16:01:12 #startmeeting watcher 16:01:13 Meeting started Wed Nov 11 16:01:12 2015 UTC and is due to finish in 60 minutes. The chair is acabot_. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:14 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:16 The meeting name has been set to 'watcher' 16:01:21 hi 16:01:26 Hi 16:01:27 hey 16:01:33 o/ 16:01:40 o/ 16:01:46 o/ 16:01:49 o/ 16:01:50 o/ 16:02:01 o/ 16:02:04 o/ 16:02:05 our agenda for today #link https://wiki.openstack.org/wiki/Watcher_Meeting_Agenda#11.2F11.2F2015_Agenda 16:02:21 hi i am here 16:02:37 Hi everybody 16:02:38 o/ 16:02:40 o/ 16:03:01 hi all - a few of us are here from ZHAW (me, bruno, vojtech) - we are not proficient with IRC meetings, so apols if we don’t know conventions etc 16:03:04 hi corentin 16:03:26 seanmurphy_: no problem ;-) 16:03:27 welcome 16:03:32 also a few of us from CREATE-NET are here 16:03:53 wlecome 16:03:57 Hello from Moscow! 16:04:00 welcoe :-) 16:04:06 Servionica is here:) 16:04:09 before we dig into the agenda, just to let you know that Monasca has been accepted yesterday in the big tent https://review.openstack.org/#/c/213183/ 16:04:11 Hi! 16:04:16 +1 16:04:26 great news for Monasca 16:04:33 congrats guys 16:04:49 cool! 16:04:56 congrats! 16:05:07 #topic Announcements 16:05:25 congrats! 16:05:35 so took the opportunity to invite our friends from ICCLab Zurich today 16:05:43 welcome folks 16:05:48 acabot_: +1 16:05:50 glad to have you attend! 16:05:52 maybe we can start by a quick intro guys 16:06:11 thanks for the invitation antoine 16:06:38 as you may see from agenda, we are a university applied research lab 16:06:48 focused on cloud computing with different aspects 16:06:52 one of which is energy 16:06:57 we have done a bit of work 16:07:10 it is modest, tbh - i don’t want to oversell it 16:07:11 ;-) 16:07:18 sure 16:07:34 we have developed an energy monitoring tool which collects energy consumption from servers 16:07:40 and shows it in a dashboard etc 16:07:47 it is based on the kwapi project 16:07:48 Greate 16:07:52 in the larger openstack ecosystem 16:08:16 it’s maybe a bit basic, but it’s open source and we can share a small video of it 16:08:31 we also have done a bit of work on delay tolerant workload mgmt as a tool to achieve energy savings 16:08:41 and have a horizon plugin to support deferred workload 16:08:54 seanmurphy_: you didn't put the link for the monitoring tool in the agenda, can you share it ? 16:08:56 so you can tell openstack to schedule launch of a VM at a specific time 16:09:07 i will do so 16:09:26 thx 16:09:29 seanmurphy_: interesting use cases indeed 16:09:46 Link for energy monitoring tool - http://blog.zhaw.ch/icclab/web-application-to-monitor-and-understand-energy/ 16:09:52 the idea with the deferred workload stuff is to add a scheduling component around it 16:10:06 which can take delay tolerant workload from users of the system 16:10:09 and scheduler - http://blog.zhaw.ch/icclab/extending-the-openstack-dashboard-to-support-delay-tolerant-workload/ 16:10:14 and schedule them using the deferred workload mechanism 16:10:21 then we have done a bit on load consolidation 16:10:40 reading usage info from a system a having a basic mechanism to support consolidation 16:10:53 it’s a bit heuristic rather than based on Hard Science 16:11:00 but it’s prob a reasonable compromise 16:11:08 yeah, most things are in this space :) 16:11:39 we also did work on live migration using so called post copy live migration rather than the standard pre copy mechanism which suffers from some issues if there is intense memory use in a VM 16:12:03 that suffers from the prob of lacking support in linux kernel and also of distributed state within the migration process 16:12:13 but is more certain to terminate than the pre copy approach 16:12:20 that’s our work in a nutshell 16:12:25 it’s not fully consolidated as yet 16:12:34 very cool, I'll read thru the material later today 16:12:47 but we plan to bring it together over the coming months 16:13:04 i think that’s abt it 16:13:10 from our side 16:13:20 acabot_: is that sufficient? 16:13:38 same here 16:13:41 thanks for the intro 16:13:41 seanmurphy_: seems great and interesting, I will look at the material 16:14:24 acabot_: shall we proceed with the agenda? 16:14:29 we will put more of our content in the minutes 16:14:46 as I see we have people from create-net, federico ? 16:14:53 we're here :) 16:15:04 we can introduce ourselve quickly 16:15:10 yes please 16:15:25 so we are from Create-Net (www.create-net.org), in Italy 16:15:32 Our area is called Smart Infrastructure (Smarti), working on cloud computing infrastructure 16:15:42 Our project is DC4Cities, which aim at using more renewable energy in data centres 16:15:55 At the moment we have several prototypes we're working on... 16:16:09 one is the EASC (Energy Efficient Software Controller): 16:16:43 it will control the workload level of flexible apps to follow some constraints, such as renewable energies 16:16:47 cdupont: could you share links on the agenda or on IRC using # link ? 16:17:00 OK 16:17:15 Also Plug4Green: a VM consolidator based on constraint programming with an energetic model 16:17:17 But it's in Java 16:17:35 Plug4Green is based on BtrPlace 16:17:37 EASC makes application adaptive to renewable energy availabilty 16:18:05 by introducing a few working modes (execution modes) for each application 16:18:18 So we have 2 ways to optimize: VM consolidation to save energy 16:18:33 EASC to optimize renewable energies 16:18:53 www.dc4cities.eu is for DC4Cities project 16:19:10 ok please share links (blogs & source code) as much as you can on the agenda and we will all look at it before our next meeting 16:19:31 there is also a tool called PaaS consolidator to consolidate containers but that might be out of topic for this meeting 16:19:47 cdupont: yes ;-) 16:20:03 #topic Review Action Items 16:20:13 #link www.dc4cities.eu 16:20:33 tnx 16:20:34 #link http://blog.zhaw.ch/icclab/extending-the-openstack-dashboard-to-support-delay-tolerant-workload/ 16:20:37 I was looking for it 16:20:49 does this sound interresting for you? How could we collaborate :) 16:20:50 #link http://blog.zhaw.ch/icclab/web-application-to-monitor-and-understand-energy/ 16:21:07 thx Ed (# link add the link to minutes) 16:21:09 EASC is instantiated for various computing styles 16:21:14 like PaaS and IaaS 16:21:19 acabot_: the watcher-specs templates are all merged now -- thanks for all the reviews from folks 16:21:30 EASC-PaaS has been implemented with Cloud Foundry 16:21:30 +1 16:22:00 cdupont: we will look at it in more detail and come back to you next week, as we have only an hour, we need to follow the agenda 16:22:11 (and big thanks to our nova colleagues allowing us to promote code reuse) :) 16:22:12 ok thanks :) 16:22:44 I am working on two things. 1. end to end install of watcher and all the pieces in a vM. I hope to have that completed today. and 2. workingon a sec with Nishi for the PoC. we are meetign with jed56 tomorrow to discuss further 16:22:47 #link http://www.fit4green.eu/ 16:23:41 sballe: ok thx for all your feedback regarding the installation process, feel free to send reviews on doc if there are issues 16:24:03 sballe: it would be better if the next one has a better doc ;-) 16:24:10 sballe: thanks for that... I will appreciate that for when I set it up soon 16:24:12 Zhenzan Zhou (is that you Kevin_Zheng ?) took care of my action item, creating the devstack plugin blueprint (#link https://blueprints.launchpad.net/watcher/+spec/devstack-plugin). i added a bit more detail and hope to start working on that soon 16:24:15 acabot_: will do. I still have to do the final step around configuration 16:24:31 No 16:24:35 no Zhenzan Zhou is with Intel and his nick is bzhou 16:24:42 ah okay 16:24:47 All... for each BP, we need to now submit the formal spec now that watcher-specs is ready 16:24:53 He is a guy from Intel 16:25:02 jwcroppe: +1 16:25:03 I'm from huawei 16:25:05 Kevin_Zheng: who are oyu with 16:25:09 :-) 16:25:18 if it's a 'small' change, then no spec is needed... otherwise we should get started on the formal spec write-up and get those approved 16:25:30 yes sorry we didn't introduced you Kevin 16:25:48 jwcroppe: we need the ceilometer integration for Devstack 16:26:21 BTW any ETA on Ceilometer integration. I believe jed56 said yes to ETA of Dec 1 16:26:23 That's ok I can do it in the open discussion 16:26:26 sballe: I agree we need to have the ceilometer integration and then we will be able to use Watcher in devstack env 16:26:41 +1000 my thoughts exactly 16:27:02 yes we plan to do it next week 16:27:12 you guys are the BEST! 16:27:14 Ok, so let's close on the specs stuff. For those LP blueprints we've opened, we should now create the specs in the watcher-specs repo now that we are all ready for that 16:27:32 we had a great discussion yesterday with jed56 and thx to Ceilometer discussion in Tokyo, it would be pretty easy to do 16:27:44 I'll work on Nishi's PoC spec with her. Should have a DRAFT early next week 16:27:49 can we review on the spec? 16:28:00 medhi__: yes of course 16:28:11 acabot_: create an action for watcher-specs submission? 16:28:24 medhi___: yes, please review all specs :) 16:28:29 sure 16:28:30 ok so I will approve BP regarding ceilometer, devstack and Nishi use case right ? 16:28:40 yes 16:28:50 and we'll work on the specs 16:28:57 sballe: wonderful 16:29:03 what's BP? 16:29:06 #action acabot_ approve BP regarding ceilometer, devstack and Nishi use case 16:29:16 BP = blueprint 16:29:19 ok 16:29:26 tpeoples: are you going to look at the devstack integration for watcher? do you want to post up the spec review for that too? 16:29:27 https://blueprints.launchpad.net/watcher 16:29:45 jwcroppe: as time allows...yes 16:29:59 tpeoples: +1 16:30:18 +1 16:30:18 fwiw, ceilometer is changing its name to 'telemetry' 16:30:39 edleafe: ok I will update the BP then 16:30:40 thx for letting us know 16:30:42 edleafe: wonderful. that's not confusing :) 16:30:54 jwcroppe: :) 16:30:59 lol 16:31:08 bug triage has been done by dtardivel 16:31:19 https://bugs.launchpad.net/watcher 16:31:27 so feel free to fix them ;-) 16:31:40 cool acabot_ 16:31:55 and of course add new one if you have troubles installing Watcher 16:32:13 Cool 16:32:28 #info change 'ceilometer' references to 'telemetry' 16:32:38 sballe: you asked for changing our meeting time 16:32:57 sballe: what would be best for US/China & Europe :-D 16:33:00 yes the main reason is that it is 12am for bzhou 16:33:25 Me too 16:33:29 I could do as early as 6 or 7 amEST 16:33:37 UTC ? 16:33:39 which is 8pm in China 16:33:50 let me find a converter 16:33:58 that'd be 5-6 am CST? 16:34:21 sballe: my biological clock does not compute anything between 5-6 am 16:34:40 lol 16:34:41 jwcroppe: :-D 16:34:59 Maybe just move to 1 or 2 hour earlier, think it will be ok for us 16:35:11 7am EST is 1pm Paris time and 8pm in China right ? 16:35:14 I understand but how if we moved to 8 am EST == 7 am CST == 9pm in China 16:35:26 I can handle up to 3 hours earlier 16:35:54 jwcroppe: what about you ? 16:36:23 perhaps we try a model like others and alternate times every other week? just a thought 16:36:24 that would be 1pm UTC per my converter 16:36:25 3h earlier is OK for France, Italy and Switzerland I suppose 16:37:19 How about 2pm UTC == 9am EST == 10pm in China = 7am CSt 16:37:41 I'm OK 16:37:43 yes - 3h earlier is fine for CH 16:37:48 In China 16:37:50 9am EST == 8am CST 16:38:06 oh yeah I went MST 16:38:07 so maybe 8am EST? 16:38:13 I'm fine with 8 am CT 16:38:26 so 2h earlier than today 16:38:29 7 am will be almost impossible for me, but I don't want to be the bottleneck. 16:38:34 2pm UTC == 9am EST == 10pm in China = 8am CSt 16:38:39 right, 2h earlier 16:38:43 +1 16:38:47 +2 16:38:55 +1 16:39:01 +1 16:39:02 +1 16:39:07 +1 16:39:23 +1 16:39:30 tpeoples: Can you check and see if you can find a day/channel for that time? You did it last time rigth? 16:39:30 #openstack-meeting is available at 2pm UTC on wednesday 16:39:41 +1 16:39:42 +1 16:39:47 acabot_: you were reading my mind :) 16:40:00 acabot_: then let's grab it 16:40:01 10 pm for EU? 16:40:15 #action acabot_ ask for a new time slot for meeting at 2pm UTC on wednesday on #openstack-meeting 16:40:18 No 10 for China 16:40:30 we can also try alternating times too eventually if needed 16:40:33 it would be 3pm for EU 16:40:45 (since the reality is there is no perfect world time) 16:40:45 acabot_: it will be early if we want people from the west coast to join. But it all about balance between all the time zones 16:40:56 jwcroppe: I agre 16:41:08 sballe: acabot_ : #openstack-meeting-4 is open during that time every week. #openstack-meeting is taken every other week during that slot by the doc team 16:41:09 I agree on the alternate times down the road 16:41:47 tpeoples: #openstack-meeting-4 it is 16:41:54 tpeoples: ok my cal is driving me crazy... 16:42:00 acabot_: :) 16:42:07 acabot_: you should use Notes :) 16:42:22 jwcroppe: lol 16:42:29 #topic Blueprint/Bug Review and Discussion 16:42:35 jwcroppe: don't be mean! 16:42:45 lol 16:42:59 I'd like to discuss BP regarding the Nova load balancer 16:43:17 acabot_: +1 16:43:22 Okay 16:43:32 to my understanding, it implies to change the nova configuration and this is something we definitely dont want to do 16:43:33 We've made videos 16:43:58 acabot_: are you referring to https://blueprints.launchpad.net/watcher/+spec/watcher-overload-underload specifically? 16:44:41 tpeoples: yes 16:45:45 i think that seems to be two specific algorithms (underload / overload) that would be written for watcher specific to the "reactive" part of watcher, which relies in getting the CEP working for that ? 16:45:48 acabot_: In Tokyo we discussed, what Watcher already have some improvements to Nova functionality what we ned 16:46:03 btw, who are Antoine Cabot and Jean-�mile DARTOIS? 16:46:07 I'm not seeing anything in there that would require a change to nova? 16:46:34 mehdi__: sorry I am Antoine, and Jean-Emile is jed56 16:46:52 tnx 16:47:18 what needs to change in nova for that blueprint samarinvv acabot_ ? 16:47:21 acabot_: time check ... 13 mins remaining 16:47:32 there is some changes in nova libvirt driver and nova db (models, api) 16:48:11 jwcroppe: yes sorry for that I looked at the videos at the end of the agenda just before the meeting and it seems to me that Nova has been modified 16:48:13 Hmm I think we have tested something similar 16:48:19 and extend nova-api 16:48:42 alexchadin: It's not clear to me why you'd need model/api changes for what's written in the description? we can take that offline tho 16:49:24 +1 jwcroppe , seems to me that could be implemented without nova changes. let's maybe get a written summary of the needed changes for next meeting and go from there? 16:50:08 alexchadin: what do you think ? 16:50:17 acabor_: Ok, we should discuss this offline 16:50:20 we are on the way to leave nova code 16:50:25 *acabot 16:50:35 jwcroppe: it could be good 16:51:00 alexchadin: samarinvv: ok that we can start iterate through a more detail spec right ? 16:51:21 +1 16:51:40 ok, right 16:51:42 as we now have a watcher-specs repo, we can start working on it and review specs 16:52:01 +1 16:52:09 it would be easier to discuss it later on 16:52:19 okay, we should prepare some for next meeting to discuss 16:52:41 acabot_: +1 16:52:43 that's what i meant by written summary :$ 16:52:46 acabot_, sballe: any further detail on mid-cycle to discuss this week? 16:52:46 but anyway thx for the videos, its really helpful to understand your current status 16:53:01 I have not heard back from Orran yet 16:53:22 give him 2 more weeks and then after we can look at planB 16:53:41 ok just to let all know about it, we plan a mid-cycle meetup end of January in Boston, MA 16:53:46 link to the video? 16:53:50 ok, was Orran going to host at Boston Univ potentially? 16:54:09 medhi__: in the agenda 16:54:16 jwcroppe: see above 16:54:32 acabot_: thanks for watching 16:55:22 samarinvv: which of your 2 BP you want to start providing detailed specs ? 16:55:55 sballe: saw that, just wasn't sure if that was your pending question to him... 16:56:03 acabot_: overload 16:56:11 ok thx 16:56:13 acabot_: overload 16:56:15 ok 16:56:28 #action acabot_ set overload BP as started 16:56:36 I sent him email several time but I know heis busy 16:56:42 np 16:57:26 ok so before the end, ICCLab & Create-net please complete the agenda with links to your current work 16:57:53 we will look at it before next meeting and see how we can iterate from it 16:58:28 OK, thanks acabot_! 16:58:52 thanks for this great meeting, I will go back to my day off ;-) 16:58:53 great 16:59:06 lol 16:59:12 have a nice evening all 16:59:22 thanks bye 16:59:26 bye 16:59:29 #endmeeting