15:59:07 #startmeeting kolla 15:59:08 Meeting started Wed Nov 30 15:59:07 2016 UTC and is due to finish in 60 minutes. The chair is inc0. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:59:09 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:59:12 The meeting name has been set to 'kolla' 15:59:22 #topic rollcall, w00t please! 15:59:23 w00t!!! o/ 15:59:28 meow 15:59:40 meow meow 15:59:41 wirehead_, meo,,,,, 15:59:49 lol 16:00:00 woot 16:00:05 o/ 16:00:52 o/ 16:00:53 I forgot or we have new kollish? 16:00:54 sorry wirehead_ I am dog guy ;-) 16:01:02 o/ 16:01:07 ^._.^ 16:01:12 o/ 16:01:14 o/ 16:01:22 finally remembered the correct time 16:01:30 pbourke_: dito :) 16:01:45 yes, pbourke_ that would be hard for someone as far away from GMT as you are;) 16:01:53 w00t 16:01:54 nice to see both of you pbourke_, berendt 16:01:59 o/ 16:02:16 inc0: :P 16:02:18 WOOT 16:02:34 sdake, WOOT finally? 16:02:38 you got it 16:02:46 #topic announcements 16:02:49 one time only :) 16:03:37 I have one. I'm going for PTO + moving across country from 6th Dec till pretty much holiday. I won't be around (will try to join in every now and then but can't promise) 16:04:01 Jeffrey4l graciously agreed to help me and cover for me during this time 16:04:12 Jeffrey4l ++ 16:04:28 ++ 16:04:38 may need some help from sdake ;) 16:04:39 so, if there is anything needing immediate attention, he's your guys. I'll also can be reached by email or just wait till I'm back 16:05:06 anyway, that being said, let's kick this off by handing this meeting to Jeffrey4l 16:05:11 #chair Jeffrey4l 16:05:11 Current chairs: Jeffrey4l inc0 16:05:18 you're up:) 16:05:40 Jeffrey4l you got it 16:05:42 our agenda is not updated recently. 16:05:55 yeah, that's not news;0 16:06:19 so I can throw first topic 16:06:34 other annoucement from community? 16:06:42 our voting of splitting core teams ends today and so far we seems to be in agreement 16:06:42 on Jeffrey4l 's point - i think it would be helpful for everyone to update the agenda 16:06:52 rather then relying on one person to do the job 16:07:03 uh. looks like i missed this mail. 16:07:06 will do :) 16:07:28 re agenda: update the wiki page https://wiki.openstack.org/wiki/Meetings/Kolla before meeting please :) 16:07:34 well not splitting cores 16:07:47 that cores form kolla-ansible woudl vote for kolla-ansible cores and so on 16:07:52 that's effectively team split 16:08:55 ok. next topic? 16:09:27 #topic heka alternative 16:09:39 this is talked in http://lists.openstack.org/pipermail/openstack-dev/2016-November/108011.html 16:09:44 recently. 16:10:00 One might say we've discussed it a heka lot of times. 16:10:15 wirehead_ lol 16:10:28 i think we need make the choice right now. 16:10:28 yep. 16:10:43 i think the fluentd win finally. 16:10:53 Jeffrey4l right now or kick it to pike 16:10:54 any idea on this? 16:11:15 i have a bit of concern about the short schedule we have in ocata 16:11:24 ok. yes. 16:11:28 and not cratering kolla in the process 16:11:29 i would prefer to work on it right now 16:11:34 +1 to Fluentd, simply beause a lot of k8s support is there 16:11:36 hurray for fluentd ;-) 16:11:39 ok - again just an opion 16:11:58 fluentd works for me, i do not have a strong behavior.. 16:12:01 yeah fluentd is cool as it's cncf project now 16:12:05 i'd recommend we work on a stack of patches rather then merging 16:12:11 my reservations for fluentd were because of billing model afair 16:12:13 sdake: +1 16:12:17 sdake, +1 16:12:27 purupose of that is to hedge, in case the implementation doesn't work 16:12:30 but as its cncf project now, that's done 16:13:06 so for Snap credit, it will allow us to have similar functionality to heka in a month or so, as far as I know from team managing it 16:13:16 but I understand, maturity 16:13:37 i think we should try to make fluentd swapable 16:13:39 we also have a general rule inc0 we adapt to our upstreams and don't wait for them to publish features 16:13:45 that it is not a big deal to replace it in the future 16:13:54 maybe we have the same issue with fluentd in a few months 16:14:01 berendt no doubt about it ;) 16:14:15 so yeah, I think that thing is swappable on it's own 16:14:17 #action migrate from heka fluentd with a stack of patches. merged until it works. 16:14:31 not merged ? 16:14:36 Jeffrey4l: will you send out an update to openstack-dev? 16:14:47 Jeffrey4l you can use #undo to undo an action 16:15:04 #undo 16:15:05 Removing item from minutes: 16:15:10 #action migrate from heka fluentd with a stack of patches. not merged until it works. 16:15:11 Jeffrey4l and then do the action again - merged until works doesn't mean what I think you meant it to :) 16:15:17 sweet :) 16:15:22 Jeffrey4l: s/untill/when 16:15:23 sdake, thanks. 16:15:31 portdirect its gefn :) 16:16:12 why openstack-dev? 16:16:18 berendt, ^ 16:16:42 Jeffrey4l: we discussed it there and i think we should note that we decided to use fluentd 16:16:46 Jeffrey4l so people know the action 16:17:03 +1 16:17:05 should we send mail to openstack-ops, right? 16:17:18 Jeffrey4l, let's send email when we merge stuff 16:17:21 to ops 16:17:27 inc0: +1 16:17:29 OK. 16:17:35 i will send the mail. 16:17:39 as we might end up having some sort of strange technical issues 16:17:40 never know 16:17:52 next topic is? i have no right now 16:18:12 I wanted to talk about changes in core polict 16:18:21 #topic core team policy change 16:18:28 #link http://lists.openstack.org/pipermail/openstack-dev/2016-November/107457.html 16:18:55 voting ends today eod, but so far it's unanimous +1 so I expect it to pass 16:19:15 that effectively means that we will have separate core teams for all 3 deliverables 16:19:28 kolla, kolla-k8s and kolla-ansible 16:19:34 does this also mean that we split the meetings? 16:19:59 no, meetings, irc channel, community is the same 16:20:04 it's just for logistics of reviews 16:20:17 sounds good 16:20:43 should we merge kolla and kolla-kubernetes? 16:20:49 we work to clean up kolla-kubernetes core team to better show who really cares about project;) 16:20:58 qwang, no.... 16:21:01 qwang, no, both of them are separate 16:21:05 same as ansible 16:21:06 now 16:21:46 inc0 the cleanup stuff is intenral core business - and shouldn't be discussed in the meeting fwiw :) 16:22:33 well, fair enough;) 16:23:04 anyway, most importantly core teams now will have ability to nominate and vote for their own memeber 16:23:05 s 16:24:01 sdake / sbezverk: wanna talk about the progress on kolla-kubernetes helm? 16:24:10 +1 portdirect 16:24:24 heh, I guess we exhausted this topic?;) ok, let's move on 16:24:42 #topic kolla-kubernetes helm 16:24:56 finally we got into the agreement how helm temaplating will look like 16:25:01 +2 16:25:12 thank sbezverk, kfox1111 on it 16:25:36 so with merging a couple of PS people will be able to replicate merged solution and start adding templates 16:26:13 we have a white board with planned services/microservices please pick the one you like and go .... 16:26:34 sbezverk i think the whiteboard is not filled out 16:26:48 sbezverk i'll do that after this meeting assuming i don't have another ;) 16:26:56 https://blueprints.launchpad.net/kolla-kubernetes/+spec/helm-microservices 16:27:12 do we have any example on how to do this? which will be helpful for others to join. 16:27:13 ya no work items 16:27:24 Jeffrey4l indeed that is what the whiteboard patchsets are 16:27:36 sdake: oh so we did not move from helm-packages BP the whiteborad we had there? 16:27:39 Jeffrey4l mind adding an action for me to fill out the work items 16:27:46 sbezverk not whiteboard, the work items 16:27:52 the whiteboared is for status udpates 16:27:58 the work items are for organizing the work 16:28:01 sdake: right 16:28:03 #action sdake add work item for bp helm-microservices 16:28:24 #link https://blueprints.launchpad.net/kolla-kubernetes/+spec/helm-microservices 16:28:35 I've also been working on a dev env, based on halycon-k8s, that I've started to docs for - so once kfox's work is merged in, i'll try and get that doc'd as well and we should be able to get a real head of steam up :) 16:28:42 portdirect ++ 16:29:05 portdirect so your dev env, i'm using it 16:29:07 what else is needed? 16:29:24 feel free to take some item from this bp. we'd better finish this before m2. 16:29:25 good job portdirect, I saw the results of your doc in sdake testbed, looks good 16:30:03 not much really - though I'm gonna retire my halycon-k8s fork (hopfully today) 16:30:21 and document getting kubectl and helm clients installed on the localhost 16:30:40 anything else on this topic? 16:30:42 from there the workflow should be really easy for a developer to use 16:30:55 do we also need a setup_dev.sh ? 16:30:59 yay. :) 16:31:24 jascott1: on it :) 16:31:29 portdirect thats a great idea - kubectl and helm clients on lcoal host 16:31:45 jascott1 defien setup_dev.sh for us :) 16:32:01 and destroy_dev.sh... 16:32:05 things from setup_gate.sh but for local dev 16:32:16 ^^ 16:32:32 one thing i'd like to ensure is we dont muck up developer systems with a bunch of crap 16:32:38 reference: devstack 16:32:45 people run devstack in vms for a reason 16:32:54 there are these things called containers... 16:33:01 i want to run kolla-kubernetes on my bare metal 16:33:09 Yeah, making kollak-k8s not be like devstack was a very large motivator. :) 16:33:39 rahter i want to do dev of kolla-kubernetes on bare metal 16:33:44 I have an idea, let's deploy it with kolla and it will be all great 16:33:46 if that means its in a vagrant box on my localhost 16:33:46 sdake: thats very close :) 16:33:47 thats cool 16:33:55 this is what I used: https://github.com/kubernetes/kube-deploy/tree/master/docker-multinode 16:34:17 portdirect what you got now is good for devs 16:34:21 only issue for multinode is you need a 2nd bm machine or a vm 16:34:28 portdirect setup_dev.sh may be good too - have to see in review :) 16:34:34 the vagrant is just setting up come vm's to then run ansible on - I'll document running on baremetal... 16:34:48 portdirect nah thats not what i meant 16:35:12 portdirect your docs are perfect from my pov +/- some deltas 16:35:30 rhallisey, it's a good option 16:35:54 portdirect devs dont want their systems destroyed by the development environment 16:36:00 if you have 1 vagrant box setup to run https://github.com/kubernetes/kube-deploy/blob/master/docker-multinode/worker.sh 16:36:00 which in essence, is what devstack does 16:36:12 & run on your host: https://github.com/kubernetes/kube-deploy/blob/master/docker-multinode/master.sh 16:36:18 sdake: +1 16:36:46 portdirect so keep at it, my only recommendaiton there is to keep setup_dev.sh whtaever that is, constrained to not installing a bajillion things :) 16:37:15 portdirect i think your docs are pretty close to mergeable now as is but we can sort that out in gerrit 16:37:27 portdirect if you do a setup_dev.sh would you do it in a followup? 16:37:56 whenever i see setup_dev.sh I think "setup_devSTACK.sh" :) 16:38:19 sdake: yeah - it wont do anything other than setup kubectl and helm clients 16:38:22 #link https://github.com/kubernetes/kube-deploy/tree/master/docker-multinode 16:38:40 rhallisey v1k0d3n and crew have produced a nice development environment that we are using as an upstream 16:38:53 cool 16:39:14 i really like it alot 16:39:22 me too 16:39:51 let's move on? 16:39:56 runs on one node - simualtes multiple nodes, fast to setup, fast to teardwon 16:39:56 rhallisey: kube-deploy is better for production (ATM at least in my opnion) - but this is just for getting a quick and dirty mutinode env up and running 16:40:02 hits all the checkboxes ;) 16:40:28 Jeffrey4l: yeah 16:40:31 Jeffrey4l sure 16:40:32 nice 16:40:37 #open discussion 16:40:46 oh wait one moment jeffrey4l 16:40:54 maybe i wasn't paying attention 16:41:08 did we have the kolla-kubernetes topic? 16:41:16 sdake, we are just now. 16:41:39 ok in open discussion - wfm :) 16:41:42 but it is marked as `kolla-kubernetes helm` 16:41:47 o/ 16:41:48 oh i see. 16:41:55 ok, well i'll do a general update here in open discussion 16:41:55 * srwilkers_ is very late 16:41:56 sup srwilkers_ 16:41:58 srwilkers_: o/ 16:42:04 srwilkers_ its ok - volunteer effort :) 16:42:20 anyone want talk? 16:42:25 basically the plan is to crank out the microservices 16:42:42 after the microservices are done 16:42:48 we will move up a layer to thte service layer 16:42:55 how that works is to be defined 16:43:00 can we set the topic back to helm? 16:43:08 thats all I have duonghq :) 16:43:16 so - I have a topic that may be controversial re ceph 16:43:31 sdake, I think we should refer to rhallisey specs 16:43:56 duonghq agree - there is no spec for this part of the system 16:44:21 in rhallisey's specs, we have 4 option, seem that we nearly have consensus no option 4 (iirc) 16:45:05 duonghq moment let me read the spec 16:45:30 should we continue to maintain kolla-k8s ceph - or make use of the work that the ceph team are doing upstream? - Not a strong opioninion either way but seems like we are replicating effort. 16:45:43 #link https://review.openstack.org/#/c/392257/ 16:45:59 #link https://review.openstack.org/#/c/392257/ 16:46:23 portdirect on the containers aspect, we must use our containers 16:46:44 portdirect the ceph containers themselves don't do ubuntu for example 16:46:45 did want to mention one thing real quick 16:46:46 portdirect, does ceph fro upstream runs on k8s? 16:47:39 Jeffrey4l: yes and they have a good set of templates 16:47:43 dealing with some internal stuff so i'll be around here and there 16:47:58 sorry I joined late 16:47:58 just wanted everyone to know where I've been 16:48:06 vhosakot, o/ 16:48:09 portdirect got a link 16:48:12 o/ 16:48:14 but sdake's point clarifyted the policy I was unsure of - so I'll keep on it 16:48:21 sdake, had kindly agreed to take the helm 16:48:36 snare drum crash 16:48:37 rhallisey, nice ;) 16:48:46 sdake: https://github.com/ceph/ceph-docker/tree/master/examples/kubernetes 16:48:52 portdirect thanks 16:48:56 i'll check it out 16:49:02 #link https://github.com/ceph/ceph-docker/tree/master/examples/kubernetes 16:49:05 wirehead_, did you see that pun ^ 16:49:13 we dont want to duplicate work that is happening upstream, no 16:49:18 where's the applause? 16:49:30 although if the work happening upstream conflcts with our rchitecture, we are forced into it 16:49:37 rhallisey: are you here all week :D 16:49:49 hehe 16:49:57 rhallisey, awesome 16:49:58 rhallisey applause :) 16:50:38 rhallisey, oh i see what you did there 16:51:02 ok duonghq when I said move up a lyaer, I wasn't talking about moving up the layer to operators 16:51:12 I was talking about moving up a "sub-layer" in layer-3 16:51:14 sdake: it would be very eay to take what we needed and put them in our containers though - I've been working on that 16:51:46 duonghq as in services next - all part of the helm layer - layer 3 16:52:18 sdake, got it 16:52:23 portdirect there are CLA problems with that 16:52:47 the work submitted to the repo must be original work of your own, not someone elses 16:53:05 so the rhallisey's operators in review queues need waiting little more time? 16:53:12 that is unless they have also signed the cla :) 16:53:18 sdake: roger 16:53:26 duonghq i think that operators can be worked in parallel 16:53:35 duonghq, it's in decent shape 16:53:39 duonghq if people want to get on that, wfm :) 16:53:42 it was meant to be a foundation 16:54:23 sdake, rhallisey we have some diverse in sort out operator and helm stuff in bp organization? 16:54:43 duonghq that didn't parse 16:54:58 specifically diverse 16:55:13 sdake, I mean for the operators, we have 1bp/operator, but for helm stuff, we have 1bp/layer 16:55:16 do you mean that the launchpad needs better organiation? 16:55:18 the bps for each operator hasn't been made 16:55:32 duonghq: confused a bit, what we do with helm now is perfectly fits with operator concept 16:55:52 should do a bp per operator service 16:55:57 since that work will be split up 16:56:00 disagree 16:56:02 where operator would call helm install microservice to deploy what ever it needs 16:56:07 the best way to handle that is work items 16:56:11 one master blueprint 16:56:39 whatever works 16:56:46 well its the openstack way :) 16:56:53 sbezverk, we are talking about "layer" in rhallisey's spec 16:57:17 I roll my own way :) 16:57:19 sdake, okay 16:57:48 time almost up 2 min 16:58:45 duonghq happy to have further conversation with you to clear things up after meeting since we are short on time 16:58:49 in #openstack-kolla 16:58:53 also 16:58:57 sdake, sure 16:59:02 got a new name for operators 16:59:04 vessels 16:59:13 mariadb vessel 16:59:16 make it harder ;) 16:59:21 rhallisey, didn't got your idea 16:59:26 *get 16:59:42 sdake, operator is hard enough 16:59:47 it means 2 things 17:00:10 ok. time is up 17:00:19 thanks people for coming 17:00:34 #endmeeting