16:01:31 #startmeeting containers 16:01:31 Minutes: http://eavesdrop.openstack.org/meetings/bgpvpn/2017/bgpvpn.2017-03-21-15.01.html 16:01:32 Minutes (text): http://eavesdrop.openstack.org/meetings/bgpvpn/2017/bgpvpn.2017-03-21-15.01.txt 16:01:34 Log: http://eavesdrop.openstack.org/meetings/bgpvpn/2017/bgpvpn.2017-03-21-15.01.log.html 16:01:35 Meeting started Tue Mar 21 16:01:31 2017 UTC and is due to finish in 60 minutes. The chair is adrian_otto. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:36 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:38 The meeting name has been set to 'containers' 16:01:48 #link https://wiki.openstack.org/wiki/Meetings/Containers#Agenda_for_2017-03-21_1600_UTC Our Agenda 16:01:54 #topic Roll Call 16:01:58 o/ 16:01:59 o/ 16:01:59 Adrian Otto 16:02:02 o/ 16:02:04 Corey O'Brien 16:02:06 Perry Rivera 16:02:10 Ton Ngo 16:02:33 hello Drago swatson hieulq_ coreyob juggler and tonanhngo 16:04:05 o/ 16:04:05 is it just me, or does IRC seem really laggy today? 16:04:19 hello jasond 16:04:24 I haven't noticed anything yet 16:05:01 i'm seeing a 0.0 second lag 16:05:04 ok, glad to hear it. Thanks swatson. 16:05:06 seems ok 16:05:23 #topic Announcements 16:05:34 1) I have appointed Spyros Tigazis as our release liaison for Magnum. 16:05:40 #link https://wiki.openstack.org/wiki/CrossProjectLiaisons#Release_management Release Management 16:06:09 this means that he will be assisting me with submitting releases in Gerrit during the Pike cycle 16:06:20 any questions/concerns/remarks? 16:06:35 congrats Spyros 16:07:15 he's normally here for our team meetings, but he did not show up today. 16:07:20 I hope he's ok. 16:07:21 yatinkarel,o/ 16:07:25 hello yatinkarel 16:07:31 Congrats Spyros 16:07:43 any other announcements form team members? 16:07:49 s/form/from/ 16:08:13 Just as an FYI, jvgrant is out this week for vacation 16:08:24 tx swatson 16:08:55 #topic Review Acton Items 16:08:57 (none) 16:09:01 #topic What should we call Magnum's generic OSC name stem? 16:09:37 ok, so swatson has started on a plugin for osc that will allow the openstack client to control magnum 16:09:55 The "openstack cluster" command is taken by Senlin, so we need an alternative. 16:10:03 #link https://review.openstack.org/#/c/446179/3/setup.cfg@33 16:10:27 the initial choice was "infra" but a few of us dislike it 16:10:46 so I emailed out ML yesterday about it,a dn got a bunch of clarity 16:11:06 first thing I want to cover is that this is not the same thing as our service catalog name 16:11:07 a dn -> and? 16:11:25 which will remain container-infra 16:11:48 and it's also not our Service Type name, which will likely end up "coe-cluster" 16:11:55 and... 16:12:00 got it thx 16:12:09 the command name MAY have a space in it 16:12:34 so I have 4 of them for us to consider as a team, and I'd like to judge if we have a consensus for any of them. 16:12:36 Ready? 16:12:51 ok 16:12:59 Choices: coe cluster, coe, infra, mcluster 16:13:15 ccluster? 16:13:25 Ones that were eliminated from consideration are: c_cluster, coe_cluster, coe-cluster, magnum, magnum cluster, ccluster 16:13:34 the rationale behind disqualifying ccluster is that it looks like a typo 16:13:43 ah ok 16:13:54 the name must be [a-z]+ only 16:14:13 now, let me explain what it means to have a command name with a space in it 16:14:25 if we select "coe cluster" as the name it will yield commands like this: 16:14:32 openstack coe cluster create ... 16:14:48 and if we chose "infra" it would be: 16:14:55 openstack infra create ... 16:15:16 would we also have "coe cluster template"? like "openstack coe cluster template create"? 16:15:26 since the noun we actually create is a "cluster" I prefer the former. 16:15:27 thoughts? 16:15:28 Is there other existing command with a space? 16:15:31 yes, coreyob. Exactly. 16:15:41 yes, volume commands have spaces in them 16:15:50 adrian_otto: We can create commands with any name after our noun. Currently my change has commands that read as "openstack infra cluster create..." and "openstack infra cluster template create..." 16:15:57 Would that go against the guidelines? 16:16:13 are there any other services that have initialisms for their command? seems like "coe" is pretty jargon-y 16:16:53 swatson: that would probably be okay too, but it is a slight departure from what dtroyer explained in our ML thread. 16:17:20 Ah. I'll have to go through it once I'm back in the office 16:17:34 I would have liked that word to be "magnum" but that's against the guideliens 16:17:47 as our project names are abstract,and they want nouns that are descriptive 16:18:28 like "object"? :) 16:18:47 heh, yeah, that's the drawback of this approach 16:18:55 Of the 4 choices I think "coe cluster" is best 16:19:22 guessing there is no way to share cluster with senlin? the "resource" we are creating is a "cluster" so it seems logical that we call it that. unfortunate that only 1 openstack service will ever have "clusters" 16:19:36 It seems that if we choose "coe" we'll end up with "openstack coe create" instead of being able to use "openstack coe cluster create". Is that right? 16:20:16 coreyob: we don't have a way to discriminate which "cluster" resource the user wants. It would require some form of an argument, or magic guessing to figure it out 16:20:28 and that would violate the principle of least surprise 16:21:25 swatson: tat' my interpretation of yesterday's discussion, yes. 16:22:09 can we have multiple nouns? could we have 'openstack swarm cluster create' and 'openstack kubernetes cluster create' etc etc? 16:22:42 from the options awailable coe cluster looks ok too me 16:22:43 just trying to think outside the box 16:23:01 coreyob: that's a really good idea, and a great question for us to pose. 16:23:02 coreyob, we can use --coe swarm for that 16:23:29 yatinkarel: but he makes a very creative suggestion to use the name of the COE to avoid the namespace collision 16:23:41 yatinkarel: --driver swarm, probably 16:23:55 Drago, yes 16:24:24 coe/driver is pluggable though, so I don't think it is feasable 16:24:45 coreyob, adrian_otto we would have multiple coe then, dcos is also coming up 16:24:55 it would pepper the namespace with all the different driver names 16:25:25 which is likely to be discouraged 16:25:48 so far we have kubernetes, swarm, mesos, and dcos 16:25:58 and the "kubernetes" one has multiple drivers 16:26:13 so you'd still need a driver argument to select whatever one is not default 16:27:02 Would "coe cluster" really be accepted? It seems to me as obscure as "magnum" would be 16:27:08 yes 16:27:34 yeah coe seems pretty odd considering it really isn't a word 16:27:40 how about "container infra"? a little wordy, but to the point 16:27:41 it's equally obscure, but there is one key difference "coe" is in fact the type of cluster it is 16:28:04 Maybe we can turn "magnum" into an acronym with the same meaning as coe, and then use it 16:28:09 the word "container" is reserved for the smae reason that "cluster" should have been 16:28:27 had we been more forward thinking back then 16:28:46 if you google "coe cluster" it's not clear what it means 16:28:55 Zun has decided to use "appcontainer" as their noun. 16:29:16 infracontainer? 16:29:20 does that imply containerinfra is possible? 16:29:25 and "container" is actually in use by swift 16:29:32 coreyob: yes 16:29:33 I like mcluster for its conciseness 16:29:47 appcontainer and not app container? 16:29:55 mcluster might be allowed, but I'm not 100% sure yet 16:29:56 juggler: +1 16:30:03 mcluster where m is for magnum? i think that only makes sense if you know you're looking for something that is called "magnum" 16:30:18 it's not obvious what m means in mcluster 16:30:19 the problem with mcluster is that it's not actually a noun 16:30:21 nothing else in osc uses the project name 16:30:28 I think "mcluster" is similar enough to "cluster" to be confusing, as well 16:30:30 coreyob: right. 16:30:36 Plus it makes me think the noun is "McLuster"... 16:30:41 lol 16:30:43 hahahaha 16:30:47 McCluster 16:30:52 a song drops into my head... :) 16:30:52 that's hilarious 16:30:56 ok, so that's out!! 16:31:00 :) 16:31:18 Hi, I think COE is getting quite accepted and it is sort of generic, except if at some point magnum cluster aren't COE 16:31:20 what is tld 16:31:30 clusters, folks, and fun...! 16:31:37 openstack tld list 16:32:13 i think coe is well known here in magnum. if you google it, magnum and rackspace are the only ones that use it. in the wider container ecosystem though it isn't well known 16:32:15 That's a good point, magnum doesn't technically only produce container infra clusters, that just happens to be the only type of clusters it produces 16:32:48 *it is able to produce atm 16:33:15 (@ArchiFleKs) 16:33:42 coreyob: that's because we made it up 16:33:44 what about cicluster? 16:33:55 Drago: atm? 16:34:01 ci stands for container infra 16:34:02 ArchiFleKs: at the moment 16:34:23 but the prevailing meaning for CI is Continuous Integration 16:34:32 hmm 16:34:39 https://www.docker.com/cp/container-orchestration-engines 16:35:05 disallowing hyphens is soo helpful... 16:35:28 Drago the technology just isn't there yet 16:35:48 underscores disallowed? 16:36:08 juggler: only ^[a-z]+$ 16:36:14 actually 16:36:16 and space too 16:36:20 juggler: only ^[a-z ]+$ 16:36:46 how bout usage of "+"... 16:36:48 :) 16:37:14 adrian_otto: did we rule out just "coe"? 16:37:25 no 16:37:28 jasond I think that one is still in the mix 16:37:31 we are still in discussion mode 16:37:58 did we rule out container cluster? 16:38:15 can't have container 16:38:26 IMO, of the 4 options presented, "coe" would be best if we could just add on to it with whatever commands, e.g. "openstack coe cluster create", "openstack coe cluster template create", etc. 16:38:26 Drago: container is a reserved work (swift) 16:38:27 adrian_otto: space are allowed but it cannot collapse with other things? Like "container orchestration" would it collapse with swift? 16:38:27 yeah but it's "container cluster" 16:38:35 i like "coe" better than "coe cluster" 16:39:25 I am in full agreement with swatson on this one 16:40:08 In my work on the plugin it seems to me we can do precisely that, but it might be against guidelines if I'm understand what adrian_otto has specified 16:40:15 "openstack coe ca show ..." 16:40:21 s/understand/understanding 16:40:40 swatson: +1 16:40:49 so that's four votes in support of coe 16:40:58 i am with mcluster(if allowed) and coe cluster 16:41:04 any opposing viewpoints to condsider 16:41:11 I'm just going to -1 the openstackclient in general 16:41:12 Yes and if magnum is not doing just coe but other cluster in the future we can change the meaning of coe to cluster orchestration engine :p 16:41:30 one last suggestion, would "containerorchestration" be better than "coe"? 16:41:53 that's pretty darn long 16:41:58 for an interactive CLI 16:42:06 how much typing (if any) of that long term would be involved? 16:42:35 at least 10 characters if word completion logic is implemented in the client 16:42:53 also bash_completion is there 16:43:25 cluster1 16:43:39 adrian_otto, bash completion didn't work for more than 10 character? 16:43:47 because everyone knows that "password1" is better than "password" 16:44:02 yatinkarel: good question, I dunno 16:44:02 another suggestion: is there a term between "containerorchestration" and "coe" which is a descriptive happy medium and yet not too long? 16:44:10 coreyob: or containercluster 16:44:31 true true 16:44:58 These rules are weird. Nouns must have spaces unless you smash two together, but then you can't use hyphens 16:45:25 they can have spaces 16:45:27 the problem is taht the word container is in use 16:45:51 that's why "coe cluster" was proposed 16:45:58 is case insensitive in this matter? 16:46:16 required lower 16:46:31 coreyob thx 16:47:21 I'm going to +1 "coe" because it is the least awful 16:47:25 ok, so we have a weak consensus around "coe" 16:47:30 Drago heh 16:47:41 we don't have to make a final decision immediately. 16:48:04 but we should give swatson a substitute for now, unless we all really want to stick with "infra" 16:48:05 adrian_otto: i'm not sure "coe cluster" makes sense, because then you'd have "coe cluster ca show ..." and "coe cluster cluster create ..." 16:48:09 yeah, I'll stick with "coe" but I'm open to hear more options and discuss 16:48:26 jasond: I agree 16:48:49 one note: Currently with "infra" I have commands like "openstack infra cluster create", but that might be against the guidelines Dean Troyer specified in the ML 16:49:09 we can ask to clarify that 16:49:29 ok, time to advance to the next agenda item 16:49:30 adrian_otto I think we should, because if we can't do something like that with "coe" then I think "coe cluster" would be the go-to 16:49:36 are ther eupdates on BP's or shold we skip those? 16:49:59 Well, can we have multiple "nouns" that basically all start with "coe"? e.g. coe cluster, coe ca, coe cluster template 16:50:00 on the name selection, please monitor the ML and participate there 16:50:15 Drago: yes, we can 16:50:33 ok 16:50:36 #topic Open Discussion 16:50:45 if you have BP updates to share, you may make them now 16:50:59 Has anyone deployed k8s-fedora-ironic cluster on devstack+ironic+magnum setup, ironic nodes are going to clean_failed state, no hosts available 16:52:33 yatinkarel: is that setup able to produce BM instances through Nova using the ironic virt driver? 16:52:57 adrian_otto, not tried yet 16:53:29 i tried but that time node were in clean_failed state 16:53:30 try that, so you can narrow it down to whether ironic is even working, or whether magnum is doing something to jam it. 16:53:51 i will try again 16:56:20 ok 16:56:41 any other discussion topics before we adjourn today? 16:58:18 Our next team meeting will be on Tuesday 2017-03-28 in #openstack-meeting-alt at 1600 UTC. See you then! 16:58:21 #endmeeting