Tuesday, 2016-06-21

*** yanyanhu has joined #openstack-higgins01:53
*** hongbin has joined #openstack-higgins02:19
*** yuanying has quit IRC02:51
hongbinHi all, our team meeting will start in about 7 mins02:53
*** yuanying has joined #openstack-higgins02:54
hongbinQiming: ping. Team meeting if you have chance03:02
*** klkumar has joined #openstack-higgins03:35
*** sheel has joined #openstack-higgins03:40
*** hongbin has quit IRC03:59
*** yuanying has quit IRC04:02
*** yuanying has joined #openstack-higgins04:03
*** klkumar has left #openstack-higgins04:11
*** coolsvap has joined #openstack-higgins04:32
*** irenab has joined #openstack-higgins05:27
openstackgerritChangBo Guo(gcb) proposed openstack/higgins: Don't include openstack/commmon in flake8 exclude list  https://review.openstack.org/33201208:41
*** coolsvap has quit IRC09:15
*** coolsvap has joined #openstack-higgins09:17
*** openstackgerrit has quit IRC09:18
*** openstackgerrit has joined #openstack-higgins09:18
*** yanyanhu has quit IRC10:30
*** sheel has quit IRC11:15
*** irenab has quit IRC13:29
*** openstackgerrit has quit IRC13:48
*** hongbin has joined #openstack-higgins14:32
-openstackstatus- NOTICE: Gerrit is being restarted now to apply an emergency security-related configuration change16:05
*** irenab has joined #openstack-higgins16:46
*** coolsvap has quit IRC17:51
hongbinharlowja: dims : FYI, there is a discussion at yesterday meeting about if Zun should support COEs or not19:00
harlowjauh ohes19:00
hongbinThe topic extend to if Zun should compete with existing COEs...19:00
harlowjai have a genuine feeling that it might be easier to just work with something like k8s instead of doing that :-P19:01
hongbinIn general, I let the team to work on this etherpad to resolve the debate19:01
hongbinYes, the development efforts is minimize if we work with COEs instead of re-inventing the whole things19:02
hongbinHowever, there are cons that has been pointed out19:02
hongbinI am divided right now19:02
dimshongbin : let's let it run its course19:03
harlowjaand don't forget to take your laxatives19:03
dimshongbin : watching the dirty-sausage-making in kubernetes, it's a monster to do a COE19:03
harlowjathou does not want to watch sausage-making19:04
dimsnot to mention docker orchestration announcements from yesterday. we'd be marginalized pretty quickly19:04
harlowjaunless we have an army appear overnight19:05
harlowja(developer army)19:05
harlowjai don't see that existing, lol19:05
hongbinBetter to avoid direct competition. That is my feeling19:05
harlowjadepends on your ambitions :-P19:06
harlowjaand reality of those ambitions, ha19:06
harlowjai mean, these COE things are in all honesty, not super-complicated, u just have to have the right core pieces, and its all the same afterwards19:08
hongbinharlowja: dims appreciate if you have chance to cast your comments on https://etherpad.openstack.org/p/zun-architecture-decisions19:08
dimshongbin : done19:16
hongbindims: thx :)19:17
hongbindims: you don't like option 4?19:19
hongbindims: Interest to know your opinions about that19:19
dimshongbin : i'd prefer we go and contribute where there is a community already. it's counter productive just to fork something19:22
hongbinI see19:22
hongbinBut we might come up with several OpenStack specific patches, that might not be easier to be acceptted19:23
harlowjatis life19:24
harlowjalife tough, but that's ok19:24
dimshongbin : yep, totally, "community over code"19:25
harlowjanow i know its not easy, but meh, people say the same when they have patches to openstack, its a urge/feeling imho that u have to get over19:26
hongbinThat is true19:26
* dims looks up his list of abandoned patches :)19:27
harlowjai don't want to look at mine19:27
hongbinThere is another difficulties to work with existing COEs. That is multitenancy19:31
hongbinCOEs generally didn't follow the OpenStack multitancy model19:31
dimshongbin : right. so i was looking at options for authentication/authorization and multitenancy in k8s and it's not pretty19:31
dimskevin fox is doing a LOT of work around this in Kubernetes19:33
harlowjadims well the only kind of multitenancy that i think can start to work in k8s is the namespace stuff, and u basically have to expand namespaces to have/be more powerful tenancy19:33
hongbinmapping a namespace to Keystone tenant, maybe19:33
harlowjaya, its not imho sufficent but its a start19:34
harlowjau almost need like a /<tenancy-paths-here/tree>/<users-namespaces>19:34
harlowjaif u want to map it that way, or perhaps it just needs a better model (inside of trying to morph namespaces into that019:34
harlowja*into that)19:35
harlowjanow maybe thats an oppurtunity for this group19:35
harlowjato fill those gaps19:35
hongbinI wonder if Hypernetes already did that19:35
harlowjathey put a thing in front of kube to do that afaik19:37
harlowjahyper-api or something19:37
hongbinI see19:37
harlowjabut imho that's a solution, although sorta imho, not the best one (the best one requires working with the k8s for a native solution)19:37
dimshongbin : harlowja : https://github.com/kubernetes/kubernetes/issues?utf8=%E2%9C%93&q=%20author%3Akfox1111%20keystone19:38
dimssee the one about authorization19:38
harlowjaya 'In future PR's, I hope to implement full Multitenancy so that additionally, one k8s cluster could be used for the entire OpenStack Cloud, and Projects are isolated from each other though some mechanism (perhaps mapping keystone project = namespaces?)'19:38
harlowjai need to get those details from kevin19:38
harlowjaoverall all of this is how i found that k8s == not simple at all19:39
harlowjain my ideal pretty world, k8s folks would of actually worked with openstack when making k8s, cause alot of the crap is the same19:39
dimswhat? blasphemy!19:40
harlowjasomeone took my keyboard over that typed that19:40
harlowjathat was weird19:40
harlowjahaving higgins work with those folks to fix that working-together-problem/situation, could be useful19:41
harlowjabut it depends on the group imho19:41
hongbinyes, depends on the team :)19:42
harlowjaya, and i don't know really how k8s stuff will work out; i can't stand github PR for example, lol19:42
harlowja*mainly the UI around github really starts to feel bad for k8s (which is to big)19:43
harlowjaso ya, hongbin its not an easy choice here :-P19:50
harlowjaimho its one magnum avoided (but times were different then)19:51
hongbinI will just let the team to decide19:51
hongbinMy personal take is we can just folk COEs to start, and contribute it back if we can (seems like another Hypernetes ...)19:53
harlowjamy personal take would be to pick k8s as the COE we will work with, work with the existing groups that are forming to do that integration, and hope for the best19:53
harlowjathat decision though wouldn't be a 'make an abstraction over COE's or ...'19:54
harlowjai'd be making the tough choice of picking one19:54
harlowjaand accepting that fate19:54
harlowja(whatever it ends up being)19:54
harlowjabut i'm also biased19:54
harlowjaif its an idea that people want to run with, i'm happy to help introduce to people that i know, help get things setup and such19:56
hongbinharlowja: How the Hypernetes folks doing. Did k8s community accept their code in upsteam?19:56
harlowjathe hyper folks changes @19:56
harlowjaHere is a list of changes made by Hyper: https://docs.google.com/spreadsheets/d/1BKQrURk31R4BvgQdCZ8LlHT9Z-bN6bEtUEQ4Il4yWXA/edit?usp=sharing ,19:56
harlowjawhich is based on our commits here https://github.com/hyperhq/hypernetes/commits/master19:56
harlowjaso no, imho they are trying to not have there startup die vs doing the k8s upstream stuff :-P19:56
harlowja(or that's my general feeling, lol)19:57
harlowjathey want that to happen (which is good), i'm just not quite sure if they have the resources to do that (upstream the changes, work through the process/changes...)19:57
hongbinThe list is huge20:00
hongbinI guess it is hard to push them all to upstream20:00
harlowjaits probably not there highest priority imho20:00
harlowja(for better/worse)20:00
harlowjathe good thing about the k8s stuff imho is there is https://docs.google.com/document/d/1iAQ3LSF_Ky6uZdFtEZPD_8i6HXeFxIeW4XtGcUJtPyU (a working group trying to better the openstack integrations)20:01
hongbinThe list is expanding or shrinking in your observation?20:01
harlowjahongbin the list just got published a week ago, so unsure ;)20:01
hongbinI see20:01
harlowjai can think of a few things that k8s needs to do better at20:02
harlowjamulti-tenancy is one20:02
harlowjahow it would/could/should integrate with nova (or ironic or?) to get 'minions' (the things that run containers) is another20:02
harlowjahow it would/could/should work with keystone (in a non-experimental way)20:02
harlowjaand then there is cinder as well (what is that plan)20:03
hongbinThe idea of integrating with Nova is cool20:03
harlowjawell it unravels the question of multiteancy at that point20:03
harlowjanot a bad problem, but one that needs likely a bunch more work vs an experimental keystone thing20:03
harlowja(that work is a good start, obviously)20:04
hongbinFrankly, it is hard to doing everything (nova, keystone, multitenancy, cinder) in upstream20:04
hongbinUnless do it in a folk20:04
hongbinLike Hypernetes, or kubernetes-mesos20:04
harlowjaya, connecting two communites that haven't been connected before is somewhat akin to surgery on brains and connecting 2 brain-halfs20:05
harlowjaalthough i am not a doctor20:05
hongbinyou are not a doctor :)20:05
harlowjabut i could be20:05
harlowjawho wants me to join there brain-halfs20:05
harlowjadims  u ok with being my guinea pig?20:06
* dims is not here :)20:06
harlowjau likely won't be after surgery either20:06
harlowjaso ya, overall this question isn't easy to weigh, its not easy in any of the choices IMHO :-P20:07
harlowja10 choices, pick one, lol20:08
harlowja(each with there unknowns)20:08
*** flwang has left #openstack-higgins21:37
*** hongbin has quit IRC22:45

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