16:02:08 <acabot_> #startmeeting watcher
16:02:08 <openstack> Meeting started Wed Nov  4 16:02:08 2015 UTC and is due to finish in 60 minutes.  The chair is acabot_. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:02:09 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:02:12 <openstack> The meeting name has been set to 'watcher'
16:02:16 <sballe> o/
16:02:19 <edleafe> o/
16:02:20 <tpeoples> o/
16:02:30 <dtardivel> o/
16:02:31 <acabot_> hi everyone
16:02:53 <acabot_> today agenda is #link https://wiki.openstack.org/wiki/Watcher_Meeting_Agenda#11.2F4.2F2015_Agenda:
16:04:06 <acabot_> #topic annoucements
16:04:45 <sballe> We had some great meetigns at the summit :-) Thanks to everyone involved
16:05:22 <sballe> Also I have an update around the Meetup
16:05:53 <acabot_> sballe: you mean the mid-cycle in january ?
16:06:07 <sballe> Orran/BU is looking at hosting it because I am not sure Intel can.
16:06:19 <sballe> Also Boston is a better place than Hudson, MA
16:06:29 <acabot_> ok
16:06:51 <acabot_> maybe we can already fix the dates today ?
16:06:56 <sballe> I will follow up with him. We talked about htis yesterday. They already hosted a midt-cycle for keysotn ehtis summer
16:07:23 <sballe> we were talking about 26 and 27 and maybe 28. Orran needs to see when he has room available
16:07:27 <sballe> s/rooms
16:07:52 <acabot_> ok
16:08:27 <acabot_> coming back to the agenda, we add an email from ttx about using "OpenStack Watcher" as the project name
16:08:52 <sballe> acabot_: I didn't see the email
16:08:57 <acabot_> its confusing because of the big tent inclusion so please do not use it anymore, we should stick to "Watcher" for now
16:09:13 <acabot_> #link http://lists.openstack.org/pipermail/openstack-dev/2015-November/078136.html
16:09:30 <sballe> oh ok so it was the "OpenStacK" piece that was the issue
16:09:47 <acabot_> yes sorry if I was unclear
16:09:55 <acabot_> Watcher is OK for the project name
16:10:00 <sballe> ok cool!
16:10:16 <acabot_> any other annoucement ?
16:10:49 <sballe> nope
16:11:09 <acabot_> #topic Review Action Items
16:11:36 <acabot_> the watcher-specs repo is now live https://github.com/openstack/watcher-specs
16:11:44 <sballe> yeah!!!
16:11:57 <acabot_> jwcroppe submit a first patch to add all templates
16:12:09 <acabot_> #link https://review.openstack.org/#/c/240475/
16:12:27 <acabot_> if everyone can have a look at it
16:13:03 <acabot_> we also have 4 core contrib on watcher-specs sable, jwcroppe, jed56 and acerbate
16:13:19 <dtardivel> acabot_: watcher and python-watcherclient internal bugs have been moved to launchpad. We have to prioritize them now
16:13:59 <acabot_> #link https://bugs.launchpad.net/watcher
16:14:16 <acabot_> does anyone wants to start a bug triage ?
16:15:14 <dtardivel> acabot_: I can do it
16:15:19 <sballe> acabot_: I think one of the b-com folks should do that given that you have it up and running.
16:15:27 <acabot_> ok
16:15:38 <sballe> As soon as dtardivel and I have my env up and running I can help with bugs
16:15:40 <tpeoples> yeah i think that makes the most sense for the time being
16:15:46 <acabot_> #action dtardivel start the bug triage
16:16:29 <acabot_> ok what about devstack plugin for watcher ?
16:16:59 <tpeoples> still a TODO for me, no updates there. hoping to work on that the end of this week and beginning of next
16:17:10 <sballe> that is important IMHO but I just realized this morning that there are many pieces to make the end-to-end wathcer thing work
16:17:23 <acabot_> tpeoples: ok thanks
16:17:27 <sballe> so I am not sure what the desvtack plug-in means
16:17:50 <sballe> does it mean watcher+wathcher metring chain + agent?
16:18:42 <acabot_> actually we didn't look at it in detail as there are many changes in devstack for Mitaka
16:18:43 <sballe> I think that reading ht metrics from ceilomters instead of having to use the metring chain+ agent might be a higher priority and would allow a clean integration wiht devstack
16:19:00 <dtardivel> +1
16:19:47 <tpeoples> ok
16:19:56 <edleafe> it would also make it easier for deployments using Ceilometer to use Watcher
16:20:05 <sballe> I am fine with using ceilometer to start  with for devstack and not ceilosca or monasca
16:20:09 <acabot_> ok so lets write a blueprint on this and give it high priority
16:20:16 <edleafe> adding agents is usually discouraged
16:20:17 <sballe> acabot_: +1
16:20:31 <acabot_> edleafe: I agree
16:20:42 <Dmitry> Hello from Moscow Nova LB team, this is Dmitry Demidov. Let me introduce my collegues Alexander Chadin (alexchad) and Alexander Stavitsky (alexstav).
16:20:44 <sballe> edleafe: +1
16:21:04 <alexchadin> hello everyone
16:21:07 <alexstav> Hi all
16:21:11 <edleafe> hello!
16:21:14 <sballe> hi.
16:21:15 <acabot_> hi guys
16:21:24 <dtardivel> hi
16:21:24 <vmahe_> hi
16:21:35 <sballe> What is the Nova LB? I thougth he LB effort were in Neutron
16:21:50 <sballe> unless lB doesn't stand for load-balancer
16:21:55 <tpeoples> it was the topic that came up in one of the nova unconference sessions
16:21:56 <Dmitry> yes
16:21:58 <acabot_> Dmitry: can you share the blueprint link ?
16:22:27 <alexchadin> one moment please
16:22:44 <sballe> ok let's do that under open topics.
16:22:53 <alexchadin> https://blueprints.launchpad.net/nova-loadbalancer/+spec/nova-lb-overview
16:22:57 <alexchadin> #link
16:23:02 <acabot_> sballe: we discussed it on friday, there are interesting things in the Nova LB and we need to see how we can integrate it in Watcher
16:23:10 <alexchadin> yeah
16:23:28 <alexstav> https://docs.google.com/document/d/1G2jYSeDzLD2JjAqZneKGQi2opVI-FIUhu6DeFhGE99Y/edit?usp=sharing
16:23:32 <alexchadin> we are about integrating our project in Watcher
16:23:34 <acabot_> sballe: the nova team agreed it should be done in an external project and suggest to join Watcher effort
16:23:35 <alexstav> Here it is our specs
16:24:15 <sballe> ok. I am just reluctant to have another LB effort beside the neutron one
16:24:35 <sballe> We need to understand if and how it overlap wth neutron LB.
16:24:37 <tpeoples> LB isn't the right term, it is the reactive part of watcher for optimization
16:24:49 <acabot_> I dont think it's LB like we use to have in Neutron ;-)
16:25:14 <sballe> ok then it should be renamed IMHO
16:25:34 <sballe> because other folks are going to not like the idea either.
16:25:35 <edleafe> or just integrated into Watcher :)
16:25:38 <acabot_> we should all take a look at the blueprint and specs and see how Dmitry's team can improve Watcher
16:25:41 <alexstav> Yep, it's not assosiated with neutron :)
16:25:44 <Dmitry> it looks like decision engine inside watcher
16:26:27 <sballe> acabot_: Can we get back to the devstack and ceilomter blueprint. Who has the action to write it up?
16:26:45 <sballe> and it sounds like tpeoples will work on it
16:27:04 <sballe> tpeoples: Will you have the time to work on it this year? I think this is super hight priority
16:27:39 <acabot_> sballe: we plan to do it at b-com so I will write down the blueprint
16:27:50 <sballe> acabot_: great!
16:28:14 <acabot_> #action acabot write the blueprint for collecting metrics with ceilometer
16:28:20 <tpeoples> i will have some time to work on it but if it is high priority then i would suggest someone from bcom since i can't be 100% yet ... acabot_  ok
16:28:22 <sballe> acabot_: so both ceilomter and watcher integration? and devstack and watcher plug-in?
16:28:39 <sballe> so two BPs?
16:28:59 <acabot_> we should definitely add a blueprint for devstack if we plan to work on it
16:29:03 <tpeoples> yes, i would say two bps
16:29:18 <acabot_> tpeoples: can you submit a blueprint for devstack ?
16:29:19 <tpeoples> i can work on the BP for devstack integration
16:29:21 <tpeoples> yes
16:29:30 <sballe> tpeoples: acabot_ +1
16:29:35 <acabot_> ok great, I will handle the other one on ceilometer
16:30:07 <tpeoples> #action tpeoples write the blueprint for devstack watcher plugin
16:30:08 <sballe> acabot_: +1
16:30:39 <acabot_> Dmitry: could you please submit a blueprint regarding how you see the Nova LB integration ? We will iterate on this later
16:30:58 <sballe> and please rename LB :)
16:31:06 <alexchadin> we have already submitted two new blueprints
16:31:17 <alexchadin> about 10 minutes ago:)
16:31:58 <acabot_> ok great, #link https://blueprints.launchpad.net/watcher/+spec/watcher-overload-underload #link https://blueprints.launchpad.net/watcher/+spec/add-new-watcherclient-commands
16:32:10 <alexchadin> so
16:32:19 <alexchadin> we have one big spec
16:32:29 <alexchadin> need link?
16:32:42 <acabot_> the one you shared previously ?
16:32:43 <alexchadin> it was introduced at Tokyo
16:32:46 <alexchadin> yes
16:33:09 <acabot_> #link https://docs.google.com/document/d/1G2jYSeDzLD2JjAqZneKGQi2opVI-FIUhu6DeFhGE99Y/edit to add it to the minutes
16:33:53 <acabot_> #topic Blueprint/Bug Review and Discussion
16:34:14 <sballe> alexchadin: Do you have a working prototype with nova or is this still a concept?
16:34:19 <alexchadin> yes
16:34:33 <alexchadin> overload underload is working fine
16:34:40 <sballe> ok cool! thx
16:34:47 <alexchadin> we have repo on github
16:35:05 <acabot_> alexchadin: please share the link
16:35:10 <alexchadin> how can we show our use cases at work?
16:35:15 <alexchadin> one moment
16:35:29 <alexchadin> record youtube video?:)
16:35:47 <acabot_> alexchadin: yes why not ?
16:35:47 <alexstav> https://github.com/Stavitsky/nova/tree/loadbalancer-client
16:35:52 <alexchadin> okay
16:35:58 <alexchadin> we will record it
16:36:09 <alexchadin> to the next meeting
16:36:17 <sballe> alexchadin: Where do you get the metrics you are using?
16:36:41 <alexchadin> we collect it from compute nodes
16:36:45 <alexchadin> to nova db
16:36:47 <alexstav> with nova-conductor
16:36:48 <sballe> via an agent?
16:36:52 <sballe> oh I see
16:37:44 <alexchadin> we have expanded set of data we get from nodes
16:38:06 <alexstav> To collect CPU and RAM
16:38:35 <alexchadin> it already balances vms among the nodes
16:38:40 <sballe> we'll need to find a way to get all the metrics we need in a standard way
16:38:56 <acabot_> sballe: +1
16:39:41 <alexstav> What do you mean by 'standard way'?
16:40:36 <sballe> common way
16:40:55 <sballe> so watcher has one way to get the data...
16:40:59 <edleafe> alexchadin: does it migrate existing nodes to balance? Or simply create them in a more balanced arrangement?
16:41:20 <alexstav> first
16:41:33 <sballe> and thta way should be extensible and easy to lugin new metrics
16:41:36 <alexstav> we use nova-migrate
16:41:58 <sballe> s/plugin
16:42:34 <alexchadin> it migrates existing ones
16:42:46 <edleafe> alexchadin: ok, thanks
16:43:37 <acabot_> ok next topic is blueprints review
16:43:39 <alexchadin> it's a good idea to use plugin way
16:43:43 <alexchadin> sorry
16:44:08 <acabot_> as we discussed in Tokyo, the first use case we want to handle is https://blueprints.launchpad.net/watcher/+spec/nishi-ahuja-energy-efficienct-dc
16:44:48 <acabot_> as soon as watcher-specs is ready, we need to define in more details what we need to do to complete this blueprint
16:44:49 <sballe> acabot_: Yeah I have asked Nishi to split it up into 3 BPs but I think she ran out of time
16:45:04 <sballe> acabot_: +1
16:45:10 <acabot_> sballe: do you think Nishi can handle specs on this ?
16:45:38 <sballe> I can help her. The first one we will be tackling is: •	Work load migration based on power and thermal data.
16:46:20 <sballe> I will work with her over the next week to get this done
16:46:28 <acabot_> #action sballe add specs for blueprint on energy efficiency
16:47:05 <acabot_> regarding jwcroppe patch, we can probably merge it tomorrow
16:47:21 <sballe> Ok. I will review it later today
16:47:22 <acabot_> and start from templates
16:47:28 <acabot_> ok great
16:47:48 <acabot_> we already talked about bugs, anything to add here ?
16:48:29 <tpeoples> acabot_:  did we get core access to the specs repo?
16:48:55 <acabot_> tpeoples: yes, jwcroppe, sballe, acabot & jed56
16:48:59 <tpeoples> ok cool
16:49:13 <acabot_> #topic Open Discussion
16:49:36 <acabot_> does anyone had a look at ceilosca ?
16:50:02 <sballe> I had extensive discussions with the team taht created ceilosca at the summit
16:50:44 <sballe> they suggested that unless we need the ceilometer APIs to just go to Monasca but that Ceilosca will scale better then Ceilomter
16:52:04 <sballe> But for devstack I would suggest we stick with Ceilomter. for production we'll need to support either Monasca or Ceilosca or both
16:52:06 <acabot_> does ceilosca provides any other advantage than scalability ?
16:52:24 <acabot_> compare to ceilometer I mean
16:52:34 <sballe> if you already have ceilomter it gives you API compatibiltiy
16:53:04 <acabot_> ok
16:53:28 <sballe> I would suggest we skip it for now
16:53:37 <acabot_> lets stick to ceilometer now with a blueprint
16:53:42 <acabot_> sballe: +1
16:53:42 <sballe> We can revisit this decision for the next release
16:54:07 <acabot_> we had a discussion with vmahe about moving our diagrams on the public wiki
16:54:20 <sballe> acabot_: Good idea!
16:54:30 <jwcroppe> here - sorry - calendar was wrong :(
16:54:34 <acabot_> but its pretty hard to maintain architectural diagrams on the wiki
16:54:40 <sballe> jwcroppe: np. Welcome
16:54:45 <vmahe_> yes, we should describe those diagrams as ascii text
16:54:50 <acabot_> any other solution on that ?
16:54:54 <tpeoples> http://asciiflow.com/
16:55:10 <jwcroppe> ascii diagrams are good, but also a nightmare to depict some things
16:55:13 <sballe> that is just horrible :-(
16:55:21 <sballe> jwcroppe: +1--
16:55:29 <vmahe_> yes but it's easier to read than XMI :-)
16:55:36 <acabot_> before we start moving all this drawing stuff, I would like to use the right tool...
16:55:51 <sballe> +1
16:55:55 <vmahe_> the following tool looks great http://asciidoctor.org/docs/asciidoctor-diagram/
16:56:18 <sballe> What are the other teams doing?
16:56:21 <jwcroppe> this is a much debated subject :)  I personally don't mind using Google slides and draw in there
16:56:29 <vmahe_> it integrates with PlantUML for describing class diagrams, sequence diagrams, ...
16:56:56 <vmahe_> It's harder to version your google slides
16:56:57 <tpeoples> once we nail down the architecture, it shouldn't be changing that much...
16:57:00 <acabot_> http://asciiflow.com/ is recommended in the template for specs
16:57:31 <acabot_> can we give a try to asciidoctor ?
16:57:42 <acabot_> for database and architecture schema ?
16:57:49 <vmahe_> the good point with ascii description is that it can be handled in the git repo and with git-review as well
16:58:08 <alexchadin> so, about integrating nova lb in Watcher, we will record video with our use cases to show how it works.
16:58:15 <vmahe_> your diagrams can be associated to a given version of the code as well
16:58:21 <sballe> I am fine with giving asciidoctor a try as long as we can change if it doesn't work
16:58:22 <acabot_> alexchadin: ok thx
16:58:37 <jwcroppe> sballe: +1
16:59:10 <acabot_> vmahe_: lets try to build 2 diagrams for next week ?
16:59:19 <vmahe_> all right
16:59:24 <sballe> acabot_: vmahe_ +1
16:59:28 <acabot_> ok great
16:59:37 <acabot_> so its time to end the meeting
16:59:42 <acabot_> thanks
16:59:42 <jwcroppe> did anyone have any major concern with the specs docs I pushed up?  Kudos to the nova team for providing the baseline :)
16:59:49 <sballe> OK. bye see you in #openstack-watcher
16:59:59 <acabot_> jwcroppe: we will merge it tomorrow
17:00:06 <jwcroppe> thx
17:00:09 <edleafe> thanks!
17:00:09 <tpeoples> ttyl
17:00:16 <acabot_> #endmeeting