14:21:58 <cdupont> acabot: OK
14:22:10 <sballe> acabot: We should bring this up as an issues at the nova mid-cycle. can you pass this on to your contact
14:22:10 <bzhou> as we use live migration in watcher, we can leverage nova's bp: check_destination
14:22:22 <jed56> IMHO, the first thing is to find the best way to collect the constraints
14:22:33 <acabot> sballe: I will
14:22:36 <bzhou> which will fail if the specified target host cannot pass filters
14:22:41 <sballe> acabot: thx
14:22:42 <jed56> if we want to solve it
14:22:54 <edleafe> bzhou: yes, that seems to be the only way. Select a host, and let check_destination reject it if necessary
14:23:08 <acabot> #action acabot give bauzas a more detailed view of our requirements on Nova scheduler
14:23:35 <tpeoples> that's definitely going to have scale issues edleafe , but i understand currently that would be our only option.  i think we should at least try to work with nova to get something better
14:23:55 <bzhou> tpeoples: +1
14:23:59 <acabot> tpeoples : +1
14:24:06 <edleafe> tpeoples: agreed. But when I spoke with people there after last week's meeting, they didn't seem very eager
14:24:35 <edleafe> Unfortunately, the scheduler is strictly internal to Nova for the foreseeable future
14:24:38 <edleafe> :(
14:24:55 <bzhou> edleafe: I read your discussion. I heard Neutron also wants a similar feature?
14:25:19 <tpeoples> :( acabot  can we work with bauzas to understand if using the scheduler filters in the way we want is out of the picture?
14:25:25 <edleafe> Neutron, Cinder, and one or two others all have generic scheduling needs
14:25:35 <acabot> bzhou : yes Neutron has pretty much the same use case and they plan to discuss it full day during the nova mid-cycle
14:25:38 <sballe> tpeoples: +1
14:25:42 <edleafe> But nova's scheduler is specific to compute resources
14:25:47 <edleafe> it can't handle anything else
14:25:54 <sballe> +1
14:25:57 <edleafe> it even has a very hard time with ironic
14:26:05 <sballe> lol +1
14:26:09 <tpeoples> is that the goal of the super scheduler by harlowja, to fill this void?
14:26:23 <tpeoples> i haven't had time to look into that much
14:26:26 <edleafe> tpeoples: not familiar with that. Link?
14:26:40 <sballe> IMO we don't need a super scheduler We just need ApI to the various schedulers
14:26:42 <acabot> https://review.openstack.org/#/c/210549/
14:26:50 <jed56> #link https://review.openstack.org/#/c/210549/
14:27:13 <acabot> fhermeni is working on this review
14:27:17 <sballe> thx I hadn't seen that work
14:27:37 <acabot> it is done on "OpenStack global specs"
14:27:53 <tpeoples> no acabot, i don't want to go into the deep end with that super scheduler, just wanted to mention it.
14:28:01 <fhermeni> tpeoples: the purpose of the « super-scheduler » is to provide a unique entry point to every scheduling aspects. The implementation at the backend might handle the scheduling completely or split and forward to different schedulers (nova, neutron …)
14:28:07 <sballe> ok just added myself as a reviewer
14:28:55 <acabot> #info there is a super-scheduler wip https://review.openstack.org/#/c/210549/
14:28:55 <fhermeni> tpeoples: it just exposes an entry point where users specify their expectations in terms of scheduling
14:29:28 <acabot> sorry, I'm coming back to the agenda
14:29:35 <tpeoples> thanks fhermeni
14:29:45 <acabot> #info devstack plugin has been merged
14:29:52 <sballe> cool! +1
14:29:58 <sballe> great job tpeoples
14:30:11 <tpeoples> do we want to send a note to the ML about this ?
14:30:19 <sballe> definetly
14:30:21 <acabot> thx for that, it was one of our Mitaka priorities
14:30:34 <bzhou> tpeoples: great, thanks
14:30:37 <tpeoples> #action tpeoples send note to ML regarding devstack
14:30:43 <acabot> thx
14:31:15 <acabot> dtardivel started working on Watcher CLI
14:31:26 <acabot> to bring Watcher evolutions in our client
14:31:29 <dtardivel> tpeoples: we have to validate again live migration triggered by Watcher on devstack
14:31:44 <acabot> #info Watcher client launchpad is now set up https://blueprints.launchpad.net/python-watcherclient
14:31:59 <dtardivel> tpeoples: this is an opened bug on it :)
14:32:07 <tpeoples> link?
14:32:31 <tpeoples> ah yes
14:32:42 <dtardivel> #link https://bugs.launchpad.net/watcher/+bug/1531912
14:32:43 <openstack> Launchpad bug 1531912 in watcher ""MIGRATE" action triggered by Watcher Applier failed on devstack" [High,Confirmed] - Assigned to David TARDIVEL (david-tardivel)
14:32:47 <tpeoples> i have a workaround for that, but i was waiting for jed56's refactoring stuff to merge first
14:32:52 <acabot> now, when you submit a BP, please think about where it is the most suitable (Watcher or Watcher CLI)
14:33:33 <acabot> #topic
14:33:39 <acabot> #topic Blueprint/Bug Review and Discussion
14:33:44 <tpeoples> same as you dtardivel, so OK :)
14:34:10 <acabot> could everyone update the delivery field of its BPs ? https://blueprints.launchpad.net/watcher
14:34:39 <acabot> it would be much easier for me to track progress !
14:34:51 <dtardivel> acabot: +1
14:35:28 <tpeoples> will do
14:35:49 <acabot> #action jed56 seanmurphy vmahe acabot tpeoples vincentfrancoise bzhou update delivery field on launchpad for affected blueprints
14:36:20 <acabot> any update on BPs from alexstav ?
14:36:43 <tpeoples> doesn't look like he's here
14:36:56 <acabot> tpeoples : thx ;-)
14:37:22 <tpeoples> if i see him on irc i'll ask him during the US day
14:37:22 <acabot> we need to create a new repo for Watcher UI (in horizon)
14:37:57 <acabot> can we agree on naming it "Watcher dashboard" as it is the common name used by other openstack projects ?
14:38:11 <tpeoples> watcher-dashboard, watcher-ui
14:38:16 <tpeoples> i'm fine with either
14:38:16 <acabot> or anyone wants another name ?
14:38:20 <jed56> watcher-ui
14:38:48 <jed56> the other projects => monasca-ui, magnum-ui ,  zaqar-ui
14:38:56 <vincentfrancoise> watcher-dashboard
14:39:11 <tpeoples> :D
14:39:12 <vincentfrancoise> but doesn't matter to me really
14:39:15 <sballe> I like to follow the rest of openstack
14:39:16 <acabot> jed56 : ok murano-dashboard, cloudkitty-dashboard ;-)
14:39:22 <jed56> yes  : )
14:39:26 <sballe> lol
14:39:31 <jed56> vincentfrancoise:+1
14:40:04 <dtardivel> I prefer watcher-ui. this is the name used internaly
14:40:05 <sballe> dashboard is probably better in case we later get a ops-console
14:40:17 <sballe> ui can be dashboard and console
14:40:35 <sballe> but I can go either way
14:41:16 <acabot> #startvote Watcher UI name ? watcher-ui, watcher-dashboard, dont care
14:41:17 <openstack> Begin voting on: Watcher UI name ? Valid vote options are watcher-ui, watcher-dashboard, dont, care.
14:41:18 <openstack> Vote using '#vote OPTION'. Only your last vote counts.
14:41:32 <dtardivel> #vote watcher-ui
14:41:34 <acabot> #vote watcher-dashboard
14:41:35 <vincentfrancoise> #vote watcher-dashboard
14:41:44 <brunograz> #vote watcher-dashboard
14:41:48 <sballe> #vote watcher-dashboard
14:41:55 <jed56> #vote watcher-ui   :p
14:41:55 <tpeoples> #vote watcher-ui
14:41:56 <openstack> jed56: watcher-ui   :p is not a valid option. Valid options are watcher-ui, watcher-dashboard, dont, care.
14:41:59 <edleafe> #vote watcher-dashboard
14:42:03 <jed56> #vote watcher-ui
14:42:13 <bzhou> #vote watcher-dashboard
14:42:26 <acabot> #showvote
14:42:27 <openstack> watcher-dashboard (6): bzhou, vincentfrancoise, sballe, acabot, edleafe, brunograz
14:42:28 <openstack> watcher-ui (3): tpeoples, dtardivel, jed56
14:42:35 <fhermeni> #vote watcher-ui
14:42:43 <tpeoples> dashboard it is
14:42:48 <sballe> :)
14:42:50 <acabot> #endvote
14:42:51 <openstack> Voted on "Watcher UI name ?" Results are
14:42:51 <tpeoples> even with fhermeni ^
14:42:52 <openstack> watcher-dashboard (6): bzhou, vincentfrancoise, sballe, acabot, edleafe, brunograz
14:42:54 <openstack> watcher-ui (4): tpeoples, dtardivel, jed56, fhermeni
14:43:10 <acabot> lets go with watcher-dashboard
14:43:12 <acabot> thx
14:43:15 <jed56> good luck dtarvidel for the refactoring !
14:43:22 <tpeoples> haha
14:43:25 <sballe> ups!
14:43:29 <jed56> I was in your side :p
14:43:30 <dtardivel> jed56: :(
14:43:40 <fhermeni> :D
14:43:46 <sballe> dtardivel: sorry about that :(
14:44:18 <acabot> #info bugs status for mitaka-2 https://launchpad.net/watcher/+milestone/mitaka-2
14:44:55 <acabot> we should freeze mitaka-2 around January 20th
14:45:38 <acabot> #info we will need someone on this one https://bugs.launchpad.net/watcher/+bug/1527189
14:45:39 <openstack> Launchpad bug 1527189 in watcher "Update the documentation to explain how to change the ceilometer API backend" [High,New]
14:46:23 <acabot> #topic Open Discussion
14:46:31 <acabot> 14 minutes left
14:46:47 <sballe> Can we talk about the talk submisisons
14:46:50 <acabot> Proposed talks are still in review https://etherpad.openstack.org/p/Watcher_abstracts_austin2016
14:47:00 <acabot> and sballe wants to submit them asap
14:47:14 <acabot> can we submit them by end of week ?
14:47:41 <sballe> I just need the email addresses people use with openstack.
14:47:48 <sballe> please add them to the etherpad
14:48:05 <acabot> ok
14:48:20 <acabot> global requirements https://bugs.launchpad.net/watcher/+bug/1533282
14:48:21 <openstack> Launchpad bug 1533282 in watcher "Watcher should use openstack global requirements" [Undecided,New] - Assigned to Taylor Peoples (tpeoples)
14:48:21 <sballe> Who else should be onthe "Watcher a Resource Manager for OpenStack: Plans for the N-release and beyond" talk
14:48:30 <sballe> I have acabot, jwcroppe and sballe
14:48:42 <tpeoples> i'd be interested in talking for the demo regarding devstack sballe
14:49:09 <acabot> tpeoples: put your name on the hands-on lab
14:49:16 <sballe> +1
14:49:27 <tpeoples> alright
14:49:43 <tpeoples> dtardivel and I discussed yesterday and agreed that we should be included in openstack/requirements project
14:49:45 <sballe> for the hands on we will need more people to help.
14:49:56 <tpeoples> which basically makes sure our requirements are always in sync with openstack
14:50:02 <tpeoples> 's global requirements
14:50:11 <vincentfrancoise> +1
14:50:15 <tpeoples> but it looks like we'll need to move our requirements to master, not liberty
14:50:18 <tpeoples> are we OK with that?
14:51:01 <jed56> what are the difference  ?
14:51:06 <jed56> the version are more updated ?
14:51:11 <tpeoples> just different versions essentially
14:51:31 <tpeoples> i can change them locally and do a smoke test and make sure nothing major is broken
14:51:36 <dtardivel> tpeoples: what is the risk ? :)
14:51:40 <sballe> acabot: tpeoples I'll check on the status of the etherpad on friday and submit EOD friday. ok?
14:51:50 <tpeoples> nothing really, i think we need to do it dtardivel
14:51:57 <tpeoples> so i'll move forward
14:52:03 <sballe> acabot: Can you add jed56 or dtardivel  to the hands on lab?
14:52:03 <acabot> sballe: submit 2 talks for the summit before friday EOD
14:52:54 <jed56> tpeoples: +2
14:52:57 <acabot> sballe: not clear about how many people we will have at the summit
14:53:39 <acabot> any other discussion ?
14:54:35 <sballe> acabot: understood. In the past various companies were only sending people ot Openstack that had a talk so this is why I have been pushing for adding people :)
14:54:49 <acabot> cdupont : do you think we could add btrPlace as a new strategy in Watcher ?
14:54:57 <acabot> sballe: thats a good way to do ;-)
14:55:09 <cdupont> I think so
14:55:13 <sballe> Rackspace and HP are operating taht way
14:55:26 <sballe> and I believe so is Intel
14:55:36 <cdupont> fhermeni might also have an opinion on that
14:56:01 <cdupont> but we need to translate the Nova constraints to BtrPlace constraints
14:56:07 <acabot> cdupont fhermeni : would you be interested in working on this ?
14:56:45 <fhermeni> acabot: it would be great but on my side, it is too soon. I am working on the scheduler spec for the moment
14:56:46 <cdupont> Actually in CN the next months will be very hard
14:56:52 <acabot> cdupont : yes, we need to solve the constraints pb before but as btrPlace is in Java, it will take time to have it in Python
14:57:05 <cdupont> yes
14:57:35 <fhermeni> acabot: I will have agenda issues as well :D. I will first make BtrPlace compatible with the spec language for fun for the moment
14:57:36 <acabot> cdupont : it was a suggestion
14:57:50 <acabot> fhermeni : ok ths
14:58:26 <cdupont> acabot: so I can work at spec level but implementation is not possible at the moment for resource reasons :(
14:58:57 <acabot> cdupont : thx
14:59:08 <acabot> time to end the meeting
14:59:12 <acabot> thanks everyone
14:59:18 <tpeoples> bye
14:59:20 <cdupont> thanks, bye!
14:59:24 <sballe> bye thx
14:59:26 <acabot> bye
14:59:35 <acabot> #endmeeting