Monday, 2015-06-15

*** zhiwei has joined #kolla00:56
sdakeevening folks01:05
mandrehi sdake01:05
sdakehey mandre01:11
sdakemandre here is my tube amp https://drive.google.com/file/d/0B8q6xDPETSkHWjdoaFFCT3RJMzUtLW1oSWlXY1lOMzVLcmJV/view01:12
mandrenice :) party time!01:16
*** erkules_ has joined #kolla01:22
*** erkules has quit IRC01:25
*** fangfenghua has joined #kolla01:35
sdakefangfenghua sup bro where ya been!01:37
*** fangfenghua has quit IRC01:44
*** alisonholloway has joined #kolla02:09
*** dimsum__ has quit IRC02:22
*** dimsum__ has joined #kolla02:36
SamYapleyo02:41
SamYaplehows it going02:41
sdakeyo samyaple02:46
sdakeany word on single nic networking? :)02:46
SamYapleheading into the office back in 1002:47
sdakelater02:47
*** dimsum__ has quit IRC02:58
SamYaplesdake: still not sure i can do single interface with linuxbridge03:04
SamYaplethere is a reason the neutron guys dont _really_ support linuxbridge03:04
SamYapleit was shoved into neutron anyway03:04
sdakecan you sync up with daneyon03:04
sdakehe is a network rocket scientist03:04
sdakei'll shoot him a mail to hunt you down03:05
SamYapleheh probably not. we are around at polar opposite times03:05
SamYapleping he pongs whe nthe other sleeps, been going on for weeks03:05
sdakemaybe email :)03:05
SamYapleyea sure. i dont normally check my email so communication with me there is spotty. ill give it a go03:06
SamYaplethe quickest way would probably just be to implement the OVS container03:07
SamYaplesince that one is actually supported all around03:07
SamYaplesdake: any idea on that spec? i cant really continue until its approved. Maybe a new version with the thoughts summarized again?03:11
sdakesamyaple you know what I did today03:12
SamYaplebuilt a house03:12
sdakefilled out 4 expense reports03:12
sdakeall day doing that03:12
sdakemonday for new spec03:12
SamYaplewell maybe dont wait til the last second03:12
SamYaplepfff this guy03:12
sdakei know your blocked03:12
SamYapleso for extracting all the crudini/autoconfiguration bits in preperation for ansible, i see things like https://github.com/stackforge/kolla/blob/master/docker/centos/binary/keystone/start.sh#L13203:13
SamYaplesince most all of _all_ of the start.sh scripts are for crudini and autoconfiguration, are we basically going to move start.sh to crudini.sh?03:14
sdakei had a similar thought03:14
SamYaplemost of the services can be started without anything else03:14
sdakeare we keeping crux as a dep with ansile03:15
SamYapleno03:15
SamYaplehttps://gist.github.com/anonymous/564e13a5c15a166998b603:16
SamYapleso that was the idea i had03:16
sdakeset_configs does what03:17
SamYapleoops let me post that code03:17
SamYaplehttps://gist.github.com/SamYaple/eddf94ecc94bf4dbe4c503:18
sdakeits more like bind-readonce03:19
sdakeand bind-readalways03:19
sdakeright?03:19
SamYapleyes, but bind-copyonce|bind-copyalways would be most accurate03:19
SamYaplebut ignore that part03:20
*** tobe has joined #kolla03:20
SamYaplei was just playing with that to see how it felt03:20
sdakeok i'll use that in the spec03:20
sdakewe need to pass the parameter (always vs once) to the bind script03:20
sdakeor have  two different scripts03:21
SamYapleit would be in $CONFIG_TYPE03:21
SamYaplelook at the case statement03:21
sdakeoh sourced03:21
sdakegot it03:21
sdakeso that model works well for everyting but keystone03:22
sdakeis that what you were pointing out?03:22
SamYapleyea03:22
SamYaplebut actually the code as it is works for keystone too03:23
SamYaplesince we exec /config-crudini.sh, it never gets back to the start.sh script03:23
SamYaplebut it branches03:23
sdakegot it03:23
SamYaplereally this feels like we are setting up crudini for deprecation03:23
SamYaplethats what im worried about03:24
SamYaplei want everyone on the same page03:24
sdakehow does it feel like we set up crudini for deprecation?03:24
SamYapleits now an entirely different script to maintain that branches away from start.sh03:24
SamYapleits not just configuration, its also setting endpoints and what not from containers03:24
SamYaplethe containers are selfconfiguring which is not the case with ansible03:25
SamYaplethat being said, last i checked no one said "self-configuring containers" in the manifesto03:25
sdakethat isn't mandatory that is just how we developed them03:25
sdakewe can just not +2 newcontainers without a  crudini start.sh script03:26
sdakerather config-crudini.sh script03:26
SamYapleso entirely seperate duplicate containers for crudini?03:26
sdakeit would be nice if we only had to support one model03:26
sdakeno same container03:27
SamYaplejust for the last layer i mean03:27
sdakelike you have implemented03:27
SamYapleim confused then03:27
SamYapleright ok. yea we can do it like i have showed03:27
sdakeconfused about which03:28
SamYaplebut that leads to new commiters having to support a script that they likely have never used03:28
*** daneyon has quit IRC03:29
sdakei agree it would be ideal if we could forget the past ever happened03:29
SamYapleno thats not needed03:30
sdakebut that isnt how reality works :)03:30
SamYapleii think it just needs to be clear. are we reruiing all new things to support crudini as well?03:30
SamYaplerequiring*03:30
sdakeyes03:31
SamYaplei thought there was talk of triple-o supporting ansible inside?03:31
SamYapleinstead*03:31
sdakethat wasn't talk coming out of the tripleo community03:31
sdakethat is a discussion that needs to be had03:31
sdakebut tripleo+heat communities have made a huge effort to support our current crudini model03:32
*** dimsum__ has joined #kolla03:58
*** dimsum__ has quit IRC04:03
*** sdake_ has joined #kolla04:56
*** sdake has quit IRC04:59
*** nihilifer has joined #kolla05:18
*** bradjones has quit IRC05:46
*** bradjones has joined #kolla05:49
*** bradjones has joined #kolla05:49
*** fangfenghua has joined #kolla05:53
*** fangfenghua has quit IRC05:58
openstackgerritSam Yaple proposed stackforge/kolla: Add openvswitch container  https://review.openstack.org/19153206:13
*** inc0 has joined #kolla06:15
inc0good morning06:15
dasmo/06:15
dasminc0: where are you?06:15
inc0home office today06:16
dasmmhm.06:16
nihiliferhello06:16
inc0hello nihilifer06:16
dasmo/ nihilifer06:18
SamYaplemorning guys06:21
*** inc0 has quit IRC06:26
*** shardy has joined #kolla06:29
*** inc0 has joined #kolla06:31
*** tobe has quit IRC06:44
*** sdake has joined #kolla06:56
*** sdake has quit IRC06:57
*** sdake has joined #kolla06:57
sdakehey inc006:58
sdakehey dasm06:58
sdakehey nih*06:58
inc0morning sdake06:58
*** sdake_ has quit IRC07:00
dasmgood morning sdake07:05
dasmwhy so early?07:05
sdakelate actually07:05
dasm(or late?)07:05
sdakeits midnight07:05
SamYaplepsshh its 2am here07:05
dasmxD07:05
dasmlast time, when i tried to attend the meeting at midnight, it was pretty hard to write something in plain english.07:06
dasmso this is hardcore hour for me.07:06
dasmkudos SamYaple07:06
dasmkudos sdake07:06
SamYaplei work 3rd shift07:06
SamYapleits not as impressive07:06
sdakei work 1,2,3,4 shift07:06
*** tobe has joined #kolla07:07
*** fangfenghua has joined #kolla07:08
sdakefor the folks in europe, i styaed up late just to link this :07:12
sdakehttps://launchpad.net/kolla/+milestone/liberty-107:12
sdakehey fangfenghua whats up bro07:12
sdakeyou can see there we have done a mountain of work07:12
sdakealmost all green :)07:12
sdakepurple is also a good color :)07:12
*** fangfenghua has quit IRC07:13
*** Haomeng|2 has joined #kolla07:15
SamYaplesdake did you say fang was here?07:16
sdakehe was and then ping timed out07:16
SamYapleah07:17
sdakeanyone not on the drivers team that would like to be07:17
sdakethe drivers team allows you to muck with launchpad07:17
dasmSamYaple: you wrote several weeks ago about 3rd shift. Still the same? or always 3rd?07:17
SamYaplealways 3rd07:17
dasmmhm.07:18
SamYaple(its a 15% bump in pay :)07:18
*** Haomeng has quit IRC07:18
sdakewhere is my 15% bump in pay for working 3rd shift07:18
SamYapleill tell you what sdake, we switch pay and ill pay your 15% out of my new salary07:18
dasmxD07:19
sdakethat sounds --- suboptimal ;)07:19
sdakeare there any blueprints targeted for l2 that folks re working on that will finish by june 25th?07:19
sdakesamyaple is this blueprint started?07:21
sdakehttps://blueprints.launchpad.net/kolla/+spec/openvswitch-container07:21
sdakeor in needs-review?07:21
sdakesamyaple can you create a blueprint for thin-ifying neutron as well?07:22
inc0sdake, I saw review up07:22
sdakeinc0 was someone working on haproxy?07:23
* sdake needs help maintaing th elaunchpad i am not all-knowing :)07:23
SamYaplesdake thinify blueprint exists07:23
inc0I think I saw something ha-proxy like in a tree?07:24
SamYaplesdake: the openvswitch one shows that there are commits in progress07:24
sdakeinc0 that needs a mulligan07:24
sdakesamyaple i set it to needs review07:24
sdakei also sety ou to the owner of https://blueprints.launchpad.net/kolla/+spec/thin-neutron-agents07:24
SamYaplesounds good07:25
sdakefor june 25th?07:25
sdakenot owner, i mean assignee07:25
SamYaplei can do it by the 25th, but it needs reviews that may not be done by the 25th07:25
sdakewfm i'll keep in l1 then07:26
*** prad has quit IRC07:27
inc0hmm, so noone works at haproxy I guess?07:27
SamYaplesomeone was i thought07:27
sdakeduring l2 we will work on ansible-ifying  the codebase07:28
sdakeprobably need a blueprint per service for that07:29
SamYapleoh please dont07:30
SamYapleat least not for the openstack services07:30
sdakeone blueprint then?07:31
sdakei want to be able to distribute the work07:31
inc0I guess writing bps will be more work than actuall work;) once we have first one incubated07:31
sdakeso folks like diga and harm can get involved07:31
SamYapleits just the code is basically the same one you get to the openstack services07:31
inc0maybe just keep Sam (or someone) as assignee and have multiple patches?07:32
sdakesamyaple thta didn't pass my syntax parser07:32
SamYaplethe openstack services require copy/paste ansible code07:32
sdakei'm open to suggestions on how to split the work up ;)07:32
sdakeyou mean copy from yaodu?07:33
sdakewe have about 5 weeks between l1 and l207:33
SamYaplehere is the flow i see: i mplement the first keystone/glance containers other people can take over from there07:33
SamYaplethe ansible stuff can stay one blueprint07:33
sdakedoesn't each container have seperate defaults/etc?07:34
SamYaplethe other things needed for the config things will be the majority of the work07:34
SamYapleugh i dont know. just get the spec approved07:34
sdakethe refactor of start.sh in all the containers can definately be one blueprint07:35
SamYaplethat more work than the ansible stuff in my opinion07:36
sdakerenaming start.sh and putting in a fresh start.sh is more work?07:36
sdakei would think sorting out the defaults would be more work07:37
SamYaplei already did the defaults07:37
SamYaplei did that a couple of days ago07:37
sdakewith config inside the container (crudini) all the work for making a new container is in the defaults07:37
SamYaplei just need the spec to be approve to commit all the defaults07:38
sdakegot it07:38
sdakewell the spec being approved for ha hasnt stopped the ha work ;-)07:42
SamYapleit does change the implementation07:42
*** prad has joined #kolla07:42
SamYaplei need to remov ethe docker compose stuff07:42
sdakeack07:42
sdakewell when I wake FDH in the morning i'll submit a new spec based upon the mountain of feedback in the original :)07:43
*** Haomeng|2 is now known as Haomeng07:43
SamYaplethoughts on getting rid of the env entirely? just using hte bind-copyonce bind-copyalways method?07:44
SamYaplethat would also protect sensitive things like passwords from the env variables07:44
sdakeagree07:45
sdakethat is what i'm going to propose in take 207:45
sdakeor whatever patch set i'm on +1 :)07:45
SamYaplei like it. i think everyone else will too07:46
inc0uhh that's a major change in approach07:46
SamYapleinc0: not major at all07:47
sdakeinc0 what samyaple failed to mention is config inside container (crudini) will also be supported07:47
sdakethere will be 3 modes07:47
SamYapleyea its not major compared to the spec07:47
sdakefeel free to comment on the ansimble-multi spec ;)07:47
inc0I did, I just misunderstood "remove env entirely"07:48
inc0;)07:48
sdakesams suggested modes are for config outside container07:48
SamYapleinc0: sorry, i meant the informal name we gave to the dynamic "env" method07:48
inc0yeah, I agree with that07:48
inc0but I've lost this one: what's exactly bind-copyalways?07:50
inc0it's syncing container07:50
inc0container's /etc/whatever.conf with /opt/kolla/whatever.conf07:50
inc0all the time?07:50
SamYapleinc0: it means the the config will be copied from the outside everytime the container restarts07:50
inc0ahh, ok07:50
SamYaplevs the config compied only the first time the container starts (bind-copy-once)07:51
inc0what's use case for bind-copy-once?07:51
SamYaplebind-copyonce allows for immutable containers07:51
SamYapleit will be default07:51
inc0I'm afraid ops will hate it07:51
inc0redeploy whenever you spot a mistake in conf?07:51
SamYaplethats why bind-copyalways exists ;)07:51
SamYapleops _will_ hate it07:52
SamYaplebut they dont have t ouse it07:52
inc0who will then?:)07:52
SamYapleidk07:52
sdakepureists :)07:52
SamYaple^07:52
digasdake: Hi07:52
digamy local firewall issue is resolved07:53
diga:)07:53
*** athomas has joined #kolla07:53
sdakediga grats!07:54
sdakemorning athmoas07:54
sdakeathomas thatis07:54
athomassdake, Good morning!07:55
sdakeya 1am - time for bed soon :)07:55
sdakeI work 6am-10am and 6pm-midnight07:55
digaI am heading for now ansible multinode07:55
sdaketo fit the dev team on kolla :)07:55
athomassdake, I was wondering whether you were very early or very late ;-)07:56
SamYapleyes07:56
sdakeathomas yes ;)07:56
sdakethis mortal form needs sleep08:02
sdakenight all08:02
nihilifergood night08:02
sdakenight nih*08:04
*** athomas has quit IRC08:05
*** sdake has quit IRC08:06
*** jmccarthy has joined #kolla08:31
*** shardy_ has joined #kolla08:38
*** shardy has quit IRC08:39
*** athomas has joined #kolla08:39
*** shardy_ has quit IRC08:44
*** shardy has joined #kolla08:44
*** athomas has quit IRC08:52
*** gfidente has joined #kolla08:59
*** inc0 has quit IRC09:17
*** inc0_ has joined #kolla09:17
*** athomas has joined #kolla09:25
openstackgerritPaul Bourke proposed stackforge/kolla: Symlink all non Dockerfile resources in images  https://review.openstack.org/19011709:38
openstackgerritPaul Bourke proposed stackforge/kolla: Symlink all non Dockerfile resources in images  https://review.openstack.org/19011709:41
*** mstachow has joined #kolla09:57
mstachowo/09:57
SamYaple\o09:57
*** zhiwei has quit IRC10:01
inc0_\o/10:02
inc0_-o-10:02
inc0_ /o\10:03
inc0_ymca!10:03
*** dimsum__ has joined #kolla10:09
*** vbel has quit IRC10:39
*** vbel has joined #kolla10:40
*** Slower has quit IRC11:04
*** Slower has joined #kolla11:04
inc0_Slower, you awake?11:05
openstackgerritPaul Bourke proposed stackforge/kolla: Add base image for oraclelinux  https://review.openstack.org/19101311:14
openstackgerritPaul Bourke proposed stackforge/kolla: Add base image for oraclelinux  https://review.openstack.org/19101311:16
SamYaplepdb: the double up of cetons-binary-centos-binary-bash comes from not properly setting up the PREFIX override11:22
pdbSamYaple: explain?11:22
SamYaplethe .buildconf file for those resets the PREFIX variable in those bases11:22
*** shardy_ has joined #kolla11:22
*** shardy has quit IRC11:24
*** tobe has quit IRC11:25
*** shardy_ has quit IRC11:28
*** shardy has joined #kolla11:29
*** rhallisey has joined #kolla11:40
*** ChanServ sets mode: +o rhallisey11:40
rhalliseymorning11:52
dasmrhallisey: hello11:53
rhalliseyhey!11:53
*** sdake has joined #kolla11:54
*** sdake has quit IRC11:55
*** sdake has joined #kolla11:55
sdakemorning11:59
*** inc0_ has quit IRC11:59
*** inc0 has joined #kolla11:59
inc0hello rhallisey I hope you follow this review?: nosetests tests/test_engine_service.py:StackServiceTest.test_stack_resource_describe_noncreated_resource12:00
inc0sry, wrong copy:(12:00
inc0https://review.openstack.org/#/c/189157/12:00
sdakemorning rhallisey12:00
sdakecan you finish the job on cinder? :)12:00
inc0sdake, awake already?12:00
sdakeunfortunately12:01
sdakemy kids jacked with my ac12:01
sdakewoke up in hot sweat12:01
sdakeac set to 85f12:01
sdake(I keep it at 68)12:01
inc0<translating 85f to celsious...>12:01
sdakehot12:02
sdake41 i think12:02
inc085 == 2912:02
sdakeya12:02
sdakehot12:02
inc0yup12:02
sdakecompared to 6812:02
inc041 would be like midday in Dubai12:02
sdakewed-fri is issupposed to get to115f12:03
sdakeac needs to be at 68 or the house is warm all day12:03
sdakeeven though I have a 5 ton ac it just can't keep up with the sun12:03
inc0I'll never move to California.12:03
sdakein arizona12:04
sdakeworlds apart from california12:04
sdakein arizona everyone packs heat12:04
sdakein california heat is banned12:04
inc0well, even worse. I don't like hot weather12:05
*** bradjones has quit IRC12:13
sdakemandre still around12:13
sdakejpeleer are you up at this godawful hour12:14
*** bradjones has joined #kolla12:15
*** bradjones has joined #kolla12:15
mstachowhi rhallisey :)12:20
*** diogogmt has quit IRC12:20
sdakeanyone want to check out my SET tube amp?  https://drive.google.com/file/d/0B8q6xDPETSkHWjdoaFFCT3RJMzUtLW1oSWlXY1lOMzVLcmJV/view12:21
sdakedual monoblock hand point to point wiring (no circuit board)12:24
inc0wow, no wonder you have high temp in home;)12:24
sdakehaha :)12:24
inc0I can only wonder how much heat this must produce12:24
sdakethose two big boxes on top are transformers ;)12:25
inc0have you seen new mad max sdake? music from this movie would be a killer on such stuff12:25
sdakeit gets warm int he 6 inches sorrouding the amp12:25
sdakebeyond that can't tell the difference12:25
sdakeand its mostly around the tubes12:26
sdakei told my kids "do you know how long your dad has wanted a tube amp?"12:26
sdake"I dont care"12:26
sdake"since I was 20"12:26
sdake"I dont care"12:26
inc0well, there is one way to make them care12:27
sdakei could have gotsomething a little less beastly and weighty if I didn't go with SET12:27
sdakebut set is the way to roll :)12:28
inc0play this loud when they're around https://www.youtube.com/watch?v=YT516h7QwA412:28
openstackgerritMichal Stachowski proposed stackforge/kolla: Galera container  https://review.openstack.org/18722512:28
inc0and in better quality than this12:28
sdakei have that lp12:28
inc0lp?12:29
*** prad has quit IRC12:30
sdakerecord12:35
sdakethe push pull amps dont weigh a million pounds12:39
sdakeor rather 121 lbs12:39
sdakei wish my kids were in school so I could get some volume rolling12:39
sdakethey complain whenever I go over .1 watts...12:39
SamYaplewow sdake up already12:40
SamYaplehey rhallisey12:40
rhalliseySamYaple, hey12:40
rhalliseymstachow, hi!12:40
sdakei'm sure i'll be back in bed in a few hours samyaple12:40
sdake4 hours of sleep does not make for a good day12:42
*** shardy_ has joined #kolla12:43
*** rhallisey has quit IRC12:44
*** rhallisey has joined #kolla12:46
*** shardy has quit IRC12:47
inc0rhallisey, do you follow ansible multi node spec?12:47
rhalliseyinc0, I haven't looked at it yet, but I know you pointed it to me earlier12:47
rhalliseyI'm going to push a new cinder patch first then I'll come back to it12:48
inc0I saw yours and Ians patchset in heat templates12:48
inc0and well, if we push this idea we're implementing, this will also affect how tripleo will deploy configs12:49
rhalliseyya it's very close.12:49
rhalliseycool12:49
sdakeit has no immediate impact12:49
sdakeinc0 ^^12:49
inc0what I mean is, crudini won't be necessary in tripleo12:49
rhalliseyok12:50
inc0in fact, I think using bind-copyalways will be best for tripleo12:50
*** shardy_ has quit IRC12:50
sdakeinc0 i think so as well12:50
sdakebut again, no immediate impact12:50
inc0and I wonder if we should start to deploy configs this way from the beggining12:50
inc0yeah, sure, but we don't have code in main tripleo code yet12:51
*** shardy has joined #kolla12:51
sdakewell if we dropped config inside container from get go that would be a big imrpovement12:51
sdakeslagle you alive12:51
inc0so we can do it the right way from start, and that's something I'd like to follow up, even help you with12:51
*** dimsum__ has quit IRC12:52
rhalliseyinc0, awesome12:52
inc0we should talk this through with Ian and maybe someone from tripleo?12:53
sdakeinc0 that is what the slagle ping was about ;)12:53
inc0well, I'd have this talk later on anyway since its like 6am on west coast;)12:54
inc0not a good time to have architectural discussions;)12:54
*** dimsum__ has joined #kolla12:54
*** dimsum__ is now known as dims12:55
*** dims has quit IRC12:55
*** dims has joined #kolla12:56
*** jtriley has joined #kolla12:56
sdakejtriley around this fine morning?12:56
*** shardy has quit IRC12:56
sdakei wonder if houses wil ever install dc stragiht in the house13:00
sdakewould make my life alot easier13:00
inc0well, Nikola Tesla would want that13:01
*** shardy has joined #kolla13:01
inc0but on the other hand, imagine how often you've tried to put usb in upside-down13:01
sdakeya that didn't even get done right13:02
sdakeusbc ftw13:02
sdakei have several busted usb connectors on my z820 because of my kids and that problem :(13:03
inc0you'd have much more broken things if it would be the case of 220V 2amp current;)13:03
inc0AC13:04
inc0DC I mean13:04
sdakeit would put the wallwart peope out of business :)13:04
inc0and probably out of existance at some percent13:05
inc0with a fair amount of somke with it13:05
sdaketrying to figure out which wall wart goes with which device is no bueno13:06
inc0btw you all know where did "smoke tests" phrase came out?13:06
*** nihilifer has quit IRC13:06
inc0all I'm saying, AC is more idiot friendly than DC, and world is full of idiots13:08
openstackgerritPaul Bourke proposed stackforge/kolla: Create keystone user in start.sh if it doesn't exist  https://review.openstack.org/19107113:08
*** jtriley has quit IRC13:08
*** dwalsh has joined #kolla13:10
SamYapleinc0: bind-copyalways also has the advantage of being a very traditional approch so lots of people will be immediately familiar with it13:17
SamYaplemodify config in a certain place on teh host and restart the service13:17
SamYaplebam13:17
inc0yeah, that's why I like this approach most really13:17
SamYapleanyones tooling can also lay down the config and work with the ansible stuff side by side if desired13:17
inc0even tho its not very container-pure13:17
SamYapleinc0: thats why the default is bind-copyonce ;)13:17
inc0people will grow to hate us for that;)13:18
*** gfidente is now known as gfidente|afk13:19
inc0we should put a outstanding bug for "my config doesn't work", so every time someone would want to file a bug with this, launchpad will point him to our "its not a bug, its a feature" outstanding bug;)13:20
*** prad has joined #kolla13:23
*** kevsi_ has joined #kolla13:31
*** kevsi has quit IRC13:32
*** pdb is now known as pbourke13:34
*** ccrouch has joined #kolla13:37
openstackgerritRyan Hallisey proposed stackforge/kolla: Cinder container  https://review.openstack.org/17096513:39
mstachowinc0 are you here ?13:39
openstackgerritRyan Hallisey proposed stackforge/kolla: Add cinder into the docker compose setup  https://review.openstack.org/17350713:44
*** absubram has joined #kolla13:56
*** athomas has quit IRC13:59
*** athomas has joined #kolla14:04
inc0mstachow, now I am14:07
*** shardy_ has joined #kolla14:13
*** shardy has quit IRC14:15
*** shardy_ has quit IRC14:19
*** shardy has joined #kolla14:19
*** jtriley has joined #kolla14:20
mstachowsdake what do You mean with INIT_DB in Your review message ? Is It prepare_db function or something ?14:22
sdakenot that init-db but ttthere was another variable like ROW14:23
sdakeiirc14:23
sdakeROW what? :)14:23
sdakea boat :)14:23
sdakeshouldn't that be CLUSTER_DB_ROW?14:23
sdakeinit-db is a global14:23
*** openstackgerrit has quit IRC14:24
mstachowhuh that one. It's just binlog format it's default value for galera here14:24
*** openstackgerrit has joined #kolla14:24
*** diogogmt has joined #kolla14:24
mstachowI can do some extra variable here for binlog, that's not problem for me (hero ;) ).14:26
*** fangfenghua has joined #kolla14:27
mstachowAnother thing is that - galera is a boat, so ROW and galera match :)14:27
*** fangfenghua has quit IRC14:31
*** jtriley has quit IRC14:32
*** jasonsb has quit IRC14:32
sdakerhallisey around?14:35
rhalliseysdake, ya14:35
openstackgerritPaul Bourke proposed stackforge/kolla: Create keystone user in start.sh if it doesn't exist  https://review.openstack.org/19107114:35
sdakere https://review.openstack.org/#/c/187313/1/docker/centos/binary/cinder-data/Dockerfile14:35
sdakethat will create an internal /etc/lvm directory that is persistent14:36
sdakeis that what you want?14:36
sdakeinternal to the container14:36
openstackgerritPaul Bourke proposed stackforge/kolla: Add keystone source image for oraclelinux base  https://review.openstack.org/19183314:36
sdakepbourke ++ :)14:36
pbourkedoing some rebasing, hate dependant patches...!14:37
sdakegit rebase -i ftw ;)14:37
sdakepbourke have a few minutes to chat about your keystone container14:40
pbourkesure14:40
rhalliseysdake, ya I just want the host to be able to do a vgremove14:40
sdakepbourke so what i'd like to see is a base image (such as ol, centos, ubuntu, debian)14:41
sdakeand then on top of that layer a distribution type "such as binary, or source"14:41
sdakefor examplle, I'd like to compose ol + rdo14:41
sdakeor centos+source14:42
sdakewithout having to have source copied to multiple places14:42
sdakedo you think that is possible?14:42
sdakei see what samyaple did, he uses links for rdo in fedora images14:45
*** absubram has quit IRC14:45
pbourkehmm yeah I thought that was just because we didnt have fedora images yet though14:46
pbourkesdake: ill have a look and see14:49
mstachowIs it necessary to check variables twice ?14:49
*** vinkman has joined #kolla14:49
sdakemstachow no14:49
sdakepbourke here is another patch to consider14:49
sdakehttps://review.openstack.org/#/c/184222/14:49
mstachowsdake thanks14:49
sdakepbourke here is the blueprint proposal14:54
sdakehttps://blueprints.launchpad.net/kolla/+spec/refactor-base-image-layout14:54
*** jtriley has joined #kolla14:55
sdakewhich is exactly what I want14:55
sdakeunder "proposed layout"14:55
sdakesamyaple still awake?14:55
*** jpeeler has quit IRC14:55
SamYapleinc0: sdake and @everyone_else, should we vote on the default method then? bind-copyonce vs bind-copyalways?14:56
SamYaplesdake: maybe....14:56
sdakesamyaple your commit referenced above doesn't match the blueprint?14:56
sdakerather review 18422214:56
sdakesamyaple i'll add to the agenda14:56
sdake(the voting)14:56
SamYaplesdake: it was a commit before we required blueprints. im piggybacking onto it14:57
rhalliseysdake, bashate is all set for cinder14:57
inc0SamYaple, I don't particuarally like democracy in terms of architectural decisions;)14:57
SamYaplethe change is within the definition of the blueprint14:57
SamYapleinc0: i personally dont care one way or the other to be honest. i will say i will ony be using bind-copyalways though14:57
inc0same as mee14:58
inc0hence my question..who *really* will use copyonce?14:58
SamYaplein that case inc0, the default should probably be the most used option, but its not a battle i care to fight14:58
SamYapleinc0: as sdake said, "pureists"14:58
inc0I mean, besides purists, let purists cofigure it themselves;)14:58
SamYapleor purists rather14:58
inc0not many purists in ops really;)14:59
sdakei was being cheeky when I said that14:59
SamYaplesdake: we know :D14:59
*** kevsi_ has quit IRC14:59
sdakecopy_once implements principle of least surprise14:59
sdakecopy_once implements one source of truth14:59
sdakecopy_always is the opposite of these two properties :)15:00
SamYapleunless of course someone changes the config outside the cotnaienr and expects it to change inside, when it doesnt that would be suprising15:00
inc0I disagree with that really..."why my configs doesn't work? I've changed, restarted servie...it should work!"15:00
SamYaple^15:00
sdakeok source of truth then ?15:00
pbourkeIm not sure its possible to have one dockerfile per service? that was why I went with the layout proposed on the whiteboard...15:00
inc0external config really will be the soruce of truth15:00
pbourkereason being you can sed in the base combo easily enough but you wont know which packaging mechanism to use15:01
openstackgerritMichal Stachowski proposed stackforge/kolla: Galera container  https://review.openstack.org/18722515:01
SamYaplepbourke: i dont think anyone was proposing one dockerfile per service15:01
SamYaplemandre and I were definetly against it15:02
sdakewhat I want and maybe its not possible15:02
*** jtriley has quit IRC15:02
inc0actually I would go futher and set ln -s between /etc/keystone and /opt/kolla/keystone...but I know that might be one step too far;)15:02
sdakeI want to say "give me distro ol + source"15:02
sdakeand out comes a bunch of images with that setup15:02
SamYapleinc0: ill tell you why that has issues15:02
sdakei want to say "give me distro centos + source"15:02
SamYapleinc0: permissions. that uses the guid and uid of the host15:02
sdakeand out comes a bunch of images with that setup15:02
SamYaplethose dont always align with the container15:02
inc0you already have SamYaple15:02
inc0thats why I'm reluctant myself15:03
SamYapleinc0: sounds like something i would have done :P15:03
SamYaplesdake: thats the plan15:03
pbourkeyou can, but you need dockerfiles for each flavor as far as I can see. basically the way I have it in ol keystone review. SamYaple am I missing something?15:03
inc0akwasne is playing around setup from source15:03
sdakepbourke can you link your review15:03
pbourkehttps://review.openstack.org/#/c/191833/15:04
sdakeso samyaple knows what your tlaking about :)15:04
sdakepbourke I was speaking of this one:15:04
sdakehttps://review.openstack.org/#/c/190117/415:04
sdakesamyaple i further dont want to have a bajiillion source directories of duplicated content if it can be avoided :)15:05
SamYaplesdake: yes  https://review.openstack.org/#/c/190117 and https://review.openstack.org/#/c/184222/ work together and dont conflict15:05
sdakeor rdo directories15:05
*** jtriley has joined #kolla15:06
SamYaplesdake: no duplicated content, just symlinks15:06
sdakei can live with symlinks15:06
sdakeand different dockerfiles15:06
sdakeis that what we get out of these two changes?15:06
SamYaplei would never vote for something that violates DRY15:06
SamYaplesdake: yes15:07
*** bmace has quit IRC15:07
openstackgerritRyan Hallisey proposed stackforge/kolla: Cinder container  https://review.openstack.org/17096515:07
sdakewhat order do they need to go in the repo15:07
SamYaplesdake: either order. the other will have to rebase15:07
SamYaplelets get the oldest merged in first15:07
sdakelol15:07
SamYapleit removes all the buildconf files15:07
inc0rhallisey, long shot here, can we specify that we don't want to install iscsi stuff while deploying container?15:08
SamYaplealso i dont want to rebase, you caught me15:08
sdakepborke you opposed to the pain of rebasing15:08
sdakepbourke15:08
sdakesorry15:08
sdaketypo15:08
rhalliseyinc0, yes, I'm going to follow up with that15:08
SamYapleinc0: it would be installed in the contaienr, but it doesnt have to be used i dont think15:08
SamYaplesay in the case of ceph15:08
pbourkemine is a little fiddily as you have to make sure all symlinks aren't broken. but i have a script so don't mind rebasing15:08
rhalliseyI want to get a 'working cinder' up15:08
*** blahRus has joined #kolla15:08
inc0SamYaple, well, we want to keep containers small and tidy right?;)15:09
pbourkeI just hope we're all clear on the scope and purpose of the refactor-base-image-layout bp?15:09
rhalliseythen scrape out that part because I've been having issues taking it out15:09
sdakepbourke thanks for taking one for the team :)15:09
inc0allright Ryan, let's get this in and file a bug to remove this dependency later?15:09
*** jpeeler has joined #kolla15:10
SamYaplepbourke: mandre and I are on the same page, sdake you seem to be there since we clarified we arent violating DRY?15:10
openstackgerritMerged stackforge/kolla: Separate base images into different folders  https://review.openstack.org/18422215:10
sdakefile a wishlist bug with title "TechDebt: this is the debt"15:10
*** gfidente|afk is now known as gfidente15:10
sdakesamyaple wfm15:10
SamYapleinc0: FAT CONTAINERS WOOOOO15:10
sdakepbourke the scope is to be able to compose any container os with any type of distribution method of openstack15:11
rhalliseyinc0, correct15:11
inc0today its one package, tomorrow another service...its like imigration policy propaganda15:11
pbourkeSamYaple: I think we're all good just I thought sdake may have been thinking we were doing service base images15:11
sdakei had thought that until I scrolled down  :)15:11
rhalliseyI'll file a bug and reference it15:11
pbourkeok15:11
inc0cool15:12
pbourkeIt may be possible to do that for the source installs given they should differ little to none between destros15:12
pbourke*distros15:12
pbourkebut for now I think it's simplest to follow the pattern the binary images are using15:12
sdakethat wfm as long as start.sh isnt' duplicated a million times15:12
sdakeit is presently in teh source base between fedora and centos15:13
SamYaplepbourke: agreed mostly15:13
SamYaplesdake: it is not15:13
pbourkesdake: right but https://review.openstack.org/#/c/190117/4 fixes that15:13
SamYaplesdake: it is not duplicated at all15:13
sdakeperhaps i am mistaken15:13
sdake4 hours of sleep - 8am15:13
SamYapleits all symlinks currently as well15:13
sdakeok cool15:13
inc0btw, crazy thought - do we want tempest to be run on our builds as CI?15:13
SamYaplei wont violate DRY not for all the +2s in the world15:13
sdakeinc0 yes15:14
sdakeinc0 we have gating of the building only atm15:14
sdaketempest gating is blocked on running kolla from a single nic with full neutron connectivity15:14
inc0yeah, I know, hence my question15:14
sdakewhich maybe blockedo n the fact that linuxbridge can't operate single nic15:14
*** jtriley has quit IRC15:15
SamYaplecorrect, SamYaple cant make it work, but samyaple doesnt do linuxbridge well, he is an OVS man15:15
SamYaplepffff SamYaple15:15
sdakesamyaple is working on ovs implementation15:15
sdakeso we can get a single nic implentation working15:15
sdakethen we can gate on ovs + singlenic + tempest ;)15:15
SamYaplecant do ovs without flat neutron containers15:16
SamYapleARRRGGG15:16
sdakea path to unblocking has been devised, matter of execution at this point :)15:16
sdakeoh ya I forgot that one15:16
SamYapledont worry samyaple is working on that too15:16
inc0noob question, what's single nic implementation? I thought with network=host we have exactly same network namespace inside containers?15:16
sdakeinc0 neutron requires two nics to operate15:16
SamYapleinc0: neutron needs a nic to route out on15:16
SamYaplesdake: wrong!15:16
SamYapleyou just cant access the box if you use a single nic with linuxbridge ;)15:17
sdakeright15:17
SamYapletechnically the floating ips should work15:17
sdakebut it busts host networking15:17
SamYapleunless its OVS. i got that stuff down15:17
inc0but why are we limited to 1 nic?15:17
SamYaplebut ovs grabs the nic at layer215:17
sdakeinc0 the gate has only 1 routable nic I guess15:18
sdakethe other nic is private15:18
*** bmace has joined #kolla15:18
inc0ah, and if we put linuxbridge over it, we lose routable nic right?15:18
*** jtriley has joined #kolla15:18
sdakeyup15:18
sdakejtriley morning around?15:18
inc0and we put linuxbridge due to contenerization15:18
sdakelinuxbridge is what danyeon implemented first15:19
inc0can't we just add virtual nic and brigdge together with container stuff?15:19
sdakedon't know the details of why but i've heard alot of complaints from the field that ovs is unreliable15:19
SamYapleinc0: yea but it breaks floating ips15:19
sdakeinc0 if you can figure it out i'll mail you 1$ USD15:19
sdakewe have had 4+ people spend 5+ days each on it15:20
sdakeal failed :)15:20
inc0haha15:20
SamYaplesdake: lots of ovs is old, that is bad. OVS >2.1 should never be used, OVS 2.3.1 is prefered. Ubuntu packages 2.0.2 and otehrs also package old ones15:20
openstackgerritRyan Hallisey proposed stackforge/kolla: Cinder container  https://review.openstack.org/17096515:20
SamYaple<2.115:20
inc0sounds like good, creative way to lose some time;)15:20
SamYapleinc0: youre welcome too it. i can get it to work upto having floating ips respond. thats where it breaks15:21
rhalliseyI think cinder is all set now15:21
SamYapleie floating ips wont respond locally or remotely15:21
inc0yeah, 2.1 was the version where they had this big perfomance leap right?15:21
SamYapleinc0: and allowed arp responder/l2pop stuff15:21
SamYaple2.3.1 is the latest LTS15:21
inc0SamYaple, do you have any code yet you'd like to share?15:22
inc0so we can coop there?15:22
SamYaple2.4 will support conntracking and all the other firewall bits so we can remove the iptables dependancy! all the secgroup rules can be implemented as OVS flows15:22
SamYapleinc0: it was all by hand15:22
sdakeinc0 hold i can share my go at it :)15:22
inc0I'd gladly take a look to get my bearings, I know thing or two about networking, but containers are new to me15:23
SamYapleinc0: the container doesnt really come into play15:23
*** sdake_ has joined #kolla15:23
inc0well, that's why I'm kinda puzzled, how does neutron CI work then?15:24
sdake_inc0 here is a couple tries:15:24
sdake_http://ur1.ca/mu10m -> http://paste.fedoraproject.org/232204/1434381815:24
sdake_http://ur1.ca/mu10r -> http://paste.fedoraproject.org/232205/3818691415:24
SamYapleinc0: pretty sure they have multinic, dont know why we dont. i _know_ they have multihost, but they dont use it15:24
sdake_let me ask on infra15:25
inc0maybe just ask them?;)15:25
*** sdake has quit IRC15:27
mstachowSamYaple, about mariadb 5.5 vs 1015:27
mstachowThis dumps was totaly random15:27
mstachowSometimes on one node I was able to join to cluster sometimes I wasn't. Logs wasn't clear for me, but I will investigate once again whole situation15:28
SamYaplemstachow: going to need more than that. everyone i know is moving to mariadb10, and ive been personally using it for a year just fine15:29
SamYaplei can help you if you tell me how to replicate15:29
SamYapleor want to share the logs if its interesting15:29
mstachowno problem, if I will had any problems again I will share logs with You15:29
SamYaplei saw the cache cause this once, galera does not work with query cache15:29
mstachowBut cache and containers ? I am removing also storage for db each time when I'm recreating galera15:31
slaglesdake_: yea, i'm alive :) was all tied up last week15:31
slaglesdake_: i did see the spec, but haven't had a chance to read it thoroughly15:32
SamYaplemstachow: its a my.cnf option15:32
sdake_slagle there is a long thread in the ansimble-multi spec15:32
sdake_slagle let me summarize it for you since i know your busy :)15:32
sdake_we are talking about having three modes of operation15:32
SamYaplepbourke: https://review.openstack.org/#/c/191833/ if you are around lets discuss that comment before i go to sleep15:32
inc0sdake_, what if we use 127.0.0.1 as "private network"?15:32
inc0can we run tests on same machine as services?15:32
sdake_inc0 let me get back to you on that point :)15:32
pbourkeSamYaple: sure - just reading your comment15:33
*** pradk has joined #kolla15:33
sdake_slagle mode 1 - is the same mode that currently exists today - config inside container using crud ini15:33
pbourkeSamYaple: all that makes sense15:33
sdake_slagle mode 2 - is a new mode that is config OUTSIDE the container using a config managmenet tool, or for the foolhardy vi :)15:33
sdake_slagle mode 3 is a new mode that is config OUTSIDE the container using a config managmenet tool, or for the foolhardy vi15:34
SamYaplepbourke: discussion over. yay :D15:34
SamYapleim goign to sleep15:34
pbourke:D15:34
sdake_slagle mode 2 and mode 3 have slightly different semantics, config files are copied from outside the container to inside either one time or mulitiple times depending on mode15:34
openstackgerritPaul Bourke proposed stackforge/kolla: Symlink all non Dockerfile resources in images  https://review.openstack.org/19011715:34
SamYaplenight all15:34
sdake_sweet dreams samyaple ;)15:34
inc0cya SamYaple15:34
sdake_slagle the key thing to think about is config inside container vs config outside container15:35
slaglesdake_: ok, not sure i grok the difference between 2 and 315:35
slagleyea, ok15:35
sdake_config inside container = docker-compose15:35
sdake_config outside container = puppet or chef or ansible or vi15:35
mstachowgood night SamYaple :)15:35
sdake_slagle difference between 2 and 3 is one maintains immutability one does not15:36
slaglesdake_: which (or all of?) the 3 is kolla planning to implement?15:36
sdake_I dont want to disrupt tripleo's efforts15:36
sdake_we already have #1 implemented15:36
sdake_we are implementing 2 and 3 anew15:36
sdake_they are very similair - its like a 3 line difference between the two models15:36
sdake_(guessing on 3 lines no actual code:)15:37
slagleok. i likely need to go read the spec to understand it better15:37
sdake_if you want to understand it you will have to read and analyze each comment in detail15:37
sdake_the big difference between mode 1, and mode 2/3 is mode 2/3 allow full customization15:37
sdake_mode 1 is totally opinionated15:38
sdake_mode 1 is what rhallisey has up for review in tripleo review queue15:38
inc0point is, how do you like idea about tripleo generating config as tripleo does now - as a single file, then just put it in specified folder on host and deploy an container?15:38
slagleright, i see how it works now with docker compose15:38
inc0instead of provide env variables and let crudini build config inside container as it is done now?15:38
slaglehonestly, i'm not sure mode 1 is the right long term thing for tripleo15:38
*** absubram has joined #kolla15:39
sdake_we are maintaining mode 1 for tripleo15:39
sdake_we prefer to go straight to mode 2/3 permanelty and drop mode 115:39
sdake_but we dont leave downstreams behind15:39
sdake_but if tripleo doesn't even want mode 1 long term, then why ahve it :)15:40
inc0do we have any downstreams *besides* tripleo?15:40
sdake_inc0 nada15:40
inc0then just drop mode 1, refactor existing tripleo patches to use mode 3 and be done with it?15:41
slaglewell really i'm just saying, tripleo doesn't have any landed code yet15:41
slaglei dont know that we've even validated the full architecture based on mode 115:41
sdake_slagle its important to know how we should proceed because it affects our strategy going forward15:42
slaglewe are doing a lot of work to deploy via puppet...maybe it makes sense for us to consider mode 2/315:42
sdake_strategy as defined by committing resources to achieve ad esired outcome :)15:42
slaglesdake_: sure, in all honestly, i'm not sure i'll be able to give you a concrete answer15:42
inc0Slower, rhallisey probably you'd like to look at this conversation as well15:43
slaglethe fact is, (aiui) no one has actually deployed a producton cloud with kolla mode 115:43
sdake_actually not accurate15:43
sdake_but they are pretty bleeding edge :)15:43
*** pradk has quit IRC15:43
slagleis there anything documented about how to do it?15:43
sdake_heavy wizardy15:44
rhalliseymode 3 would allow puppet -> kolla15:44
rhalliseywhich would be nice15:44
slaglei'm just curious if mode 1 is going to break down if we start to do HA/multinode/orchestrating updates/ etc15:44
slagleya know, all the long tail stuff that is actualy the 90% of the work to deploy the cloud :)15:44
inc0slagle, that will require some handy env tinkering15:44
slagleand if mode 1 isn't the right model for that...i don't know that tripleo would use it15:44
*** jtriley has quit IRC15:45
sdake_slagle it would be fantastic if you could say that in the review :)15:45
slaglesure, i need to review it15:45
sdake_rhallisey if we drop mode 1 - is that going to cause a bunch of disruption for your work?15:45
rhalliseysdake_, no, because all we do it pull containers and run them15:46
sdake_slagle when can you get er done, people chomping at bit to get started15:46
rhalliseywe don't configure15:46
sdake_where does yoru openstack.env file come from?15:46
slaglesdake_: i'll do it today15:47
sdake_slagle sounds great15:47
slagleif not in detail, i'll at least add my thoughts15:47
sdake_slagle i am working on a spec update - but would appreciate the comments in the previous review15:47
inc0slagle, don't worry about ansible-technicalities, we are in agreement that ansible will not be hard requirement there15:47
sdake_th eone with 30+ comments15:47
rhalliseysdake_, oh I see that will get yanked.. that's ok if we adjust15:48
rhalliseythe change won't be hard15:48
sdake_rhallisey your statement was unclear could you restate15:48
inc0rhallisey, I've said that before - if you need any help, just honk15:48
rhalliseysdake_, if mode 1 is removed we will have to change it a bit, but it won't be too bad15:49
rhalliseybecause what we're doing is pretty simple, we can instead give a template openstack.env or something15:49
sdake_slagle this model should work well with puppet buti am no puppet expert15:50
rhalliseysdake_, I agree, if puppet can gen the config it removes that step where we add openstack.env15:50
inc0sdake_, puppet, at the end, just generates files, that will remain the same15:50
slagleyea, i'm just curious, how is puppet, doing something like configuring nova.conf15:51
sdake_puppet also starts services and other things15:51
inc0only difference might be it will put it to /opt/kolla instead of /etc15:51
slagledoes a puppet apply run in the container?15:51
slagleor is puppet itself aware of the containers?15:51
sdake_it runs on the host15:51
inc0slagle, it doesn't have to know about containers15:51
sdake_i wouldn't think puppet would be aware of th econtainers15:51
sdake_but it could be if you wanted it to do so and had people to write the code :)15:52
inc0I mean, we can then run container with puppet15:52
inc0like ensure => run...or whatever you deploy container with puppet15:52
slaglei'm not asking for that, i was just asking how it works15:52
inc0but that's completely decoupled from config15:53
sdake_the host ends up with keystone.conf for example in /opt/kolla/openstack/keystone/keystone.conf15:53
sdake_the container reads /opt/kolla/openstack/keystone/keystone.conf on startup15:53
sdake_it either copies it once on continer create15:53
sdake_or copies it every container start15:53
slagleok. are the puppet openstack modules used to configure that keystone.conf?15:53
sdake_(modes 2 or 3)15:53
inc0slagle, yes15:54
inc0just running keystone itself will differ15:54
sdake_would require r&d I think15:54
sdake_i am not a uppet expert - get one in channel and we can talk about it :) dprince perhaps :)15:54
slaglesdake_: ok, so when you say kolla is implemented mode 2/3...what does that mean?15:55
slagleis someone actually doing the puppet work? or...15:55
sdake_nobody is doing puppet work for kolla15:55
slaglethere's some interface there that shows how it might work15:55
sdake_so puppet has a keystone module i assume, and that configures keystone.conf15:55
sdake_atm in tripleo that occurs on the host correct?15:56
slagleyes15:56
sdake_that would work the same way although the keystone.conf may be in a different place then puppet expects15:56
sdake_this is the r&d I was tlaking about15:56
sdake_the running of the container could be done by puppet as well, this requires r&d15:56
inc0slagle, we will just mount folder from host:/opt/kolla/openstack/keystone/ from host to container, and then cp /opt/kolla/openstack/keystone/keystone.conf /etc/keystone15:57
slaglei see15:57
inc0and whatever was in this keystone.conf (even Lorem ipsum) will land in /etc15:57
sdake_hopefully it doesn't convert my ipman quote into "Horizontal." :)15:57
slagleinc0: i'm not sure the puppet openstack modules are setup with this flexibility.15:57
inc0slagle, they have to generate config somehow right?15:58
slaglesure, but they expect to do things like install packages, restart services, sync db's, etc15:58
slagleespecially if you were to make a change to keystone.conf15:58
sdake_right, the config step isn't broken out into one concrete piece15:58
inc0mhm, then we would have to decouple all additional logic15:58
sdake_we being the puppet community in this case :)15:59
slagleright, so i guess that's what i'm trying to understand15:59
slaglethis was a huge stumbling block for tripleo IMO16:00
sdake_which part slagle16:00
slaglejust saying we integrated with puppet/chef/whatever, didn't actually turn out to be true16:00
slagleso i'm just trying to understand what mode 2/3 is actually going to enalbe someone to do16:00
sdake_slagle can you ping dprince to join the convo16:00
sdake_mode 2/3 do no configuration16:01
sdake_they just copy the config file on the host into the container and start the service16:01
sdake_the service is installed in the docker fiels during the container build operation16:01
sdake_http://blog.circleci.com/its-the-future/16:02
*** jtriley has joined #kolla16:02
slaglelol16:02
sdake_worth a read if you haven't :)16:02
sdake_so basically puppet/chef/whatever do a whole bunch of things all in order16:03
sdake_like install packages, sync db, etc16:04
sdake_install packages -> dockerfile16:04
slaglebut you wouldn't want to do any of that on the host16:04
sdake_run service -> start.sh16:04
sdake_agree16:04
sdake_but you solved  this problem for tripleo, how?16:04
*** bmace has quit IRC16:05
slaglewhat problem? we run puppet apply on the host itself16:05
slagleif it's a compute node, we appy the puppet manifests for nova compute, etc16:05
sdake_you said it was a stumbling block?16:05
sdake_we dont want to create more hurdles16:05
slaglethe stumbling block was saying "this works with puppet/ansible/chef" without anyone actually validating that16:06
*** bmace has joined #kolla16:06
slagleas it turns out, the interface was actually quite poor to non-existant16:06
slaglewe didn't start to refine the interface until we started to think about and actually making it work with a 2nd implementation16:06
slaglethat's really what proved it out16:07
slagleso (i don't mean to nitpick)...but if kolla says mode 2/3 is going to enable all these things...i'm just curious what that actually looks like16:07
slagleis puppet enabled if i go off and write a bunch of custom puppet?16:07
sdake_it is going to enable them for ansible16:07
slagleor do the puppet openstack modules just plug in16:07
sdake_i suspect that would be an r&d exercise upstream slagle16:08
slagleyea, i think i've got that now :)16:08
sdake_the ansible part, we are handling16:09
sdake_brb need to switch networks16:10
*** sdake has joined #kolla16:11
*** mstachow has quit IRC16:13
sdakeslagle if mode 1 is not viable long term for tripleo16:13
sdakeand mode 2&3 require puppet r&d which our community is unequiped to contirbute to16:13
sdakewhat do you suggest ;)16:13
slagle1) tripleo continues to use mode 1 so we can see how viable long term it might be16:14
slagle2) we involve the tripleo/puppet folks to see what the r&d entails16:14
*** sdake_ has quit IRC16:14
slagleat that point, we could probably make a decision about going with mode 2/3 long term16:14
sdakewfm16:15
slaglelike i said, not a concrete answer :)16:15
sdakerelated question about big tent16:15
slaglei do think though that the puppet folks are interested in container integration and making the modules work with containers16:15
sdakehow does tripleo intend to handle big tent new components16:15
slaglewe'll deploy them if theres interest16:16
slaglebut tripleo isn't just adding everything big tent just for fun16:16
openstackgerritPaul Bourke proposed stackforge/kolla: Symlink all non Dockerfile resources in images  https://review.openstack.org/19011716:17
openstackgerritPaul Bourke proposed stackforge/kolla: Add base image for oraclelinux  https://review.openstack.org/19101316:17
sdakekolla is adding all openstack namespaced projects just for kicks16:17
sdakeatleast long term16:17
sdakei could put a requirement in there "dont do in-container config for new containers"16:17
sdakethis would drastically reduce the amount of work we have to do16:17
sdakeand we could address the ones you have interest in on an as needed basis16:18
slaglei think that would be fine16:18
slaglei'm not going to ask people to do more work just for tripleo's sake, when i don't even know if it would ever be used16:18
sdakeslagle can we somehow get the puppet cats engaged in this conversation16:19
sdakesooner rather then later16:19
sdakeso we can get to a decision point about mode 1 deprecation (if there is to be one) sooner rather then later16:19
slaglesure, i can ask16:19
slaglei can't speak for when anyoen would actually do it16:20
sdakeyar16:20
sdakewho can speak for anyone else?16:20
sdakeIcan't even speak for my 10 and 12 year olds :)16:20
inc0let's just try to scope out how much work that would be16:20
sdakeyar if we can get a viable technical plan in place that would be helpful16:21
sdakeslagle thanks for your time this conversation has been emmensly helpful16:23
slagleyep, thanks as well. i'll read over the spec and ping folks on it today16:24
sdakethe second version will look drastically different from the first16:24
sdakebe prepared :)16:24
sdakeit is based upon a week of open debate on irc + the review16:24
*** jtriley has quit IRC16:25
*** daneyon has joined #kolla16:26
*** dasm is now known as dasm|afk16:26
*** daneyon has quit IRC16:26
*** daneyon has joined #kolla16:27
inc0uhh, fruitful discussions16:27
inc0let's start another one! 127.0.0.0/24 as private network for endpoints?:)16:28
inc0for gates?16:28
sdakebrain tired16:28
sdakeneed to refresh :)16:28
inc0haha16:28
inc0I'll think also how to build gates based on option 2/316:29
inc0I mean we probably will just end up providing hardcoded config on infra16:29
sdakeif you use the flat interface for the routed internet network, you lose connectivity to the vm16:30
inc0why do we need connectivity at the end?16:31
inc0once we build all?16:31
inc0uh, nvm, giving any test feedback is useful16:32
sdakeya we need to know the tests success or failure :)16:32
inc0we need to dig out how neutron does that16:33
sdakethey do it via single nic via ovs via the unblocked workflow i've sorted out :)16:34
sdakeit cannot be done apparently with linuxbridge16:34
*** jtriley has joined #kolla16:36
*** fangfenghua has joined #kolla16:37
*** fangfenghua has quit IRC16:42
jpeelersdake: if this patch gets merged one day (https://github.com/docker/docker/pull/13525) would you be opposed to running systemd in containers?16:47
sdakein kolla containers?16:52
sdakepretty much, kolla uses thin containers16:52
sdakei.e. one process per container16:53
jpeeleryes, i know how it works now16:54
openstackgerritPaul Bourke proposed stackforge/kolla: Symlink all non Dockerfile resources in images  https://review.openstack.org/19011716:54
sdakeoh for some reason I saw jpeeler and read "slagle"16:54
sdakebrain malfunction :)16:54
sdakejpeeler why doth thou ask :)16:55
jpeeleri assume this post is still relevant: https://blog.phusion.nl/2015/01/20/docker-and-the-pid-1-zombie-reaping-problem/16:55
jpeelerthat's why i ask16:55
sdakethere are might lighter weight solutions to solve that problem then starting 20+ processes16:56
sdakeeven supervisord is a superior solution16:56
jpeelersure, it doesn't have to be systemd necessarily16:56
pbourkejpeeler: the docker baseimage guys dont get containers imo...16:57
sdakeif you read that blog they have a nice tidy 200 line python program that solves it16:57
sdakei am open to solving the zombie reaping problem but not really with systemd16:58
sdakereason being systemd will start all the stuff in the /lib/systemd directories on start16:58
sdakewhich is huge bloat for a container16:58
sdakeif systemd didn't do that,i'd probably have a different viewpoint16:58
jpeelerpbourke: maybe i don't either! what makes you say that?16:59
pbourkejpeeler: i haven't read the blog you posted (yet) it maybe makes some good points. but I find https://phusion.github.io/baseimage-docker/ sort of annoying17:01
pbourkeanyway have to head home, conversation for another day :) sdake, I rebased the symlink so it should be hopefully good to go17:03
pbourke*symlink patch17:05
*** dwalsh has quit IRC17:07
openstackgerritHarm Weites proposed stackforge/kolla: Check if tools/genenv binary dependencies are met.  https://review.openstack.org/19158017:10
harmwmandre: you were absolutely right!17:10
harmwI took the wrong file :)17:10
rhalliseyharmw, updated the cinder container17:16
*** nihilifer has joined #kolla17:17
harmwyea, just looking through my e-mail :)17:17
rhalliseyno prob, thanks man17:18
harmwah yes, the massive cinder review17:21
harmwI remember that17:21
harmw:p17:21
*** jasonsb has joined #kolla17:24
*** dwalsh has joined #kolla17:24
jpeelersdake: I just updated https://review.openstack.org/#/c/191133/, but I can't decide if that's right or not17:27
*** dims has quit IRC17:27
jpeelerafter reading: http://docs.openstack.org/infra/system-config/zuul.html17:27
*** dims has joined #kolla17:28
jpeelerthe question is, do we think people will be uploading broken images? if not, then the review is good. if we do, then maybe putting the job in the check section is better than waiting on core reviews to get verification17:28
harmwnever trust input or users!17:28
jpeelerwell ultimately we don't! it's just a matter of when the feedback comes. one is faster merge time if we assume people are mostly good17:29
*** jasonsb has quit IRC17:29
jpeelerin fact, it's theoretically at least double the speed...17:30
harmwrhallisey: it scares me I discovered new 'issues' :p17:38
harmw*again17:38
harmwanyway, nothing big but one or two functional things17:39
rhalliseyharmw, I'll look17:39
rhalliseyharmw, :) I didn't know how copy works never used it17:40
harmw:)17:41
*** fangfenghua has joined #kolla17:43
*** shardy_ has joined #kolla17:44
*** mstachow has joined #kolla17:45
*** shardy has quit IRC17:45
*** fangfenghua has quit IRC17:48
*** shardy_ has quit IRC17:50
*** shardy has joined #kolla17:51
*** jasonsb has joined #kolla17:55
sdakejpeeler I saw but I don't know either17:58
sdakelook at how you reverted it?17:58
sdakeoh it should be a check as well17:58
*** jasonsb has quit IRC18:00
jpeeler... so you mean you expect people to submit broken images?18:22
sdakeit could happen18:25
sdakeanything that has a gate should have a check imo :)18:25
jpeelerthat sort of defeats the point of the check pipeline18:27
jpeelermaybe i'm misunderstanding, no idea why you would run the exact same tests. what am i missing?18:28
*** sdake_ has joined #kolla18:29
sdake_jpeeler the check pipeline tells the core reviewers a patch is ready for review18:30
sdake_jpeeler the gate pipeline tells gerrit the patch si ready for merge18:30
sdake_that is the significant difference18:31
sdake_make sense?18:31
jpeeleryes18:33
jpeeleri guess it's possible that code changes underneath that wouldn't require a rebase to the review gives value to double testing18:33
jpeelerthat's something i just realized (if it's correct)18:33
sdake_that is a technical reason but you asked why have two :)18:33
*** sdake has quit IRC18:34
jpeeleryes, i couldn't figure out the value. but now i can! i have to understand why i do things most of the time. not trying to be a pain!18:34
sdake_all good18:35
sdake_i'm here to please :)18:35
*** jasonsb has joined #kolla18:37
*** mstachow has quit IRC18:41
openstackgerritMichal Rostecki proposed stackforge/kolla: [WIP] Add designate-sink service  https://review.openstack.org/18939318:47
*** fangfenghua has joined #kolla18:50
*** mugsie has joined #kolla18:52
mugsienihilifer: hey - you around?18:53
harmwoh noes, designate ppl here!18:53
mugsieharmw: we are like a stray dog18:54
harmw:p18:54
mugsiejust follow you around if you talk to us :)18:54
*** fangfenghua has quit IRC18:54
harmwhaha18:54
*** dwalsh has quit IRC18:56
mugsiemaybe you can anser one thing harmw - https://review.openstack.org/#/c/189393 - none of the domain names have trailing dots (eg. "example.com.")18:58
mugsiehow are the designate client calls succeeding? is there a wrapper somewhere I cant see thats is appending it?18:58
harmwtrailing dots are most probably not there18:58
harmwgood one18:58
harmwthis is the only designate container where it actually creates domains and such, and I missed it19:00
harmwdamn19:00
mugsie:)19:00
harmwI still need to go through that code again, so I'll add a comment :)19:01
harmw*over19:01
nihilifermugsie: yes19:01
mugsieit was disccusion on the designate sink container patch ^19:01
mugsiewas confusing myself19:01
*** absubram has quit IRC19:01
mugsiebut it seems you *are* missing the trailing dots19:02
nihiliferyes, nice catch, will add the dots19:02
mugsieto be fair, it is the one thing that gets everyone when they use designate19:03
harmwlol, yes sire to that ! :p19:03
mugsieeven me sometimes, even though I wrote the majority of the v2 API, and called it 100s of times a day19:04
nihiliferI was adding dots when I runned this container with my environment, that's the worst thing :P19:06
*** dwalsh has joined #kolla19:09
mugsie:D19:09
openstackgerritMichal Rostecki proposed stackforge/kolla: [WIP] Add designate-sink service  https://review.openstack.org/18939319:11
openstackgerritMichal Rostecki proposed stackforge/kolla: [WIP] Add designate-sink service  https://review.openstack.org/18939319:19
openstackgerritRyan Hallisey proposed stackforge/kolla: Cinder container  https://review.openstack.org/17096519:23
inc0good night everyone19:24
*** inc0 has quit IRC19:24
*** mstachow has joined #kolla19:28
*** mstachow has quit IRC19:47
*** fangfenghua has joined #kolla19:54
*** sdake has joined #kolla19:54
*** gfidente has quit IRC19:58
*** sdake_ has quit IRC19:58
*** daneyon_ has joined #kolla19:59
*** fangfenghua has quit IRC19:59
*** sdake_ has joined #kolla20:01
*** daneyon has quit IRC20:02
*** sdake has quit IRC20:05
*** nihilifer has quit IRC20:38
openstackgerritSteven Dake proposed stackforge/kolla: Ansible multi-node specification  https://review.openstack.org/18915720:43
*** sdake has joined #kolla20:44
*** sdake_ has quit IRC20:48
*** mstachow has joined #kolla20:52
*** dims_ has joined #kolla20:56
jpeelerit appears ironic also requires openvswitch20:57
sdakesamyaple is on it20:58
sdakefirst we neeed to de-fatify neutron20:58
sdakewhich he is working on atm20:58
jpeeleryeah, maybe it'll be done by the time i'm ready to test20:59
*** fangfenghua has joined #kolla20:59
sdakei think he has openvswitch working20:59
sdakebut it conflicts with the neutron agent container20:59
sdakehow do you know ironic requires ovs?21:00
sdakeis that documented somewhere?21:00
jpeelerhttp://docs.openstack.org/developer/ironic/deploy/install-guide.html#configure-neutron-to-communicate-with-the-bare-metal-server21:00
*** dims has quit IRC21:00
*** dims_ has quit IRC21:00
sdakeonly documented forovs21:01
jpeelerneutron is the last thing to configure before trying to figure out the PXE stuff21:01
sdakejpeeler take a stab at the ansible-multi spec :)21:02
openstackgerritimain proposed stackforge/kolla: Set up glance to use a data container.  https://review.openstack.org/19197521:03
daneyon_jpeller do u have a sec?21:04
*** fangfenghua has quit IRC21:04
jpeelerdaneyon_: what's up21:05
daneyon_Re: https://review.openstack.org/#/c/189974/321:05
daneyon_Looks like the patch works when using the binary images but not the centos/rdo images21:05
daneyon_would you like me to submit an updated PS with the centos/rdo fixes?21:06
jpeelersure, but i'm confused. you're talking about in the docker directory with the rdo symlinking to binary? which binary images are you referring to?21:07
*** rhallisey has quit IRC21:07
daneyon_I'm probably confused with the new DIR structure of the project. What is the diff between docker/centos/binary and docker/centos/rdo21:09
jpeeleri didn't think there was any currently21:10
daneyon_So right now rdo images are using images from binary21:10
*** dwalsh has quit IRC21:10
daneyon_why is rdo a sym to binary?21:10
daneyon_should i use docker/centos/binary for building images and not docker/centos/rdo?21:11
jpeeleri really thought currently they were supposed to function exactly the same. i can't remember which i used21:11
* jpeeler looks for a commit21:12
jpeelerhttps://github.com/stackforge/kolla/commit/13062e23e86b5d86020949754540460cb5b0b4fe21:14
jpeelerthe idea i thought was to eventually have source installs21:14
jpeelerwhat breaks when using the rdo directory?21:14
daneyon_ok, i think i get it now. let me try your patch again, recheck and i'll let you know if i have any problems. thx21:16
*** jasonsb has quit IRC21:17
sdakejpeeler the plan is to be abel to select base container os and select source/binary and have the right thing happen21:26
sdakejpeeler rhallisey can you guys take a look at https://review.openstack.org/#/c/189157/ with tripleo in mind21:33
sdakejtriley if you want to take a look with puppet in mind that would be helpful too :)21:33
jtrileysdake: yep i'll give this a look when I get a chance21:34
*** dims has joined #kolla21:35
*** jtriley_ has joined #kolla21:38
*** rhallisey has joined #kolla21:38
sdakerhallisey can you take a look at https://review.openstack.org/#/c/189157/ with an eye towards triipleo when you hve a moment21:39
sdakeit is a bit different then take 1 :)21:39
rhalliseyI'll take a look tmr21:40
rhalliseyor later.. I gota head home21:40
sdakewfm21:40
sdakethanks21:40
sdakegood work on cinde as well :)21:40
sdakenow just need to make it thin :)21:40
rhalliseythanks21:40
*** rhallisey has quit IRC21:41
*** jtriley has quit IRC21:41
*** jasonsb has joined #kolla21:42
*** jtriley_ has quit IRC21:42
*** jasonsb has quit IRC21:43
*** rhallisey has joined #kolla21:46
openstackgerritRyan Hallisey proposed stackforge/kolla: Cinder container  https://review.openstack.org/17096521:57
*** fangfenghua has joined #kolla22:05
*** fangfenghua has quit IRC22:09
jpeelersdake: i'll look at it later on tonight22:12
sdakejpeeler wfm22:17
*** dims has quit IRC22:34
*** jtriley has joined #kolla22:35
*** pbourke has quit IRC22:43
*** pbourke has joined #kolla22:44
*** Haomeng|2 has joined #kolla22:50
*** rhallisey has quit IRC22:53
*** Haomeng has quit IRC22:53
*** diogogmt has quit IRC22:53
sdakepbourke around?22:57
*** dims has joined #kolla23:03
*** dims has quit IRC23:03
*** dims has joined #kolla23:04
openstackgerritDaneyon Hansen proposed stackforge/kolla: Fixes MariaDB to Support Heat  https://review.openstack.org/19201123:10
sdakeloth you about23:19
lothWhats up?23:20
sdakemind taking a look at https://review.openstack.org/#/c/189157/ with a vote :)23:22
*** blahRus has quit IRC23:29
*** jtriley has quit IRC23:29
*** diogogmt has joined #kolla23:30
*** jtriley has joined #kolla23:33
*** daneyon_ has quit IRC23:43

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