Tuesday, 2016-05-31

*** flwang1 has joined #openstack-higgins00:09
*** Qiming has joined #openstack-higgins00:10
*** Wang_jian has quit IRC00:18
*** Qiming has quit IRC01:17
*** yanyanhu has joined #openstack-higgins01:36
*** hongbin has joined #openstack-higgins01:37
*** shu-mutou-AFK is now known as shu-mutou02:02
*** Qiming has joined #openstack-higgins02:13
hongbinHi all, our team meeting will start in 20 minutes at #openstack-meeting02:40
hongbinAlso, I plan to send a new project announcement to the ML. Here is the draft: https://etherpad.openstack.org/p/higgins-announcing-statement Comments are appreciated02:42
flwang1hongbin: sure, thanks for the reminder02:44
*** yuanying has quit IRC02:47
*** mkrai has joined #openstack-higgins02:47
*** sheel has joined #openstack-higgins02:47
*** yuanying has joined #openstack-higgins02:52
*** adisky has joined #openstack-higgins02:59
openstackgerritOpenStack Proposal Bot proposed openstack/higgins: Updated from global requirements  https://review.openstack.org/32306503:00
hongbineliqiao: Qiming yanyanhu harlowja ping03:01
hongbinThe team meeting is now on the #openstack-meeting channel03:01
yanyanhuhi, hongbin , has been there :)03:02
hongbinflwang1: hey, the meeting is on. Joint #openstack-meeting if you have a chance03:04
*** yuanying has quit IRC03:43
*** yuanying has joined #openstack-higgins03:47
mkraiyanyanhu, Agree04:00
*** hongbin has quit IRC04:01
mkraiadisky, Hi04:01
eliqiaoI just wonder how can we start the port of nova virt-driver part of work.04:01
openstackgerritEli Qiao proposed openstack/higgins: DO NOT MERGE  https://review.openstack.org/32310604:01
yanyanhumkrai, that's an important basement :)04:02
yanyanhuwill leave, ttyl04:02
mkraiBye yanyanhu04:02
mkraieliqiao, You mean the unified nova APIs04:02
*** xuhaiwei_ has joined #openstack-higgins04:10
adiskyhi mkrai04:37
*** chandankumar has joined #openstack-higgins04:43
*** yanyanhu has quit IRC04:53
openstackgerritaditi sharma proposed openstack/higgins: Add db migrations  https://review.openstack.org/32208505:33
adiskyhi, i have create initial migrations and now all higgins-db-manage command are working.05:35
openstackgerritMerged openstack/higgins: Updated from global requirements  https://review.openstack.org/32306505:43
eliqiaomkrai: talked with ironic guys, ironic virt-driver is just a api-proxy (will proxy nova requests to ironic-api). the benefit of adding a virt-driver of nova is to leverage nova's scheduler06:07
eliqiaoI have no problem to adding a virt-driver of higgins, but my concern is if nova team allow us to do such implementation.06:08
mkraiIt is good not to reinvent the wheel06:09
mkraiBut I have doubt whether nova scheduler will suit containers requirement?06:09
mkraiIf yes, then it is good to use nova scheduler which is robust06:09
mkraieliqiao, I think they will not have any issue06:10
eliqiaomkrai: there is a spec which proposed by a nova core. they are trying to refine nova scheduler to support different resource(suchu as ironic node)06:24
mkraieliqiao, so there should be same for containers06:25
eliqiaoI don't think it's good idear to depend on nova if we are only want a scheduler06:25
mkraiCould you share the link?06:25
mkraieliqiao, Do you think it will be good to write our own scheduler?06:26
eliqiaomkrai: check this https://review.openstack.org/#/c/312696/06:26
eliqiaoand https://review.openstack.org/#/c/183837/06:26
eliqiaomkrai: not sure.06:27
eliqiaoI meant, it will be very hard to depend on nova.06:27
mkraieliqiao, Thanks06:28
mkraieliqiao, I agree that it is always good to be an independent project06:29
mkraieliqiao, But that depends on what is our requirement06:29
mkraiIf we can implement our own scheduler that would be great but at the end we would need nova to provide us hosts06:29
eliqiaomkrai: yep.06:31
mkraiWe should discuss more on scheduler part in meeting and see if we can remove this dependency06:32
eliqiaomkrai: good idea, glad to talk with you.06:34
mkraiSame here eliqiao06:34
openstackgerritMerged openstack/higgins: Add db migrations  https://review.openstack.org/32208506:38
openstackgerritEli Qiao proposed openstack/higgins: WIP fix in devstack lib  https://review.openstack.org/32319708:17
*** Qiming has quit IRC08:50
*** Qiming has joined #openstack-higgins09:01
*** shu-mutou is now known as shu-mutou-AFK09:49
*** chandankumar has quit IRC10:09
*** Qiming has quit IRC10:46
*** chandankumar has joined #openstack-higgins11:16
*** Qiming has joined #openstack-higgins11:59
*** Wang_jian has joined #openstack-higgins12:09
*** Wang_jian has quit IRC12:09
*** Wang_jian has joined #openstack-higgins12:09
*** mkrai has quit IRC12:11
*** Wang_jian has quit IRC12:26
*** Wang_jian has joined #openstack-higgins12:29
*** chandankumar has quit IRC12:42
*** Wang_jian has quit IRC13:13
*** Wang_jian has joined #openstack-higgins13:27
*** Wang_jian has quit IRC13:29
*** Wang_jian has joined #openstack-higgins13:38
openstackgerritMerged openstack/higgins: Add steps to setup devstack environment  https://review.openstack.org/31967513:43
*** chandankumar has joined #openstack-higgins13:46
*** Wang_jian has quit IRC13:47
*** Wang_jian has joined #openstack-higgins14:13
*** Wang_jian has quit IRC14:16
*** Wang_jian has joined #openstack-higgins14:22
*** hongbin has joined #openstack-higgins15:13
*** chandankumar has quit IRC15:24
hongbindims: Hey Dims, I am going to send out a new project announcement. Here is the draft: https://etherpad.openstack.org/p/higgins-announcing-statement15:32
*** openstackgerrit has quit IRC15:33
hongbinFeel free to edit it once you have a chance15:33
*** openstackgerrit has joined #openstack-higgins15:33
*** Wang_jian has quit IRC15:38
dimshongbin : will try to do it later this afternoon.15:44
dimshongbin : we should run it by the team before we send15:44
*** Wang_jian has joined #openstack-higgins15:45
hongbindims: I sent to the team yesterday15:45
hongbindims: Some people has editted it15:45
hongbinYou are the last one I think :)15:46
hongbinMaybe I can send a ML in case some folks didn't notice the IRC activity15:47
dimshongbin : i meant we can use the meeting tonight15:47
hongbindims: It is yesterday night :)15:48
dimsoops. i am a day off :)15:49
hongbinI thought it was mentioned in the Open discussion15:49
dimsi did not make it to last night's chat15:49
dimsif a lot of folks have already looked at it then we should just go ahead15:49
*** Qiming has quit IRC15:49
dimsonly thing that stands out is, post the date/time for meeting and possibly the links to etherpads15:50
*** Wang_jian has quit IRC15:50
hongbingood points15:50
*** chandankumar has joined #openstack-higgins17:09
*** Wenzhi has quit IRC18:30
*** Wenzhi has joined #openstack-higgins18:30
*** Wang_jian has joined #openstack-higgins18:49
*** Wang_jian has quit IRC18:50
*** chandankumar has quit IRC18:53
*** sheel has quit IRC19:25
*** Wenzhi has quit IRC19:47
*** Wenzhi has joined #openstack-higgins19:47
flwang1hongbin: ping22:28
flwang1i'm reading the api design spec22:28
flwang1and i'm wondering if we should split it into two bp but keep this one as a parent bp if we want22:29
harlowjaother thing i guess i'd like in https://etherpad.openstack.org/p/higgins-announcing-statement is a high level definition of the apis that will be higgins22:31
harlowja ie /create_container22:31
hongbinflwang1: which one is the parent bp?22:35
flwang1hongbin: i mean if we agree to place higgins as  a similar position like Ironic22:36
hongbinharlowja: that is undecided yet...22:36
flwang1then that means  some api will be native api, called by end users22:36
hongbinharlowja: but let me know what I can said in the annoucement22:36
flwang1some api(endpoint) are called by nova22:37
hongbinflwang1: ack22:37
flwang1just my 2 cents22:38
hongbinflwang1: Then there are two BPs, 1) the nova API, 2) the Higgins API22:38
hongbin#2 should go first, then have #1 depends on #222:38
hongbinIs that correct?22:39
flwang1or, at least add a note for the current bp, say it's for native Higgins API consumed by end users22:40
hongbinflwang1: sure. I can do that22:40
hongbinflwang1: In addition, I think the nova API spec should be on the nova repo if there is one22:42
hongbinThen, we can definetly link these two BPs together22:42
flwang1hongbin: hmm.. i don't think there is a Nova API22:44
flwang1let me explain a bit, we basically need to define some API so that nova can call Higgins like this https://blueprints.launchpad.net/nova/+spec/add-ironic-driver22:45
flwang1so in the future, there could be a Higgins virt driver in Nova22:46
hongbinThen, there is a nova blueprint for implementing a Higgins virt driver22:46
flwang1and IMHO,  it would be great if nova can support multi virt drivers(can it now? )22:47
hongbinIn addition, there is a blueprint in Higgins to expose a (REST? RPC) API for higgins virt-dirver?22:47
flwang1so that VM, baremetel and containers are all compute resources of nova22:47
flwang1and nova can schedule the instance request based on image type or22:47
flwang1a pre-defined tag on image22:48
hongbinNo sure about that22:48
flwang1hongbin: yep, does that make any sense :)22:48
flwang1we can discuss, just dump my brain22:48
hongbinyou mean multiple virt-drivers?22:49
flwang1harlowja: any comments? ^22:49
flwang1hongbin: yep22:49
hongbinYes, definitely make sense. Just not sure if it is supported22:49
hongbinIn about two years ago, I tried to configure Nova to schedule VM and baremetal. but I failed22:50
flwang1so in that Utopia, user don't have to care about the resource types, they just use it based on their requirement22:50
hongbinYes, that is optimal22:51
hongbinflwang1: The Higgins virt-driver should call Higgins REST API or RPC API ? (in your opinions)22:52
flwang1hongbin: ah, i remember now22:53
flwang1hongbin: nova can do that22:53
flwang1nova can schedule request based on the architecture-type22:53
hongbinI see22:53
harlowjai'd still try to find the best/most useful thing that higgins can do, and make it the best at that22:54
flwang1that said, if we can define the container image in glance with a predefine architecture-type, let's say 'container', then nova can schedule it22:54
harlowjathat may require data from users, as to what they actually want22:54
harlowja(not all users are showing up to the dev mailing list and such either, so its hard)22:55
hongbinflwang1: sounds reasonable22:55
hongbinharlowja: ack22:56
harlowjadoes the foundation have any way to get survey questions to said users?22:56
harlowjai could almost see a container survey being useful22:56
harlowja(in figuring out the higgins goals and such)22:56
hongbinThere is a user survey, but not container specific22:57
*** Qiming has joined #openstack-higgins22:57
harlowjai'd be more interested in using that same mechanism to get information about containers22:57
flwang1harlowja: do you still remember the discussion in the mail list about unified the compute resource(VM, baremetel, and containers)? i think higgins can fit it22:57
harlowjain general every project should probably use that same mechanism to learn from users22:57
flwang1harlowja: i like that idea22:58
harlowjaflwang1 sure, if u consider containers as the lowest level resource, then they can fit it22:58
harlowjadepends on how u nest all these things (like a russian doll, ha)22:58
harlowjai'd be concerned honestly if the foundation doesn't have a way for projects to use that survey mechanism to reach a larger audience and get feedback (because ummm, without it, u just sort of hope that features u are doing are right?)22:59
harlowjabut meh, i've never been impressed with what the foundation makes possible (the infra team is super, but besides that, meh)22:59
flwang1harlowja: yep, currently, the survey is too general23:00
flwang1we just ask user if they're using xxx, but we don't know what and why user want xxx23:00
harlowjahard to gain insights from that info23:00
hongbinIf there is no data, we have to guess what users want23:00
hongbinThen, let's guess :)23:01
harlowjaand sample size needs to be bigger than say a few :-P23:01
harlowjawell did we try to get the data and failed at getting it already??23:01
hongbinPersonlly, I don't know how to get it23:01
harlowjaif we did that, then sure guess, spin a dice, play the lotto...23:01
harlowjaright, maybe that can be a question to the ML?23:01
harlowjawith [foundation] tags23:02
harlowjaor something23:02
hongbinYes, I can try that23:02
* harlowja i'd like to know that answer also, ha23:02
harlowjait seems like something that the foundation should be helping provide23:02
harlowjacause otherwise projects be doing <whatever they want here>23:02
flwang1hongbin: another possible options is sending mail to ops ML23:03
harlowjais that the right audience?23:03
flwang1they're the users not far aways form us23:03
flwang1harlowja: at least, they may know if they want to provide such kind of service :)23:03
harlowjasure, agreed, useful data, unsure if u should derive much value from it though23:04
harlowjawe'd still be on BBS and such if someone didn't push past ops ;)23:04
flwang1technically, most of the target users of higgins are app developers or similar guys23:04
flwang1but it's hard for us, even foundation to get feedbacks23:04
harlowjaidk, i've never seen the foundation try in this regard23:05
harlowjamaybe they have tried and i just don't know, lol23:05
flwang1since most of the users we(or foundation) can reach are dev, ops and sales :D23:05
flwang1not the real users23:05
harlowjaya, so i guess u get the data u can get, try to make an educated guess, and see where that goes23:05
flwang1and i'm not sure if foundation realized a thing23:06
flwang1the users of nova are different from the users of zaqar23:06
hongbinOK. THen let's send two ML, one for foundation, one for openstack-ops23:06
flwang1harlowja: that's silly, but the only thing we can do now.23:06
harlowjaya, can't hurt trying23:07
hongbinin my TODO list23:07
harlowjathe foundation really should have a way to do this (imho); cause like what else are they doing :-P23:07
harlowjafoundation rant over, lol23:07
flwang1harlowja: openstack has done a great job on the IaaS layer23:08
harlowjahmmmm, no comment23:08
flwang1haha, BUT23:08
flwang1as for the PaaS layer23:08
flwang1for those services like Trove, Sahara, Zaqar, Mangum, etc, etc23:09
harlowjawhen the stuff gets cloudy, the cloudy people get confused, lol23:09
flwang1there is a long way to go23:09
harlowjaand all those are cloudy :-P23:09
flwang1harlowja: yep23:09
flwang1harlowja: my point is23:09
harlowjacloudy confused people collborating confusingly23:09
flwang1the bad thing is we can't get feedback from the real user of those PaaS  services23:09
harlowjaya, cause they using other companies stuff :-P23:10
flwang1harlowja: don't say the name, pls23:10
harlowjathat's why u sneak subliminal messages and popups into other companies websites23:10
harlowjapopups like 'HAVE U THOUGHT OF OPENSTACK, CLICK HERE'23:10
hongbinvery funny23:11
harlowjathat's also why i was poking adrian to get some user study rackspace did for magnum to be released23:13
harlowjacause they did something and that's why they are more magnum focused, but nobody else knows quite what that something was23:13
harlowja(aka the data would be useful for the whole community, not just rackspace)23:14
harlowjai mean, knowing from what i've heard at godaddy, people really would like something like ECS, but people also want something ike K8s23:14
harlowja(but i don't have a big sample of people that i've heard this from, so...)23:15
hongbinfolks. I need to go now23:16
hongbinSee you all23:16
harlowjaif u must23:16
harlowjano no don't go23:16
harlowjaok u can go23:16
*** hongbin has quit IRC23:17
*** sputnik13 has joined #openstack-higgins23:31

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!