Monday, 2016-11-28

« Sunday, 2016-11-27 Index Tuesday, 2016-11-29 »
*** hfu_ has joined #openstack-kolla00:00
*** hfu has quit IRC00:01
*** hfu_ has quit IRC00:13
*** sdake has joined #openstack-kolla00:15
*** sdake_ has joined #openstack-kolla00:18
*** sdake has quit IRC00:19
*** jtriley has joined #openstack-kolla00:35
*** jtriley has quit IRC00:41
*** tovin07 has joined #openstack-kolla00:48
*** yingjun has joined #openstack-kolla00:49
*** tovin07_ has joined #openstack-kolla00:49
*** zhubingbing has joined #openstack-kolla00:59
sdake_zhubingbing sup dude01:01
zhubingbinghi01:01
zhubingbinggood morning01:01
sdake_say, are you interested in tackling monasca01:01
zhubingbingyeah01:02
sdake_its really hard i think because htey have a bajillion repos01:02
sdake_i'd like to see it in action01:02
zhubingbingnp01:03
zhubingbingI will integrate it into kolla01:03
zhubingbinghttps://wiki.openstack.org/wiki/Monasca01:03
*** tovin07 has quit IRC01:04
zhubingbingthis week，i will finish it01:04
sdake_well i wouldn't put such a tight timer on it01:06
sdake_it will probably take awhile01:06
sdake_it has a slew of dependencies01:06
zhubingbingunderstand01:06
zhubingbingI'll deal with it  as the top priority.01:07
sdake_zhubingbing also do you have any interest in learning about kubernetes arches?01:07
sdake_top priority is actually probably barbican - to enable magnum security01:07
zhubingbingK8s I was very familiar with it.01:07
zhubingbing；）01:07
sdake_ya i was too01:07
sdake_its changed quite a bit01:07
sdake_reason i ask is we have the kolla-kubernetes work as well01:08
sdake_atm we aer cranking out helm charts01:08
zhubingbingI've done k8s related work 6month ago01:08
sdake_which is a whole new brave world to learn01:08
*** goldyfruit has joined #openstack-kolla01:10
*** portdirect is now known as harbor01:13
*** tonanhngo has joined #openstack-kolla01:13
*** tonanhngo has quit IRC01:16
openstackgerritLi Yingjun proposed openstack/kolla-ansible: Update repo in documentation  https://review.openstack.org/40001701:17
*** tovin07 has joined #openstack-kolla01:17
*** zhurong has joined #openstack-kolla01:27
*** jtriley has joined #openstack-kolla01:30
*** jtriley has quit IRC01:36
*** sdake_ has quit IRC01:37
*** hfu has joined #openstack-kolla01:37
*** hfu has quit IRC01:40
*** brad[] has quit IRC01:40
*** hfu has joined #openstack-kolla01:41
*** brad[] has joined #openstack-kolla01:42
*** hieulq has joined #openstack-kolla01:44
*** sdake has joined #openstack-kolla01:46
*** Pavo has quit IRC01:47
sdakesbezverk about?01:47
sbezverksdake: yes but for 2 minutes, what is up?01:50
sdakenothing just wondering why working on a sunday :)01:51
sdakesaw your review of rabbitmq01:51
sdake;)01:51
sdake!!!does not need reviews!!!01:51
openstacksdake: Error: "!!does" is not a valid command.01:51
*** Pavo has joined #openstack-kolla01:52
sbezverksdake: :-) I needed to push it so I could get on another host where I do debugging01:52
sdakecool01:52
sdakei am going to copy that structure for mariadb01:52
sbezverkas soon as I clean it up I will remove first line01:52
sbezverkI hope I did not ofend anybody ;-)01:52
sdakenah i thought it was funny :)01:53
sdakeenjoy sleep time ;)01:53
sbezverksdake: no I am watching "the medici" series..01:54
sbezverkmy wife is a fanatic of Florence01:54
sdakeraining dogs and cats here01:55
sbezverksdake: well it is better than snow01:55
sdakepossible - miss four seasons01:55
sdakealthough i'll never miss shoveling snow01:55
*** liyifeng_seu has joined #openstack-kolla01:56
sbezverksdake: oh man do not remind me. the one winter before last was really bruttal. had to shovel almost every week01:56
sdakedude if i lived somewhere where it snowed ever agagain01:57
sdakesnowblower01:57
sdakeno iffs ands or butts01:57
sdakebuts that is01:57
*** liyifeng has joined #openstack-kolla01:57
sbezverksdake: got to go, have a great evening, talk to you tomorrow..01:58
sdakelater dude01:58
*** jtriley has joined #openstack-kolla02:01
*** tovin07_ has quit IRC02:05
openstackgerritwangwei proposed openstack/kolla-ansible: Add remove-garbage function in kolla-ansible  https://review.openstack.org/40236202:06
*** tovin07_ has joined #openstack-kolla02:08
*** duonghq has joined #openstack-kolla02:10
*** tonanhngo has joined #openstack-kolla02:11
*** jtriley has quit IRC02:12
*** tonanhngo has quit IRC02:12
*** goldyfruit has quit IRC02:16
*** sdake has quit IRC02:21
*** hfu has quit IRC02:22
*** hfu has joined #openstack-kolla02:24
*** Jeffrey4l has joined #openstack-kolla02:28
*** sdake has joined #openstack-kolla02:35
*** sdake_ has joined #openstack-kolla02:41
*** hfu has quit IRC02:41
*** liyifeng_seu has quit IRC02:45
*** hfu has joined #openstack-kolla02:45
*** sdake has quit IRC02:45
*** liyifeng_seu has joined #openstack-kolla02:47
*** jtriley has joined #openstack-kolla02:53
*** tonanhngo has joined #openstack-kolla02:57
*** zhurong has quit IRC02:58
*** zhurong_ has joined #openstack-kolla02:58
*** tonanhngo has quit IRC02:58
*** tonanhngo has joined #openstack-kolla03:03
*** tonanhngo has quit IRC03:07
*** ayoung has joined #openstack-kolla03:38
*** Pavo has quit IRC03:47
*** Pavo has joined #openstack-kolla03:47
*** hfu has quit IRC03:56
*** f13o has quit IRC04:02
*** jtriley has quit IRC04:18
*** jtriley has joined #openstack-kolla04:18
*** mdnadeem has joined #openstack-kolla04:28
*** mdnadeem has quit IRC04:28
*** jtriley has quit IRC04:30
sdake_duonghq around?04:31
*** mdnadeem has joined #openstack-kolla04:31
*** mdnadeem_ has joined #openstack-kolla04:32
*** mdnadeem has quit IRC04:32
*** jtriley has joined #openstack-kolla04:34
*** yuanying has quit IRC04:57
openstackgerritJeffrey Zhang proposed openstack/kolla: Add ironic-ui to horizon image  https://review.openstack.org/40344505:06
openstackgerritJeffrey Zhang proposed openstack/kolla: Install sahara dashboard into horizon image  https://review.openstack.org/40350205:06
*** AnswerGuy has quit IRC05:12
*** jascott1 has joined #openstack-kolla05:15
*** jascott1 has quit IRC05:15
*** jtriley has quit IRC05:17
*** jtriley has joined #openstack-kolla05:17
openstackgerritJeffrey Zhang proposed openstack/kolla-ansible: Enable sahara dashboard when enable_sahara is yes  https://review.openstack.org/40350405:24
openstackgerritJeffrey Zhang proposed openstack/kolla-ansible: Enable sahara dashboard when enable_sahara is yes  https://review.openstack.org/40350405:24
*** jascott1 has joined #openstack-kolla05:35
*** yingjun has quit IRC05:38
*** yingjun has joined #openstack-kolla05:38
*** jrist has quit IRC05:39
*** tonanhngo has joined #openstack-kolla05:43
*** tonanhngo has quit IRC05:45
*** Pavo has quit IRC05:47
*** Pavo has joined #openstack-kolla05:51
*** hfu has joined #openstack-kolla05:58
*** yuanying has joined #openstack-kolla06:02
*** hfu has quit IRC06:04
*** hfu has joined #openstack-kolla06:06
*** hfu has quit IRC06:09
*** hfu has joined #openstack-kolla06:14
*** saneax-_-|AFK is now known as saneax06:16
*** jtriley_ has joined #openstack-kolla06:19
*** jtriley has quit IRC06:22
sdake_duonghq you about06:30
*** hfu has quit IRC06:31
duonghqsdake_, helol06:31
duonghqhello06:31
*** hfu has joined #openstack-kolla06:32
sdake_hi fine sir06:32
sdake_wondering if your interested in getting your dev env setup for kolla-kubernetes dev06:33
sdake_requirements are workstation running linux with 16-32 gb of ram06:33
*** jtriley_ has quit IRC06:34
openstackgerritJeffrey Zhang proposed openstack/kolla: Bump Ubuntu cloud-archive repo to Ocata  https://review.openstack.org/40351606:34
duonghqyou mean the minimum is 16GB of RAM?06:34
sdake_not sure, it works on 32gb06:34
sdake_we can probably make it work on 16gb if that is what you got06:34
sdake_it may work on 16gb out of the box06:34
duonghqso, is there any reasons for it requires more RAM than kolla-ansible?06:35
duonghqas I tested, the k8s does not consume much resource06:35
sdake_the dev environment sets up 4 virtual machines06:35
sdake_the 4 virtual machines make up your kubernetes cluster06:35
duonghqhmm, so the requirements should be 4 nodes with resource for each node is ....?06:35
sdake_our dev environment at present is virtual06:36
sdake_focused towards developers that only have one server to work with06:36
sdake_relies on vagrant06:37
openstackgerritJeffrey Zhang proposed openstack/kolla: Install cloudkitty dashboard in horizon image  https://review.openstack.org/39653506:37
sdake_tell me what you have to work with06:38
duonghqso, we decided using Vagrant for kolla-k8s dev env?06:39
sdake_for the moment06:39
sdake_makes things easier, we have code ready to go for that06:39
sdake_and it works - i have a dev env on my gear in about 30 minutes-1hr06:40
duonghqI do not have experience w/ Vagrant yet, still use kube-deploy to setup k8s cluster06:40
*** jtriley has joined #openstack-kolla06:40
sdake_not familiar with kube-deploy06:40
sdake_if you have a 4 node cluster with kubernetes running on it your all set06:40
sdake_we don't dictate a development environment06:40
sdake_its only to help people that don't have acccess to multiple servers06:41
sdake_which is most people ;)06:41
openstackgerritJeffrey Zhang proposed openstack/kolla-ansible: Enable cloudkitty ui when enable_cloudkitty is yes  https://review.openstack.org/40352006:41
duonghqbasically, developers need 4 node k8s and then everything is the same?06:41
sdake_right06:41
duonghqgot it06:41
sdake_we may need more nodes in teh future06:41
sdake_got it s in you have a 4 node k8s cluster?06:42
duonghqatm, and in my company, it's easier to get 2 16GB workstation than 1 32GB ws, so I think I can test if I can get 4 VM on 1 16GB VM, it's more popular than 32GB ws? (at least for vagrant)06:43
duonghqsdake_, yup, as many nodes as the RAM can hold on06:43
duonghqkube-deploy is nice06:43
sdake_you dont need vms, you could deploy to bare metal06:44
sdake_not sure if vagrant can cross two nodes06:44
sdake_so basically you would need 4 16gb ws or 1 32gb ws06:44
sdake_or maybe it would run on 1 16gb ws06:44
duonghqfor many people, it should work on 1 16GB ws06:45
sdake_ok, well you want a walk thru on that?06:45
duonghqsure,06:45
sdake_now?06:46
*** sdake_ is now known as sdake06:47
*** jtriley has quit IRC06:48
*** zhurong_ has quit IRC06:56
*** zhurong has joined #openstack-kolla06:57
*** msimonin has quit IRC07:02
*** yuanying has quit IRC07:04
sdakeyou seek yoda!07:16
*** sdake_ has joined #openstack-kolla07:22
*** sdake has quit IRC07:24
*** f13o has joined #openstack-kolla07:30
*** harbor has joined #openstack-kolla07:33
*** harbor is now known as portdirect___07:33
portdirect___morning o/07:34
*** Jeffrey4l has quit IRC07:35
*** liyifeng has quit IRC07:37
*** liyifeng has joined #openstack-kolla07:39
duonghqmoring portdirect___07:39
duonghqnice to see you07:39
sdake_sup portdirect___07:39
sdake_do you sleep ;)07:40
portdirect___lol, I deserved that - 7:30 am here :)07:40
sdake_ahh thats a normal time07:40
sdake_i am usually up at 6am07:41
portdirect___just going through logs - duonghq/sdake that vagrant setup only really requires 1gb of ram per machine when just running k8s- though i think we'll need 4gb for the workers once we have some services running07:42
sdake_portdirect___ cool07:42
duonghqportdirect___, so people should have more than 16GB RAM box if they do not want swap is utilized?07:43
*** Jeffrey4l has joined #openstack-kolla07:43
sdake_sp_ pingola07:44
*** hfu has quit IRC07:44
sp_sdake_:07:44
sp_sdake_: hi07:44
sdake_sp_ you put your john hancock on https://github.com/openstack/kolla-kubernetes/blob/master/specs/kolla-kubernetes-arch.rst07:45
*** hfu has joined #openstack-kolla07:45
sdake_sp_ if your still interested, ready to ramp you up on how to get your dev environment setup07:45
sp_sdake_: sure07:46
sdake_sp_ what type of gear do you have07:46
portdirect___duonghq: currently, though we only need 3 workers for ceph and mariadb - so we should be able to get a two (or even single) node setup in less than that07:46
sdake_sp_ typically we expect a single machine with 16gb of ram07:46
sdake_sp_ hopefully running centos although ubuntu should work07:46
*** Pavo has quit IRC07:47
sp_sdake_: though this time dont have 16GB machine, but 1-2 day i can arrange that for env dev07:47
sdake_sp_ what do you have available?07:48
sp_sdake_: its 8gb, 2 core 100GB HDD07:49
sdake_portdirect___ will that work?07:49
sp_sdake_: yes07:49
sdake_sp_ running which os?07:51
portdirect___sdake_/sp_: wouldn't want to use that for more than a single (maybe 2) machine really - but thats still usefull.07:51
sp_sdake_: ubuntu 16.0407:51
portdirect___sp_: virtualbox?07:51
*** Pavo has joined #openstack-kolla07:52
sdake_portdirect___ ubuntu rolls with libvirt07:52
*** matrohon has joined #openstack-kolla07:52
sdake_portdirect___ are ou asking if sp_ is running on bare metal?07:53
duonghqanybody try to add gpg key manually for centos base (w/o download from server)?07:53
sdake_duonghq that works07:53
sdake_duonghq that is how it used to be setup07:54
portdirect___sdake_: should have phrased that better, was meaning to ask if he had (or was using) virtualbox07:54
sp_portdirect___: yes it is VM07:54
duonghqbut we don't have option for custom key atm? (from local file)?07:55
sdake_sp_ what does your bare metal look like?07:55
sdake_duonghq not sure, ask inc0 - he wrote all the customization stuff, I think its possible via customization to do what you want07:55
duonghqok, I'm reading the source code and see that code fragment, only wonder if I can get it done through kolla-build.conf07:56
sdake_duonghq no idea - not sure its documented07:56
duonghqroger07:56
sp_sdake_: are you talking about H/W spec. Seems like i didn't get you07:57
Jeffrey4lduonghq, a workaround https://github.com/openstack/kolla-ansible/blob/master/tools/setup_gate.sh#L4907:57
sdake_sp_ yup hardware spec07:58
sdake_yo uwill want to run the dev env on bare metal07:58
sdake_because it uses virtual machines07:59
sdake_virt on virt on virt not ideal ;)07:59
sdake_we got virt for your virt for your containers to run virt in, YO DAWG!07:59
portdirect___vagrant can also pull the strings on via openstack - never tried it though08:00
sp_sdake_: ok, i will try to arrange it soon and will make sure that env get setup08:00
sdake_sp_ you will need help setting up the env08:00
sdake_as it is undocumented08:00
sdake_our deadline is 20th for ocata-208:00
sdake_we dont hvae alot of  time to jerk around :)08:00
*** DTadrzak has joined #openstack-kolla08:01
portdirect___sdake, what should my prority be this morning? documenting the dev env?08:01
sp_sdake_: roger08:01
duonghqJeffrey4l, thank you,08:01
sdake_portdirect___ nah srwilkers is doing that08:02
portdirect___great :)08:02
sdake_portdirect___ i dunno what your priority should be08:02
sdake_work on what you think is important :)08:02
*** tonanhngo has joined #openstack-kolla08:02
portdirect___did you manage to have time to look at maridb?08:02
sdake_portdirect___ i looked at the k8s templates08:03
sdake_looked like greek to me :)08:03
sdake_but i'll get to it tomorrow morning08:03
sdake_it is afterall sunday08:03
sdake_;)08:03
sdake_and i've totally brain dumped since thursday08:04
*** tonanhngo has quit IRC08:04
portdirect___:)08:04
sdake_portdirect___ one thing we need is to ramp up peoples dev environments08:04
sdake_duonghq and sp_ are interested in contributing08:04
sdake_sp_ may i guide you to the following launchpad08:05
portdirect___great - duonghq/sp_ hit me up if you need anything08:05
duonghqsure, portdirect___08:05
duonghqwhat is your tz, portdirect___?08:05
portdirect___sp_: I may be able to get you some hardware in an hour once I'm in the office if that helps08:05
portdirect___duonghq: GMT08:05
duonghqroger08:06
sdake_https://launchpad.net/kolla-kubernetes/+milestone/ocata-208:06
*** yingjun has quit IRC08:07
sdake_sp_ if you want to choose a thing to work on that would rock :)08:07
sp_sdake_: ok08:08
sdake_sp_ are you in the drivers team?  if not, what is your launchpad id08:09
sp_sdake_: yes I am in driver team08:09
*** jtriley has joined #openstack-kolla08:09
sdake_sp_ nic08:09
sdake_e08:09
sp_sdake_: yes08:10
sdake_anyone else in channel interested in contributing to kolla-kubernetes at this time?08:10
sp_sdake_: nic "sp_"08:10
berendtmorning08:11
*** shardy has joined #openstack-kolla08:11
sdake_sup berendt08:12
*** jtriley has quit IRC08:15
berendtopened a wishlist bug for nova vmware driver a few days ago and they implemented the features as requested 2 days later. very cool.08:15
sdake_berendt i thought nova already had a vmware driver?08:19
berendtyes of course, i needed a new config parameter to be able to change the scheduling behavior08:19
sdake_i see08:19
sdake_nice08:20
*** egonzalez90 has joined #openstack-kolla08:20
sdake_sup egonzalez9008:20
sdake_you are beaten08:20
sdake_its useless to resist08:20
sdake_don't let yourself be destroyed as obi-one did!08:20
duonghqJeffrey4l, the workaround only disable apt key check, but if the key cannot be retrieved, the build process still fails?08:21
egonzalez90morning08:21
sdake_ok peeps well got to hit the rack08:22
Jeffrey4lduonghq, sorry, didn't u want add some file into images?08:22
duonghqJeffrey4l, I want to add some key the to image08:22
portdirect___laters sdake_08:22
sdake_portdirect___ if you see folks kicking around that are signed up as contributors on the specification, could you get them setup dev env wise?08:22
duonghqJeffrey4l, got you idea08:22
duonghqthank you08:22
Jeffrey4lduonghq, add the file and import it ;)08:22
portdirect___np - I'll be communiting for the next 45 - but will be here for the rest of the day08:23
sdake_portdirect___ roger - ok night all08:23
*** sdake_ has quit IRC08:23
*** yingjun has joined #openstack-kolla08:28
openstackgerritVladislav Belogrudov proposed openstack/kolla-ansible: Add tool to merge passwords during release upgrade  https://review.openstack.org/40253408:32
openstackgerritMerged openstack/kolla-ansible: Update repo in documentation  https://review.openstack.org/40001708:40
*** portdirect___ has quit IRC08:48
openstackgerritjimmygc proposed openstack/kolla: Add strongswan/openswan package to vpn-agent source image Lacking of strongswan/openswan package make it impossible to use create vpn connections on images built from source. Adding strongswan/openswan package fixes this bug.  https://review.openstack.org/40356108:50
*** unicell1 has joined #openstack-kolla08:55
*** unicell has quit IRC08:57
*** jtriley has joined #openstack-kolla09:03
openstackgerritzhubingbing proposed openstack/kolla-ansible: Fix kibana command path error  https://review.openstack.org/40356909:05
*** jtriley has quit IRC09:09
*** Serlex has joined #openstack-kolla09:13
berendtneutron vpnaas is not loger part of the neutron governance, we should keep this in mind09:16
*** portdirect_away is now known as portdirect09:17
berendtBesides, I fear my implementation in kolla has never worked properly :(09:17
*** DTadrzak has quit IRC09:20
*** nradojevic has joined #openstack-kolla09:21
openstackgerritVladislav Belogrudov proposed openstack/kolla-ansible: Add tool to merge passwords during release upgrade  https://review.openstack.org/40253409:23
*** gfidente has joined #openstack-kolla09:24
*** mgoddard has joined #openstack-kolla09:24
*** rmart04 has joined #openstack-kolla09:36
*** liyifeng has quit IRC09:42
*** liyifeng_seu has quit IRC09:42
*** yingjun has quit IRC09:43
*** yingjun has joined #openstack-kolla09:43
*** DTadrzak has joined #openstack-kolla09:43
*** yingjun has quit IRC09:46
nradojevicHi there, I want to deploy an additional neutron container (lbaas) into my environment. I built the neutron images and started a deployment for neutron, but this did not affect my old neutron containers. Now I have the problem that I get a new UID for the neutron user in the lbaas container, but in the old ones there is still the old UID. Is there a way to handle this without deleting all neutron containers?09:46
*** yingjun has joined #openstack-kolla09:46
*** Pavo has quit IRC09:47
*** athomas has joined #openstack-kolla09:47
*** mdnadeem_ has quit IRC09:50
*** yingjun has quit IRC09:51
*** Pavo has joined #openstack-kolla09:51
*** mdnadeem_ has joined #openstack-kolla09:53
*** jtriley has joined #openstack-kolla09:58
*** hfu has quit IRC09:59
*** zhurong has quit IRC10:02
*** jtriley has quit IRC10:03
*** mgoddard_ has joined #openstack-kolla10:03
*** jascott1 has quit IRC10:05
*** mgoddard has quit IRC10:06
*** DTadrzak has quit IRC10:09
nradojevicAny ideas?10:14
*** stevemar has quit IRC10:19
pbourkeberendt: vpnaas works fine, I tested it just last week10:20
pbourkeberendt: well a few patches were needed for redhat but overall it works10:20
berendtpbourke: good to know, thanks10:20
berendtI am actually not sure if we should deprecate the feature because of the missing neutron governance10:20
pbourkeberendt: na I think we should keep it unless it goes into complete disrepair10:21
berendtmakes sense, what about a warning note in the docs?10:21
egonzalez90berendt: is there any info about that, not hear about deprecation of vpnaas10:22
egonzalez90maybe they move into a separate repository?10:22
berendtegonzalez90: it is already in a separate repository10:23
berendtegonzalez90: http://lists.openstack.org/pipermail/openstack-dev/2016-November/107384.html10:23
egonzalez90berendt: thanks10:25
*** jmccarthy has quit IRC10:30
*** jmccarthy has joined #openstack-kolla10:31
*** lukl has quit IRC10:31
egonzalez90I think we should deprecate the service, not this release. but in next releases will be potentially broken.10:31
berendtegonzalez90: this is my fear..10:32
*** lukl has joined #openstack-kolla10:34
pbourkewe could move it to the contrib directory10:35
*** ccesario has quit IRC10:36
*** DTadrzak has joined #openstack-kolla10:38
*** pcaruana has joined #openstack-kolla10:45
*** ccesario has joined #openstack-kolla10:46
*** tovin07_ has quit IRC10:50
*** duonghq has quit IRC10:53
*** mgoddard_ has quit IRC10:54
*** mgoddard has joined #openstack-kolla10:56
*** Bico_Fino has joined #openstack-kolla10:59
*** lukl has quit IRC10:59
*** zhubingbing has quit IRC11:09
*** mliima has joined #openstack-kolla11:22
*** portdirect is now known as portdirect_away11:22
*** Bico_Fino has quit IRC11:23
mliimamorning guys11:24
egonzalez90morning mliima11:27
*** lukl has joined #openstack-kolla11:31
openstackgerritjimmygc proposed openstack/kolla: Add strongswan to neutron-vpn-agent source image  https://review.openstack.org/40183411:39
*** derekjhyang has joined #openstack-kolla11:43
*** rock_ has joined #openstack-kolla11:44
*** Pavo has quit IRC11:47
*** sdake has joined #openstack-kolla11:48
*** slagle has joined #openstack-kolla11:49
*** Pavo has joined #openstack-kolla11:51
*** duonghq has joined #openstack-kolla12:00
pbourkeharlowja: ping12:05
*** yingjun has joined #openstack-kolla12:10
*** portdirect_ has joined #openstack-kolla12:15
*** dave-mccowan has joined #openstack-kolla12:19
pbourkeegonzalez90: berendt: any of you guys actively using kibana?12:24
egonzalez90pbourke: no, zhubinbing is using it IIRC12:24
*** portdirect__ has joined #openstack-kolla12:31
* portdirect__ slaps portdirect_ around a bit with a large trout12:31
*** yingjun_ has joined #openstack-kolla12:32
*** mdnadeem_ has quit IRC12:35
*** yingjun has quit IRC12:36
*** mdnadeem_ has joined #openstack-kolla12:39
*** pbourke has quit IRC12:41
*** pbourke has joined #openstack-kolla12:42
berendtpbourke: yes12:43
pbourkeberendt: do you have any visualisations/searches you recommend? Trying to make use it to see if I can more easily diagnose problems with my deploys12:44
sdakemorning peeps12:48
sdakepbourke sup dude12:48
pbourkemorning sdake12:48
sdakeportdirect__ egonzalez90 berendt yo12:49
sdakepbourke why the q about kibana12:49
pbourkesdake: trying to start actually making use of it12:50
sdakei have some experience with it  - setup a dashboard - its a pretty cool tool12:50
sdakeit took me about 8 hours to setup a dashboard12:50
pbourkesdake: so far I've been using the ol ssh n' grep12:50
sdake6 of those was learning12:50
pbourkewhich leaves a lot to be desired12:50
portdirect__hey sdake12:51
sdakepbourke the dashboard functionality is super impressive12:51
sdakeone of hte things that was on akwasnie 's plate was making an autoloading dashboard12:52
sdakeperhaps she can share her dashboard she used for her demo12:52
sdakeperhaps elemoine_ can share it as well12:52
sdakeI no longer have the dashboard I created12:52
pbourkesdake: yeah Im touching up the docs as part of my learnings12:52
sdakeor if I do, its on a laptop half disassembled12:52
*** tonanhngo has joined #openstack-kolla12:53
sdakewhat is really needed there is to autoload the dashboard12:53
sdakeso we ship a default12:53
sdakei dont have the dashboard that was used for austin summit12:53
*** tonanhngo has quit IRC12:55
sdakealso perhaps inc0 hsa the dashboard12:55
sdakesomeone of those 3 who presented the "These are not the logs you are looking for" talk would hopefully have the kolla dashboard12:55
*** lamt has quit IRC12:58
openstackgerritMauricio Lima proposed openstack/kolla-ansible: Load manila-ui in horizon  https://review.openstack.org/40368013:01
openstackgerritMerged openstack/kolla: Add strongswan to neutron-vpn-agent source image  https://review.openstack.org/40183413:02
*** DTadrzak has quit IRC13:09
openstackgerritMauricio Lima proposed openstack/kolla: Add manila-ui in horizon image  https://review.openstack.org/40368913:13
*** nradojevic has quit IRC13:13
berendtAt the moment I only have basic dashboards. I think it makes most sense when we add basic dashboards to the Kibana Ansible role. This way we can ship some basics.13:15
berendtIs anybody using irccloud? I am using the trial version at the moment... Are they stable and trustworthy?13:20
*** liyifeng has joined #openstack-kolla13:29
*** dims has quit IRC13:33
*** jtriley has joined #openstack-kolla13:40
*** mdnadeem_ has quit IRC13:41
*** tovin07_ has joined #openstack-kolla13:42
*** dims has joined #openstack-kolla13:44
*** athomas has quit IRC13:44
*** zhubingbing has joined #openstack-kolla13:45
*** jtriley has quit IRC13:45
*** Pavo has quit IRC13:47
*** tonanhngo has joined #openstack-kolla13:51
*** Pavo has joined #openstack-kolla13:51
*** portdirect__ has quit IRC13:52
*** tonanhngo has quit IRC13:52
*** sp_ has quit IRC13:53
*** jheroux has joined #openstack-kolla13:53
sdakepbourke berendt can I get some action on https://review.openstack.org/#/c/399582/13:56
sdakeegonzalez90 too :)13:56
*** fguillot has joined #openstack-kolla13:56
*** yingjun has joined #openstack-kolla13:59
*** lamt has joined #openstack-kolla14:00
*** BIOS_Hazard has joined #openstack-kolla14:01
*** inc0 has joined #openstack-kolla14:01
inc0good morning14:02
*** yingjun_ has quit IRC14:02
*** BIOS_Hazard has left #openstack-kolla14:03
inc0guys, I'm a bit sick, so please give me a slack this week;)14:04
sdakesup inc014:05
inc0started from food poisoning, now muscle inflammation, popping codeine14:05
inc0mild fever14:05
openstackgerritMerged openstack/kolla: Clean up README.rst in docker repo  https://review.openstack.org/39958214:06
sdakesounds like painful and fun at same time :)14:06
inc0nothing too bad, but thinking is hard14:06
sdakesurprised your do gave you codeine for inflammation14:06
sdakethere re much better meds for that14:06
inc0yeah me too14:06
inc0but after food poisoning options are limited too14:07
sdakeright- not a doctor myself :)14:07
inc0also, I'm not sure (neither they were) that it is inflammation14:07
inc0so just for the pain;)14:07
sdakejust armchair phsychiatrist ;)14:07
*** liyifeng_seu has joined #openstack-kolla14:07
duonghqI build kolla image successfully on Arch linux w/ Python 2.7, anybody try Python 3.5? I got error relate to bytes/str compatible14:08
sdakeduonghq our build script doesn't work oon python 3.5 iirc14:08
sdakeit does work on python 3.4 i think tho14:08
sdakeit needs some porting i think14:08
duonghqdo we have plan migrate it to python 3.5?14:08
*** goldyfruit has joined #openstack-kolla14:08
sdakeduonghq that was part of the goals discussion for ocata14:09
sdakethe goal got shot down14:09
rock_Hi. I want to know one thing.  who is using kolla mostly?14:09
sdakeor pushed down the road14:09
duonghqsdake, so, I'll write a bp or bug for this?14:09
sdakerock_ operators14:09
sdakeduonghq whatever you think makes sense I think14:09
rock_sdake: may I know who were those operators? In which cases kolla is more suitable for them?14:10
duonghqsdake, roger, so I'll make it in the same way as other project14:11
sdakerock_ you said 1 tihng not 2 :)14:11
sdakerock_ so its not like we have a master list of operators using kolla14:11
inc0rock_, mostly when they want to deploy openstack;)14:11
sdakekolla has no typical analytics companies use14:11
sdakerock_ because kolla is a project not a product14:12
*** hvlad has joined #openstack-kolla14:12
sdakerock_ so I can't answer #2 for you :)14:12
inc0rock_, what is exactly you want to know?14:12
rock_sdake/inc0 : Hi. i am new to Kolla, magnum and kuryr. I want to know which one is used by industries mostly and why?14:15
inc0rock_, they solve *totally* different problems14:16
*** lrensing has joined #openstack-kolla14:16
inc0I suggest to watch "Containers: mapping the landscape" keynote from last summit14:16
sdakemagnum -> join chnanel #openstack-containers14:16
sdakekuryr -> join channel #openstack-kuryr14:17
sdakekolla is used by operators to deploy openstack as per our mission - why do operators use operators to deploy Kolla?  We don't have hard fast data, but my opinion it all comes down to ease of use14:18
duonghqsdake, inc0 https://blueprints.launchpad.net/kolla/+spec/support-python-3.5 here is the bp I set the milestone to o-3 due to we have many work has higher priority for o-214:19
inc0duonghq, thanks14:19
rock_inc0/sdake: OK. Thank you.  Yes I had a look on "Containers: mapping the landscape" keynote .14:19
sdakerock_ we are not experts in magnum nor kuryr14:20
sdakethey have their own irc channels, if you want ot understand their roles, ask there14:20
duonghqanybody think the magnum channel's name is somewhat misleading?14:20
rock_sdake: OK. thank you .14:20
egonzalez90duonghq: yes, sound more like a general containers in openstack14:21
*** yingjun has quit IRC14:22
*** yingjun has joined #openstack-kolla14:22
duonghqinc0, np14:22
duonghqmy keyboard has one more key screw up14:22
sdakeduonghq pretty much :)14:24
openstackgerritPaul Bourke (pbourke) proposed openstack/kolla: Update Kibana documentation  https://review.openstack.org/40372914:24
*** yingjun has quit IRC14:26
*** yingjun has joined #openstack-kolla14:27
*** openstackstatus has quit IRC14:28
*** openstackstatus has joined #openstack-kolla14:29
*** ChanServ sets mode: +v openstackstatus14:29
*** yingjun has quit IRC14:31
*** rhallisey has joined #openstack-kolla14:32
*** inc0 has quit IRC14:37
mliimaguys, need some act here https://review.openstack.org/#/c/403680/ and https://review.openstack.org/#/c/403689/14:40
mliimaand here https://review.openstack.org/#/c/403445/214:40
*** v1k0d3n has joined #openstack-kolla14:42
*** liyifeng has quit IRC14:45
*** sdake_ has joined #openstack-kolla14:46
*** mgiles has joined #openstack-kolla14:48
*** ccesario has quit IRC14:48
*** sdake has quit IRC14:49
*** fguillot has quit IRC14:52
*** fguillot has joined #openstack-kolla14:52
*** liyifeng has joined #openstack-kolla14:53
openstackgerritSerguei Bezverkhi proposed openstack/kolla-kubernetes: WIP Rabbitmq HELM Chart  https://review.openstack.org/40345614:57
*** TxGirlGeek has joined #openstack-kolla15:00
*** ccesario has joined #openstack-kolla15:00
openstackgerritSerguei Bezverkhi proposed openstack/kolla-kubernetes: WIP Rabbitmq HELM Chart  https://review.openstack.org/40345615:02
*** tonanhngo has joined #openstack-kolla15:03
*** tonanhngo has quit IRC15:04
*** saneax is now known as saneax-_-|AFK15:11
openstackgerritMerged openstack/kolla: Add ironic-ui to horizon image  https://review.openstack.org/40344515:14
*** v1k0d3n has quit IRC15:15
*** v1k0d3n has joined #openstack-kolla15:16
*** hvlad has quit IRC15:17
*** v1k0d3n has quit IRC15:19
*** v1k0d3n has joined #openstack-kolla15:19
*** f13o has quit IRC15:22
*** liyifeng_seu has quit IRC15:25
*** v1k0d3n has quit IRC15:27
*** v1k0d3n has joined #openstack-kolla15:31
*** portdirect__ has joined #openstack-kolla15:31
sdake_mgiles up for ramping up your dev env for kolla-kubernetes dev?15:33
*** jtriley has joined #openstack-kolla15:33
*** portdirect_ has quit IRC15:35
*** dims has quit IRC15:36
*** liyifeng_seu has joined #openstack-kolla15:40
*** dims has joined #openstack-kolla15:41
*** mliima_ has joined #openstack-kolla15:41
*** awidders_ has joined #openstack-kolla15:42
*** sdake has joined #openstack-kolla15:42
*** mliima has quit IRC15:44
*** awiddersheim has quit IRC15:44
*** sdake_ has quit IRC15:45
*** Pavo has quit IRC15:47
*** harlowja_at_home has joined #openstack-kolla15:51
openstackgerritMerged openstack/kolla-ansible: Add tool to merge passwords during release upgrade  https://review.openstack.org/40253415:51
*** Pavo has joined #openstack-kolla15:52
*** hvlad has joined #openstack-kolla15:52
*** sayantan_ has joined #openstack-kolla15:53
*** derekjhyang has quit IRC16:01
*** hieulq has quit IRC16:01
*** tovin07 has quit IRC16:01
*** cliles has quit IRC16:01
*** david-lyle has quit IRC16:01
*** nihilifer has quit IRC16:01
*** karlamrhein has quit IRC16:01
*** Kimmo__ has quit IRC16:01
*** zigo has quit IRC16:01
*** tovin07_ is now known as tovin0716:01
*** zigo has joined #openstack-kolla16:01
*** Kimmo__ has joined #openstack-kolla16:02
*** cliles has joined #openstack-kolla16:02
*** tovin07_ has joined #openstack-kolla16:02
*** david-lyle has joined #openstack-kolla16:02
*** hieulq has joined #openstack-kolla16:02
*** nihilifer has joined #openstack-kolla16:02
*** Satya_ has joined #openstack-kolla16:03
*** jtriley has quit IRC16:04
openstackgerritMauricio Lima proposed openstack/kolla: Add manila-ui in horizon image  https://review.openstack.org/40368916:04
*** karlamrhein has joined #openstack-kolla16:07
*** shardy has quit IRC16:10
jmccarthyHiya I'm looking at an mtu related change in this commit here https://github.com/openstack/kolla-ansible/commit/a627681eec09eacd208b21d35ab7ad76245319e4#diff-4bf34bbf9cbc26c3eeb2f87a100bc63116:11
jmccarthyThere is some accompanying docs that say advertise_mtu is True by default:16:11
jmccarthyhttps://review.openstack.org/#/c/302165/4/doc/networking-guide/source/adv-config-mtu.rst16:11
jmccarthyDoes this mean it should be specified in ansible/roles/neutron/templates/neutron.conf.j2, or that even if it is omitted from here (which it looks like it is) that it will advertise by default ?16:11
jmccarthyI'm thinking we may need to have an entry in here to explicitly set this to True ?16:11
*** vhosakot has joined #openstack-kolla16:14
openstackgerritMerged openstack/kolla-ansible: Updated from global requirements  https://review.openstack.org/40218716:15
*** f13o has joined #openstack-kolla16:18
sdakejmccarthy openstack-neutron is where i'd ask - tell em your doing dev on kolla should help you get some kind of answer ;)16:18
sdakejmccarthy otherwise they will consider it some kind of support question16:19
sdakewhich neutron doesn't do :)16:19
sdakedave-mccowan sbezverk ^^ any idea on jmccarthy 's question?16:19
jmccarthysdake: Ok, thanks ! I'm going to test it out a bit more myself also and see what happens :) Will be able to tell soon hopefully16:20
sdakejmccarthy sounds good16:20
sdakejmccarthy note the defaults are printed in the container logs16:21
sdakelooking at the neutron logs should help you understand the defaults as well16:21
*** zhubingbing has quit IRC16:21
*** v1k0d3n has quit IRC16:22
openstackgerritMerged openstack/kolla: Install sahara dashboard into horizon image  https://review.openstack.org/40350216:23
sbezverkjmccarthy: sdake: It makes sense to have mtu advertisement on by default as there are lots of higher level protocols in networking world relying on this feature16:23
*** MarMat has joined #openstack-kolla16:24
vhosakotback from a great vacation.  Hope everyone is doing great and had a great Thanksgiving!  I needed all day yesterday to catch up on 1 month's emails16:24
vhosakotlothian [11:22 AM]16:24
vhosakot@vhosakot you were gone?16:24
vhosakothaha, bad paste!16:24
duonghqhmm, I use all-in-one playbook to deploy to one node and then run post-deploy con deploy node, it still try to connect to target node with local connection -> failed16:28
duonghqtarget node != deploy-node16:29
duonghqiirc, we do not have this issue before?16:29
*** kristian__ has joined #openstack-kolla16:30
duonghqah, see the problem16:30
sdakevhosakot you been out for a mo?16:30
duonghqneed sleep now, bye all16:30
sdakevhosakot wtb your schedule ;)16:30
*** duonghq has quit IRC16:30
vhosakotsdake: haha, yes.. was out for a month... couldn't chat with you in Barcelona.. attended all your talks.. they were great16:31
sdakethanks16:31
sdakedidn't see ya in audience but there were slews of people16:31
openstackgerritMauricio Lima proposed openstack/kolla: Add manila-ui in horizon image  https://review.openstack.org/40368916:32
sdakevhosakot you making it to PTG?16:32
*** kristian__ has quit IRC16:35
*** kristian__ has joined #openstack-kolla16:35
sbezverksdake: I cleaned up rabbitmq charts, still WIP but please share your thoughts if you have time.16:37
sdakesbezverk sure you got it16:37
sdakewould like kfox1111 's input as well, since he started the helm work16:38
sdakesbezverk note i noticed a patch was +w'ed with only 1 +216:38
sbezverksdake: that was the agreement we had for kube ps16:38
sdakesbezverk because core team was too small?16:39
*** schwicht has joined #openstack-kolla16:39
sbezverksdake: because there were only 3 guys working on it before16:39
*** eaguilar has joined #openstack-kolla16:41
*** v1k0d3n has joined #openstack-kolla16:41
*** jascott1 has joined #openstack-kolla16:41
*** schwicht has quit IRC16:42
*** harlowja_at_home has quit IRC16:42
kfox1111morning.16:42
*** schwicht has joined #openstack-kolla16:42
sdakeok - well there is bound to be some deltas in the patches - lets sort out the delta with 2 +2s :)16:43
sdakekfox1111 sup dude - can you hit the review queue, have some qs in your review, and had a slew of em in sbezverk 's as well16:43
kfox1111sdake: yeah. was going to try and do that this morning.16:43
sdakekfox1111 also - thanks to portdirect__ got my env setup16:43
kfox1111along with everything else. :)16:43
sdakekfox1111 tell me about it :)16:43
kfox1111portdirect__: thanks. :)16:44
sdakekfox1111 working on mariadb chart today if the stars align (as in I can fix my $460 water bill leak)16:44
kfox1111ouch.16:44
kfox1111k.16:44
sdakenormally my water bill runs about 80 bucks16:44
sdakeya the 380 isn't isn't too bad but hate to waste resources :)16:45
sdakethe 380 being resources and the water being resources16:45
kfox1111yeah. thats a fair amount. several hot tubs worth I think.16:45
sdakeits about 120*8 showers :)16:46
sbezverkkfox1111: good morning!16:46
kfox1111sbezverk: good monring. :)16:46
sdakekfox1111 sbezverk you fellas might consider beefing up the core team with nominations at some point16:47
kfox1111sdake: yeah.16:47
kfox1111need to see more 3rd party reviews. have had some. hopfully more this last week?16:47
sdakeyup i think srwilkers and portdirect__ going to hit it harder ;)16:48
*** v1k0d3n has quit IRC16:48
vhosakotsdake: I'll let you know about PTG.. sure16:48
kfox1111sdake: +116:49
sdakeand possibly dounghq and mgiles :)16:49
kfox1111sdake: yeah. PTG would be good to know about. I think I'm tentitively scheduled to go.16:49
*** tovin07 has left #openstack-kolla16:49
sdakekfox1111 did you get a ticket yet?16:49
sdakekfox1111 they may be sold out16:49
sdakethere were only 500 available16:49
kfox1111sbezverk: so, have you tried to implement a service package yet? I don't think your global aproach will work with them.16:49
kfox1111sdake: already?16:50
kfox1111wow...16:50
kfox1111hmm...16:50
sdakei dont know for sure if they are sold out16:50
sdakethe tickets are 100$16:50
kfox1111that would be unfortunate16:50
kfox1111looks like they may still be open.16:50
kfox1111so is anyone else going?16:51
sdakeya i think core team members would not be part of the limit but what do I know16:51
sdakekfox1111 don't have that information - inc0 may16:51
sdakeinc0 out sick today i think16:51
kfox1111k. cause that may infuence if I can go.16:51
kfox1111and if they are hard to come by, I shoudl get them asap.16:52
kfox1111k.16:52
sdakekfox1111 all I can tell you about my attendance at this point is I've bought a ticket ;)16:52
sdakehowever kolla has scheduled time at the ptgs16:52
sdakeso it would be a serious shame to waste that scheduled f2f time16:52
sdakekfox1111 service package = e.g. nova?16:53
kfox1111sdake: yeah. thanks.16:53
kfox1111sdake: yes.16:53
sbezverkkfox1111: I am waiting for helm 2.1 it should bring missing piece, lots of people asking for the same functionality16:54
kfox1111sbezverk: have an issue link?16:54
sbezverkkfox1111: yep, one sec16:54
kfox1111there's a huge amount of duplication in here without templating charts. :/16:54
mliima_sdake, can you review https://review.openstack.org/#/c/403689/ and https://review.openstack.org/#/c/403680/16:54
sbezverkkofx1111: https://github.com/kubernetes/helm/issues/1568#issuecomment-26208650416:55
mliima_thanks sdake16:56
sdakemliima_ both lgtm16:56
kfox1111sbezverk: thanks. looking...16:57
*** rmart04 has quit IRC16:57
openstackgerritMerged openstack/kolla-ansible: Load manila-ui in horizon  https://review.openstack.org/40368016:58
kfox1111sbezverk: that aproach you linked to sounds like a variant of my first prototype but with a plugin.17:00
kfox1111I made a single package conditionally include subresources using just the engine.17:00
kfox1111I don't like the proposed solution in the issue though. it would make it very difficult on automation that could do upgrades.17:00
kfox1111I'd rather be able to actually launch the packages without having a parent at all.17:01
*** liyifeng_seu has quit IRC17:01
kfox1111and it doesn't address having 2 layers of parents.17:01
kfox1111which I think will break the globals idea.17:01
*** liyifeng_seu has joined #openstack-kolla17:02
kfox1111for my use case, I need to be able to lauch the individual microservices as helm instances, so I can upgrade them or change them a single piece at a time.17:03
kfox1111I need 0 orchestration that could break things running.17:04
*** matrohon has quit IRC17:04
*** egonzalez90 has quit IRC17:04
sdakekfox1111 can we get htis information in the reviews17:05
sdakekfox1111 in irc is cool too, but without reviews, people reviewing won't have this context if they aren't in irc at this particular point in time :)17:05
kfox1111yup.17:05
sdakekfox1111 justto be clear, you want to launch each individual microservice such s nova-proxynovnc as a helm chart?17:07
sdakeor whatever that component is called :)17:08
kfox1111sdake: correct.17:08
kfox1111then when I update nova-novncproxy, there should be 0 risk of nova-api getting affected.17:08
kristian__Hi sdake please may you help me?17:09
kfox1111helm list should show them all.17:09
kristian__We may have a solution17:09
sdakekristian__ solution to what17:09
kristian__Change libvirt to 1.2.1717:09
kristian__https://bugs.launchpad.net/nova/+bug/164241917:09
openstackLaunchpad bug 1642419 in kolla-ansible "GPU Passthrough isn't working" [Low,Confirmed]17:09
sdakekristian__ i don't see that solution in the bug log - where did you come up with that17:10
sdakenote we have no control over libvirt versions - those come from upstream distros17:10
kristian__from nova team17:10
kristian__oh17:10
sdakewhich version of libvirt is in your container17:11
kristian__dunno17:11
*** eaguilar has quit IRC17:11
portdirect__kfox1111: word up!17:11
sdakedocker exec nova_libvirt rpm -qi libvirt17:11
sdakekfox1111 that sounds pretty cool17:11
kristian__1.3.117:11
kfox1111portdirect__: morning.17:12
sdakekfox1111 that isn't something that helm supports correct?17:12
sdakekristian__ so we have a newer version then 1.2.1717:12
kristian__than 1.2.1717:12
kristian__yes17:12
sdakedownstreams follow upstreams, not the other way around17:12
kfox1111sdake: well, each helm root chart is 'instantiated'. its given a name and can then be acted upon.17:12
kfox1111I don't think subcharts get names.17:12
sdakenova should be following the latest upstream, not some specific version that is old :)17:12
kristian__he did that gpu passthrough with 1.2.17 and it worked for him17:13
sdakedid he see the ethtool spam17:13
kristian__yes17:13
sdakehis take on that was what?17:13
kristian__kristian__: it doesn't makes sense that libvirt tries to do something with ethtool on this kind of device17:13
sdakedberrange is the fella to talk to in the nova channel17:14
sdakehe maintains libvirt17:14
sdakeget em to put the stuff in the buglog17:14
sdake:)17:14
sdakether eis no wy to run 1.2.17 of libvirt17:15
sdakethat I know of17:15
sdakenot sure how the person that tested gpu passthrough has 1.2.1717:15
kristian__ok going to ping him17:15
sdakemaybe thats what is shipped with ubuntu?17:15
kristian__maybe before they've updated the repos17:15
sdakeactually he may not maintian libvirt anymore17:15
kristian__ok17:15
sdakebut he will know who does ;)17:15
kristian__first Im going to manually install that version17:16
*** sdake has quit IRC17:16
kristian__and then we will see17:16
kfox1111we've not had very good luck over here with gpu passthrouhg. :/17:16
*** sdake has joined #openstack-kolla17:16
kfox1111redhat's been recommending to use efi to make it work smoothly, but nova doesn't support it well yet. :/17:17
kristian__I had but dunno on what libvirt17:17
*** eaguilar has joined #openstack-kolla17:18
*** pcaruana has quit IRC17:19
*** portdirect__ has left #openstack-kolla17:23
*** liyifeng has quit IRC17:29
*** liyifeng_seu has quit IRC17:29
*** unicell1 has quit IRC17:29
*** sdake_ has joined #openstack-kolla17:34
*** v1k0d3n has joined #openstack-kolla17:35
*** v1k0d3n has quit IRC17:35
*** Guest81558 has joined #openstack-kolla17:36
*** sdake has quit IRC17:36
kfox1111is there a length limit to a review?17:37
kfox1111:/17:37
sdake_kfox1111 how do  you mean17:42
sdake_the contents of a review?17:42
kfox1111yeah.17:43
sdake_I am not sure - but i suspect there may be - there are always limits in all kinds of software :(17:43
kfox1111well, it took it. :)17:43
*** Guest81558 has quit IRC17:44
*** gfidente is now known as gfidente|afk17:45
*** Pavo has quit IRC17:47
*** Pavo has joined #openstack-kolla17:52
openstackgerritsatya proposed openstack/kolla-ansible: Added separate interface for rabbitmq  https://review.openstack.org/40195717:56
*** aric49 has joined #openstack-kolla17:59
Satya_hi jeffrey4l17:59
Satya_hi sdake17:59
Satya_hi sdake_18:00
harlowjapbourke whats up18:00
*** sayantan_ has quit IRC18:01
*** sayantan_ has joined #openstack-kolla18:01
*** Serlex has quit IRC18:01
Satya_hi pbourke18:03
pbourkeharlowja: just wondering if you guys are using kibana actively in godaddy?18:03
pbourkeSatya_: hi18:04
*** portdirect_away is now known as portdirect18:04
Satya_Please review18:04
harlowjapbourke yes18:04
harlowjaklindgren ^18:04
harlowjapbourke klindgren knows more of the godaddy history than i do though :-P18:04
klindgrenhi18:05
pbourkeharlowja: I was just wondering if you had any useful searches or dashboards you could point me at18:05
harlowjaklindgren 'harlowja: just wondering if you guys are using kibana actively in godaddy?'18:05
pbourkesince this morning I've managed to learn a bit more though18:05
harlowjapbourke redirect to klindgren  :-P18:05
klindgrenso we moved over to a kibana4 infra18:05
klindgrenbut I can point you to our stuff that we did for kibana2 and logstash18:06
*** tonanhngo has joined #openstack-kolla18:06
klindgrenhttps://github.com/godaddy/openstack-logstash18:06
klindgrenwe use a similar one to that just done up in kibana4 now days18:07
harlowjawhy isn't that updated klindgren  :-P18:08
*** sayantan_ has quit IRC18:08
klindgrenbecause I didn't do the move to the kibana4 infra.  and kibana4 took something simple and made it so where you need a phd in kibana to do crap in it :-D18:08
harlowjalol18:09
klindgrenas I understand kibana4 bascially you have to create views.  That individually give you parts of the dash board.  Then you have to create a dashboard that references the views.  Since we dont run our own elk infra anymore.  I am not sure wher those views/dashboard get sotred or if tis possible to save them off into a single uploadedable json format anymore.18:10
klindgrenwhich also means - I am pretty sure that stuff is not under source control. :-/18:11
sbezverkkfox1111: sorry was on the meeting. Yes, I took your common and placed it in _helpers.tpl which will make it avaiable to all sub charts18:14
sbezverkkfox1111: when that plugin is avaiable then we can deploy parent/child/grand-child independently but using globally defined code..18:15
sbezverkkfox1111: also it should apply to packaging18:15
*** rhallisey has quit IRC18:16
klindgrenpbourke, I think the osops repo also has other peoples logstash/kibana configs as well.18:16
*** rhallisey has joined #openstack-kolla18:16
klindgrenhttps://github.com/openstack/osops-tools-logging/tree/master/logstash18:18
*** sdake has joined #openstack-kolla18:18
*** sdake_ has quit IRC18:21
aric49Hi everyone --- quick question about Kolla OS support.   Can anyone here confirm that the kolla build and deploy works on Ubuntu 16.04?18:25
*** jascott1_ has joined #openstack-kolla18:26
*** jascott1 has quit IRC18:26
*** jascott1_ has quit IRC18:26
*** jascott1 has joined #openstack-kolla18:26
portdirectaric49: though I use CentOS, I am aware of people on here who use Ubuntu 16.0418:29
*** inc0 has joined #openstack-kolla18:30
portdirectaric49: there is also gates that run for both building and deployment of kolla on ubuntu18:30
*** tonanhngo has quit IRC18:30
inc0hey, I'm back18:31
*** kristian__ has quit IRC18:31
sdakearic49 can fonfirm it works18:31
aric49@portdirect - Thanks.   Have you successfully built and deployed kolla on 16.04?18:37
inc0aric49, I did, multiple times18:38
inc0xenial works fine18:38
portdirectaric49: ^^18:39
*** kristian__ has joined #openstack-kolla18:39
inc0careful tho, I think aufs is a no-no with xenial18:40
inc0btrfs should be ok tho18:40
*** unicell has joined #openstack-kolla18:41
aric49perfect. Thank you18:42
*** rmart04 has joined #openstack-kolla18:42
*** v1k0d3n has joined #openstack-kolla18:42
*** mgoddard_ has joined #openstack-kolla18:44
*** tonanhngo has joined #openstack-kolla18:44
*** tonanhngo has quit IRC18:45
kfox1111back.18:46
kristian__sdake: are you here?18:46
sdakeOTP18:46
sdakekristian__ there is plenty of other people here who can help :)18:46
*** rmart04 has quit IRC18:47
*** lbeliveau has joined #openstack-kolla18:47
kristian__ok lbeliveau from nova team18:47
*** mgoddard has quit IRC18:47
kristian__could you two or with portdirect do a script to load an older version of libvirt?18:48
kfox1111sbezverk: having a plugin change the behavior of helm in the client seems kind of sketchy to me.18:48
*** mgoddard_ has quit IRC18:48
kfox1111I'd rather have the safest unit of deployment be just the thing I'm trying to update.18:48
inc0kristian__, like build image with older libvirt?18:48
kristian__yes18:49
kristian__because with the newer one gpu passthrough doesnt work18:49
inc0http://docs.openstack.org/developer/kolla/image-building.html#dockerfile-customisation18:49
inc0ahh so it was issue with libvirt version?18:50
kristian__yes18:50
lbeliveauwe still need to validate that, but it looks like it18:50
lbeliveauI booted guest with non-NIC passthrough on 1.2.1718:51
inc0https://github.com/openstack/kolla/blob/master/docker/nova/nova-libvirt/Dockerfile.j2#L24 <- change this package to version you want18:51
*** aric49 has quit IRC18:52
kristian__to libvirt-bin=1.2.17?18:52
*** v1k0d3n has quit IRC18:52
*** v1k0d3n has joined #openstack-kolla18:52
inc0kristian__, you're running ubuntu?18:52
kristian__yes18:53
*** v1k0d3n has quit IRC18:53
lbeliveauFYI on latest centos, libvirt version is 1.2.1718:53
*** v1k0d3n has joined #openstack-kolla18:53
sdakelbeliveau we use virtualization sig version18:54
inc01.3.1 for ubuntu xenial18:54
kristian__so should I try and build for centos?18:54
sdakewhich is 1.3.z18:54
sdakekristian__ aren't you already on centos?18:54
kristian__no18:54
kristian__Im on ubuntu18:54
lbeliveauyeah, I've never tried passthrough on newer versions of libvirt (1.3.x)18:54
sdakelet me check qemu sig version - moment18:55
inc0and yes kristian__ create your customization file with nova_libvirt_packages_remove = ['libvirt-bin'] and nova_libvtirt_packages_append=['libvirt-bin=1.2.17']18:55
kristian__inc0 where to create that file?18:56
*** aric49 has joined #openstack-kolla18:56
inc0it's in docs, you need to build it with customization line, but for now maybe just edit our dockerfile18:56
inc0to quickly test if it works18:56
kristian__ok18:57
sdakevirt sig uses 1.2.1718:57
inc0if it's case indeed, I'll walk you through adding this customization18:57
inc0and probably write it down as workaround for this bug18:57
sdakeif it works on 1.2.17 looks like a bug not in kolla to me ;)18:57
inc0yeah, it's libvirt bug18:58
inc0we can hardcode version of libvirt for ubuntu, but I'd rather not18:58
sdakepossibly not - libvirt does a whole bunch of trickery in docker containers18:58
inc0well not something like that tbh, but let's see18:58
kristian__quick question where is kolla from pip located?18:58
sdakekristian__ on pypi18:59
kristian__path on machine18:59
kfox1111if the bug can be tested automatically, someone should add a gate job. :)18:59
inc0in site-packages I'd think18:59
sdakeon centos its in /usr/lib/python2.7/site-packages/kolla for cmd line code19:00
sdakeon centos /usr/share/kolla for ansible playbooks/dockerfiles19:00
sdakeon ubuntu no idea19:00
kristian__ok I will look19:01
sdakekfox1111 that is a dream that never becomes reality ;(19:01
portdirectWe need a newer version of libvirt to support efi guests (Clear Linux)19:01
sdakeand it also can't be tested automatically since no infra gear has gpus19:01
openstackgerritMauricio Lima proposed openstack/kolla: Add manila-ui in horizon image  https://review.openstack.org/40368919:02
*** jgriffith is now known as jgriffith_away19:02
inc0 /usr/local/share/kolla for buntu19:03
portdirectkristian__: what version of libvirt is in your containers? what version of the containers are you using?19:03
kfox1111sdake: yeah.... but if we don't try at least for known issues, then we're at least doomed to repleat the same mistakes. :/19:03
sdakekfox1111 agree - hopefully someone will take on writing some gating for kolla-ansible19:04
kristian__3.0.1 the containers and the latest libvirt 1.3.119:04
kfox1111sdake: and kolla containers.19:04
sdakekfox1111 what is needed there is cross-deliverable gating19:05
kfox1111I kind of hoped the package tests by rdo/ubuntu would be enough though. :/19:05
kfox1111+119:05
sdakethe real only way to learn gating is to actually do the job yourself19:05
sdakei've attempted training19:05
sdakethat didn't work out19:05
kfox1111yeah. :/19:05
kristian__building19:06
kristian__Version '1.2.17' for 'libvirt-bin' was not found19:06
portdirectkristian__: both the centos binary and source packages are currently building with 1.2.17, you can pull it from the docker hub.19:07
kristian__ok19:07
inc0you can try to run centos on ubuntu19:07
inc0but well, usually not great idea19:07
inc0I'd rather look for different version that is available19:07
kristian__pulling it right now19:08
sdakekristian__ atleast that will answer if its a libvirt version problem19:08
kristian__yes19:08
kfox1111inc0: had an issue last t ime I tried that.19:08
kfox1111ubuntu container expected something of the centos kernel that just wasn't there.19:09
sdakekfox1111 this is centos containers on ubuntu host - maybe different result19:09
kfox1111oh. k.19:10
inc0fun fact, it will be libvirt 2.1 in Yakkety19:10
kristian__if it will be a libvirt problem then we will need to talk with the maintainers19:10
portdirectkfox1111: I've had the same but the other way round is usually a bit better - as ubuntu packages often expect a newer kernel19:10
kristian__or we yan build yakkety containers19:10
kristian__currently pulling libvirt :)19:11
portdirectman - I never worked so hard to get doom working - software rendering FTW :)19:12
kfox1111portdirect: makes sense19:12
kristian__be right back (in max 1-2 hours)19:13
*** kristian__ has quit IRC19:13
*** sayantan_ has joined #openstack-kolla19:14
kfox1111ugg.. a revew is a horible way to do this... :/19:15
sdakekfox1111 on this plugin thing - assuming that works, does it meet the microservice targeted deployment requirement you have?19:15
*** Satya_ has quit IRC19:16
kfox1111sdake: It makes me really uneasy as an op.19:17
sdakekfox1111 better then a spec ;)19:17
sdakecould you expand as to why19:17
kfox1111sdake: but a spec, I can reply inline. we're starting to have to copy/paste/manually append who said what now. :/19:17
sdakethat is what the reply button is for, not the post button :)19:18
*** kristian__ has joined #openstack-kolla19:19
kfox1111sdake: can't do that in the main message. :/19:20
sdakeright - the main message is not really intended for debate - that is what the reply button is for ;)19:20
kfox1111but thats what we're dealing with now. :/19:21
*** hvlad has quit IRC19:21
kfox1111I have to respond to a response in the main message now. :/19:21
sdakei see19:21
sdakewell it helps if everyone uses the same workflow ;)19:21
kfox1111true. :/19:22
kfox1111but the issue is, there's not code to comment on. :/19:22
sdakeas in debate with reply leave comments in review on technical matters19:22
kfox1111its a more general discussion. :/19:22
*** kristian__ has quit IRC19:24
*** kristian__ has joined #openstack-kolla19:30
*** tonanhngo has joined #openstack-kolla19:30
sbezverksdake: so how do you propose to reply to other people comments? It is difficult without quoting since comments are huge.19:37
sdakein the main message you can use the reply button19:37
sdakethat will quote them19:37
sdakethat can be used for whatever - generally i try to avoid it at :)19:38
*** schwicht has quit IRC19:39
*** schwicht has joined #openstack-kolla19:40
kfox1111sdake: k. I'll try that next time.19:41
kristian__Hi19:41
kristian__it looks to be the same error on centos19:41
kristian__gonna check libvirt version19:42
kristian__finally got another error19:42
kfox1111sdake: ah. I see. it doesn't look like a button, so I missed it. :/19:43
kristian__Exceeded maximum number of retries. Exceeded max scheduling attempts 10 for instance f39d1172-8ded-4dce-b54a-173a48e11934. Last exception: internal error: process exited while connecting to monitor: 2016-11-28T19:42:08.353179Z qemu-kvm: -device vfio-pci,h19:43
kfox1111I thought that was part of the collapse message functionality they have.19:43
*** matrohon has joined #openstack-kolla19:44
kfox1111sdake: fair enough. didn't know the procedure. figured -2 was for strong disagreement without further convincing. will remove it.19:45
sdakekfox1111 np - -2 means "GTFO" :)19:45
sdakethe only time i ever -2 a review is if it is flat out wrong or for e.g. when we had the docker repo removal of ansible, people had a ton of reviews outstanding for kolla repo19:46
sdakethose needed -2s to block their merge19:46
kfox1111sdake: ah. k. I think your definition might be stronger then some other projects I've worked on. but works for me. :)19:47
*** Pavo has quit IRC19:47
kfox1111I've seen it as a way to get more discussion amungst cores when 2 +2's wouldn't be enough.19:47
kfox1111or my memory's just shotty. :)19:47
inc0-1 from core should be enough to not merge it19:51
inc0-2 is "never merge it"19:52
*** Pavo has joined #openstack-kolla19:52
inc0we should have -3 too "and don't come back"19:52
inc0(that was a joke...)19:52
kristian__inc0: what do you think of the error I pasted?19:52
inc0kristian__, seems like qemu from centos can't work with kernel from ubuntu19:53
kristian__ok19:53
inc0(you tried to run centos on ubuntu right?)19:53
kristian__then what19:53
kristian__yes19:53
inc0hmm, soo I have idea19:54
kristian__then edit all dockerfiles to use 17.04?19:54
inc0I was thinking 14.0419:54
kristian__to use the new libvirt and it may be fixed there19:54
kristian__or19:54
inc01.2.2 for trusty19:54
kristian__download that kolla image?19:55
inc0won't work19:55
inc0we changed few things since then19:55
kristian__does it have newton?19:55
inc0might not work at least19:55
inc0no19:55
inc0you can try to curl + dpkg tho19:55
inc0with dockerfile customization19:55
kristian__but what about 17.0419:56
kristian__it should work19:56
inc0well, might19:56
inc016.10 tho19:56
inc017.04 is future19:56
kristian__ok so need to edit all dockerfiles to 16.10 right?19:57
inc0so, not really19:57
inc0hold on19:58
kristian__ok19:58
inc0https://github.com/openstack/kolla/blob/master/docker/base/sources.list.ubuntu <- you need to prep sources.list for yakkety19:58
inc0and in your kolla-build.conf change base_tag=16.1019:59
inc0this *might* work19:59
inc0never tried that19:59
*** schwicht has quit IRC19:59
inc0depends how much stuff canonical broke between xenial and yekkety20:00
inc0I hope not too much20:00
inc0and that's assuming libvirt 2+ will have it fixed20:00
*** mgiles has quit IRC20:00
inc0you're doing major version upgrade of libvirts20:00
kristian__inc0: where is kolla-build.conf?20:02
*** schwicht has joined #openstack-kolla20:04
*** mnasiadka has joined #openstack-kolla20:05
sdakekfox1111 apologies otp - 2 +2s are always enough20:13
sdakekfox1111 its not like we rollcall reviews for code ;)20:13
sdake1 +2 isn't sufficient because people can ram shit through that is busted20:13
inc0kristian__, check out building containers doc  section20:13
inc0I gtg - doctors appointment20:14
kfox1111yeah, but something this fundimental shouldn't be left up to just 2 devs.20:14
inc0be back later20:14
kfox1111arguably needs a spec. but we're trying to shortcut that.20:14
kristian__inc0: ERROR: unknown environment 'genconfig'20:14
sdake2 +2s should be sufficient in all cases for code20:15
sdakeif you really want a rollcall vote on the topic, the best way to do that is via spec20:15
sdakebecause rollcall votes are reserved for architectural decisions not implementation decisions20:15
kfox1111sdake: this is like, ansible vs chef though. its really really inherent in which direction the code base will go in. that isn't a 2 +2 thing I think.20:15
sdakei think in the case of helm the issue is of implementation not architecture20:15
kfox1111(not that it matters much, considering how many cores kolla-kubernetes has)20:15
kfox1111sdake: no, its kidn of a architecture thinig. operators vs helm/endpoint like.20:16
kfox1111my poc has tried to allow both. sbezverk's is more oriented towards helm/endpoint from my feel of it. (my 2 cents. I can be wrong)20:17
kfox1111I want to be able to completely devorce orchestration from packages when desired.20:17
kfox1111so I want as fine grained packages as I can get. microsservices level,20:18
kfox1111not service package, with flags that turn off stuff.20:18
*** inc0 has quit IRC20:18
*** jgriffith_away is now known as jgriffith20:18
sbezverkkfox1111: I could not care less about helm and endpoint, I was really happy with jinja2 :-) but since community decided to go with helm, my point is to use it as much as possible before inventing new tools20:18
kfox1111that fits the operator workflow too.20:18
kfox1111sbezverk: but the way your are aranging it is the heml/endpoint aproach.20:19
kfox1111service packages.20:19
kfox1111I'm with you, it would be nice if helm supported natively everything we want.20:20
sbezverkkfox1111: that is not true. the only reason I used entrypoint was because of real dependency20:20
kfox1111Id be 100% on board with having helm do both package build templateing and package instantiation templating if it supported it today.20:20
kfox1111sbezverk: I'm not talking etnrypoint specificially. just the pattern discussed in the spec.20:20
sbezverkbut helm install microservice could be executed from operators as well20:20
sbezverkthat was my understanding20:20
kfox1111sbezverk: complexity.20:20
kfox1111I don't want the complexity of the orchestration engine of select super package, cut out junk just to get the single package I really want getting in the way.20:21
kfox1111thats complicated and potentially dangerious.20:21
kfox1111I'd rather just have it do the 1 exact thing I need it to do.20:21
sbezverkkfox1111: and then invent something on top of that to get everything together?20:22
kfox1111temporarlily, until helm can do it itself.20:22
kfox1111I need the packages to have a good api. the build procedure can evolve over time.20:22
kfox1111if helm can do it natively at some point, I'm all for switching to native tools.20:23
sdakeapis are permanent things :)20:23
kfox1111but  it cant today.20:23
kfox1111exactly. build tools come and go. but api's are semi-forever. :)20:23
sbezverkkfox1111: if I get confirmation from helm that new release will be released before next holidays, would it be soon enough for you?20:24
kfox1111the packages provide an api. that of the vars used when instantiating them.20:24
kfox1111sbezverk: I think that means we can continue the discussion. I'm still not sold on having to use a super package to deploy the primary pacakges.20:24
kfox1111that api seems very wrong.20:25
sdakeok just for reality check, goal is to release this stuff in ocata20:25
sbezverkkfox1111: your manipulation with helm files to get them into package does not sounds safe to me20:25
sdake1.0.020:25
kfox1111sbezverk: is it different then how our containers are being built?20:25
kfox1111it parallels that workflow.20:26
sdakebuild.py is a monster20:26
kfox1111I'm totally ok with redefining how we build it.20:26
kfox1111sdake: yeah. it could use some love.20:26
kfox1111it is a poc after all.20:26
sbezverkkfox1111: we are trying to simplify things not complicate with new tools20:26
sdakeso i mean kolla repos build.py20:26
kfox1111sdake: no, I disagree. simplification is just one aspect of writing sofware.20:27
sdakeso/no20:27
kfox1111not turning off users by lack of features that meet their needs is another.20:27
kfox1111sdake: ah.20:27
sdakeif we can't make a decision we wont have any users to turn off :)20:27
kfox1111we have to ballance features that let users use the system with maintainability (code complexity, maintainability, etc)20:28
sdakei think life was simpler pre-helm ;)20:28
sbezverksdake: +220:28
kfox1111sdake: orchestration in packages for me is a pretty strong showstopper for us.20:28
kfox1111thats why we agreed to helm being a simple packaging mechanism in the spec. :/20:28
sdakeright20:29
sbezverkkfox1111: I do not see orchestration in a package20:29
sdakewho said anything about orchestration in packages?20:29
kfox1111requiring super packages to depoy the real ones (from our perspective) is not following that agreement.20:29
sdake(minus the entrypoint thing rolling in sbezverk's patch that needs attention)20:29
kfox1111imo20:29
sbezverkif you are refering me using entrypoint once for vswitchd, then it is a requirmenet and not orchestration20:29
sdakeour=Kolla perspective?20:29
kfox1111sorry. my empoyer.20:29
sdakeok so lets leave the entrypoint discussion in vswitchd out for now - that is an annoying special case20:30
sdakesince it just complicates the discussion with no real value20:30
kfox1111I want to be able to deploy nova-novncserver package. if I have to jump through nova to do so, it is involving orchestration like things in the mix.20:30
kfox1111I have to trust helms orchestration features won't mess with my deployment. :/20:30
*** eaguilar has quit IRC20:30
kfox1111I'm not that trusting.20:31
sbezverkkfox1111: nova only holds global values20:31
kfox1111and the source to all the other services.20:31
sbezverkwhich you can easily override at lower levels20:31
kfox1111if helm decides wronglyh to update something else, I'm screwed.20:31
kfox1111its far safer/simpler to just not have that code involved so it can't screw things up.20:31
sbezverkkfox1111: if you decide wrongly ;-) ??20:31
sdakeyup i had hoped helm would have slayed the problem you want solved kevin20:31
kfox1111sbezverk: then its on me. thats my job.20:32
kfox1111sdake: it can. my prototype does do it that way.20:32
kfox1111it just can't do it completely 100% with helm's build pipeline as the root of the build process.20:32
sdakehelm doesn't slay it your build tool does so20:32
sbezverkkfox1111: I do not get you do not trust helm but explicitely trust your builder packager?? I am not sure if it a right approach20:32
kfox1111it still uses helm to do the build. ist just assembles the chart before telling helm to build it.20:33
sdakesbezverk i think the builder package tool would be trustworthy as long as helm doesn't change directory layout and structure20:33
kfox1111sbezverk: heh. I especially trust that, condisering I did not write helm, but am writing the packages. ;)20:33
kfox1111but thats an edge case. ;)20:34
kfox1111sdake: and their layout is their contract.20:34
sbezverkkfox1111: what stopps other people not to trush your software like you do to helm?20:34
kfox1111and we unit test things.20:34
sdakethe code looks straightforward enough - have concerns of it turning into a frankensystem if helm doesn't grow the necessary sub-deploy features20:34
sdakesbezverk nothing really stops that20:34
sdakesbezverk either people trust until given reason not to, or dont trust to begin with :)20:35
sdakeis portdirect alive20:35
portdirecthi guys!20:35
sdakeportdirect are you following the conversation20:35
portdirect2 mins20:35
sdakek20:35
kfox1111sdake: sub-deploy is out I think for me.... its like building a clock, throwing 10x the number of gears in it, and only ever using a subset of them. sure, they probably won't mess things up.... but they can get conected accidently for numberious reasions.. better just to leave them out. less complicated.20:36
portdirectlots to parse20:36
kfox1111sdake: sbezverk: The primary differece beteen devs and ops is this20:36
sdakeportdirect take your time :)20:36
kfox1111devs are in the job of making the impossible possible. therefore, they tend to be optimists.20:36
sdakei'm a dev and i'm definately a pragmatist20:36
kfox1111ops are tankless. they are either in 1 of 2 states. they are doing their job well, so nothings broken, and then people ask, "why do we need an op? Things are working great"20:37
kfox1111or, the system is broken, and you are the badguy.20:37
sdakeok so lets stick on this sub-deploy idea20:37
sdakeyou mention its out for you - so that means if helm ever implements subdeploy, that won't be an option we can use?20:37
sbezverkgents I have meeting for 2 hours. Hopefully you will not the decision yet :-)20:37
kfox1111ops, trying to make the best out of the bad situation try hard not to get noticed, so try really hard to figure out what twill break things and avoid it.20:37
sbezverkI meant take the decision..20:38
kfox1111sdake: I'm pretty woried about that, yeah.20:38
sdakekfox1111 so to take that to its logical conclusion we are going to be stuck with a custom build tool forever?20:38
kfox1111sdake: no, we just need to ensure that the functionality of our custom build tool makes it into helm.20:39
sayantan_Hi guys, I am trying to work on a multinode deployment using kolla-ansible. I am getting an error when I execute tools/build.py -p default20:39
sayantan_Traceback (most recent call last):20:39
sayantan_  File "tools/build.py", line 39, in <module>20:39
sayantan_    sys.exit(main())20:39
sbezverkno until kfox1111 starts trusting helm ;-)20:39
kfox1111the features that are lacking get implemneted.20:39
sayantan_  File "tools/build.py", line 30, in main20:39
sayantan_    statuses = build.run_build()20:39
sayantan_  File "/root/kolla-ansible/kolla/image/build.py", line 960, in run_build20:39
sayantan_    kolla.create_dockerfiles()20:39
kfox1111there are 2 as far as I can tell.20:39
sayantan_  File "/root/kolla-ansible/kolla/image/build.py", line 693, in create_dockerfiles20:39
sdakesayantan_ use a paste service plz20:39
sayantan_    content = template.render(values)20:39
sayantan_  File "/usr/local/lib/python2.7/dist-packages/jinja2/environment.py", line 989, in render20:39
sayantan_    return self.environment.handle_exception(exc_info, True)20:39
sayantan_  File "/usr/local/lib/python2.7/dist-packages/jinja2/environment.py", line 754, in handle_exception20:39
sayantan_    reraise(exc_type, exc_value, tb)20:39
sayantan_  File "/tmp/kolla-2016-11-28_20-37-04_quyOIE/docker/base/Dockerfile.j2", line 6, in top-level template code20:39
sayantan_    {% import "macros.j2" as macros with context %}20:39
sayantan_  File "/usr/local/lib/python2.7/dist-packages/jinja2/loaders.py", line 187, in get_source20:39
kfox11111. common packages.20:39
sayantan_    raise TemplateNotFound(template)20:39
sayantan_jinja2.exceptions.TemplateNotFound: macros.j220:39
sdakesayantan_ i can tell you your not going to get an answer with a spam of 30 lines int he channel :)20:39
kfox11112. templating for Chart/Variables.20:39
sayantan_How do I do that? I am new to IRC20:40
kfox1111sbezverk: you assume I don't think like other operators. ;)20:40
sdakepaste your stuff into paste.openstack.org20:40
sdakeit prints out a link20:40
sdakeput the link in irc20:40
kfox1111sbezverk: in a op world, trust is erned.20:40
sayantan_Thanks :P20:41
sdakekfox1111 can you break out the commmon packages part a bit20:41
sbezverkkfox1111: on this matter no. I think if we selected a tool, we need to leverage it as much as we can, otherwise why we selected it at the first place20:41
kfox1111sbezverk: because it solves real problem. just not every problem. dont use a hammer to screw in a nail.20:41
sdakeok so - lets assume we use this custom build tool20:41
kfox1111sdake: talk more about it?20:41
sdakequestion20:41
sdakewill we be able to deploy from one master chart?20:42
sdakeand also deploy from subcharts?20:42
kfox1111sdake: yes. I proof of concepted one master chart deploy and individual deploys.20:42
sdakehow is that different then having one package that has a subdeploy as a feature?20:42
sayantan_http://paste.openstack.org/raw/590710/20:42
kfox1111sdake: there is no risk of nova update nova-novnc-proxy messing with nova-api if they are seperate packages.20:43
portdirectsdake: this is the question I have - I think we need to see how helm implemnts the subdeploy feature20:43
kfox1111well, or should I say, drastically reduced.20:43
*** gfidente|afk has quit IRC20:43
sdakei think portdirect is spot on here20:44
kfox1111helm is really doing someting weird when it comes to package managers.20:44
portdirectkfox1111: unless I've missed somthing helm is essentially just a templating engine that keeps track of the templates its thrown at the k8s api?20:44
sdakeits doing bundled libraries in essence (in the rpm world)20:44
kfox1111it is mixing package management, (yum like) with orchestration (puppet, deploy complex applications using packages)20:44
kfox1111its considerd bad form in the rpm world to make one rpm that goes and rpm -Uvh's a bunch of other rpms.20:45
kfox1111(after many years of blood due to peple trying it and getting hurt. :/ )20:45
kfox1111but helm's doing that. deploy one super package, and it deployos lots of little packages.20:45
kfox1111I get the desire to do it, but to me, thats a strong antipattern. I'm ok supporting those that want to do it that way,20:46
kfox1111but I have no desire to get bit by that way of doing things. seen it go horribly wrong too many times.20:46
sdakeback on portdirect 's point, if subdeploy only affects the subdeployed package, then the argument that orchestration is involved doesn't exist?20:46
*** rhallisey has quit IRC20:46
*** aric49 has quit IRC20:46
sdakekfox1111 ^^20:46
kfox1111sdake: I'm trying to come up with an analogy that can explain that...20:47
kfox1111its basically telling the orchestration system to try and ignore parts of itself.20:47
kfox1111"lets make the complicated orchestration system more complicated to try and get it to do something simple, and hope it doesn't get in the way"20:47
kfox1111I just want to do the basic thing without any of the complexity in the first place.20:48
sbezverkkfox1111: not orchestration but rendering system20:48
kfox1111what helm was originally intended to do.20:48
kfox1111here's a simple package, load it up for me.20:48
portdirectkfox1111: I think I see where you are coming from, but i think this is a matter of perspective and how we build/layout the packages20:48
kfox1111it grew the idea of superpackages later, when someone thought that was a good idea. :/20:48
kfox1111portdirect: I think those are two seperate issues.20:48
sdakewell we do need subdeploy20:49
kfox1111portdirect: build/layout of the files in the packages20:49
sdakethat is an essnetial feature of our architecture20:49
kfox1111sdake: I think we're not matching up on terminology.20:49
sdakelet me defifne subdeploy then20:49
portdirectbear with me here, it will take me a min to get this out20:49
sdakesubdeploy is something like helm deploy openstack --sub-package=nova-novncproxy20:50
kfox1111sdake: that is one way to do it.20:50
kfox1111the other is just helm deploy nova-novncporxy20:50
kfox1111helm deploy openstack --sub-package=nova-novncproxy makes me nervious.  helm deploy nova-novncporxy does not.20:50
sdakeok so boiled down to essence these are the two options?20:50
portdirecti think that it may be possible to get the best of both worlds using a lightweight build tool - one that build's a super package and another that builds micro-pakages for each service20:51
kfox1111there is no reason, you can't do a stand alone nova-novncproxy that can get bundled into a package 'openstack'20:51
kfox1111so that 'helm deploy openstack' works too.20:51
portdirectas the superpackae is essentially just composed of the micropackages20:51
kfox1111portdirect: yeah. thats what mine does actaully.20:51
kfox1111portdirect: it build the micropackages. the super packages pretty well normal helm packages. would just have dependencies on the micropackages.20:52
sdakeok i guess i'm good with that model - if the build system doesn't spiral into a frankensystem20:52
kfox1111the tradeoff is, where is the complexity. at build time, where it can be easily tested afterwards, or instantiation time.20:53
sdakekfox1111 however the patch you submitted does not follow the "slew of packages" lets build them into a metapackage20:53
kfox1111I'd rather see a bit more pushed towards build time. thats all.20:53
portdirectmy issue with your method is with the directory structure - as what I'm proposing I think cloud be achived in about 30 lines of bash - if we went with a similar layout to sbezverk's20:53
sdakeright directory structure20:53
sdakeexactly my problem too :)20:53
kfox1111sdake: portdirect that could be. I tried something like that initially, but was harder then it looked when I got into it.20:54
kfox1111I'm not tied to structure.20:54
kfox1111it can be anything that works.20:54
kfox1111I kind of like ryan's idea of making a templates subdir.20:54
portdirectso looking at: https://review.openstack.org/#/c/401394/20:55
portdirectif we symlinked the _helpers.tpl into the right place for each set of templates, and then added the appropriate values (via som method, to each subchart) we could then spit out a bunch of packages, or leave it as-is and spit out a single meta-package - unless I'm mistaken20:56
kfox1111portdirect: I reversed the flow, and did it with a common package.20:57
kfox1111put the shared library in a single package, then made it a dep for each microservice.20:57
kfox1111it worked well, except that vars would have to be done as globals, which I think is problematic for other reasons.20:57
sdakeshared library seems like a premature optimizatio nto me20:57
kfox1111so I wroked around that with the common-vars stuff being merged into vars.20:57
sdakewhich is a super anti-pattern in softwar eengineering20:57
sbezverkkfox1111: what you did with common it looks more like a workaround solution than the right use of tool..20:57
kfox1111sdake: we've had a shared lib for a while now. it has sped thigns up significantly.20:57
kfox1111sbezverk: quite possibly. they really need a concept of a lib/common package.20:58
kfox1111not a sub program package.20:58
sdakesped up development?20:58
kfox1111sdake: we didn't have to keep updating dozens of files to add common things.20:58
kfox1111and dealing with the review of all of that.20:58
sdakeok kfox1111 - can you write down in the whiteboard what exactly helm needs with exquisite detail from your pov20:59
sdakei guess this build system will have to do - with a resorted directory structure20:59
sdakei think each subpackage should have a consistent directory structure that is used in helm proper21:00
sdakee.g. nova-novncproxy21:00
sdakekfox1111 roger - apologies for not being able to focus any attention on kolla-kubernetes during my ptl-ship - it was literally a 20 hr a day 7 day a week job just to stay on top of the kolla repo21:01
kfox1111sdake: so, what I think the build should do, and what helm should do, or staring at the goals I stated in the review, and showing which peices solve which problems?21:01
kfox1111sdake: no worries. we're all busy.21:02
sdakekfox1111 too many statements in your question for me to parse properly could you restate?21:03
kfox1111sdake: "ok kfox1111 - can you write down in the whiteboard what exactly helm needs with exquisite  detail from your pov21:03
kfox1111"21:03
kfox1111I'm having trouble parsing that.21:03
sdakeok above in the scrollback you listed 2 things21:03
kfox1111"what exactly helm needs"21:03
kfox1111helm's a tool, without needs. I'm confused.21:03
sdakewhat we need exactly from helm21:04
sdakewe have needs of helm that it doesn't fulfill21:04
kfox1111ah.21:04
kfox1111gotcha.21:04
kfox1111that, I can do. :)21:05
sdakewhiteboard section of launchpad on the blueprint21:05
sdakeso i think where we stand re helm packaging is we need the following21:05
sdakea directory that contains pristine helm charts for each single microservice21:05
kfox1111I don't think we can make prestine helm charts without some templating.21:06
kfox1111which then arn't exactly helm charts.21:06
sdakeand meta packages which are built21:06
sdakeok lets tackle templating next up21:06
kfox1111k21:06
sdakeso pristine helm charts21:06
sdakethen some type of chart that packages them into services, like nova, neutorn, etc21:07
sdakethen some type of chart that packages all the subpackages into one big master chart21:07
sdakethen eveyrone gets what they want and the architecture is consistent with the implementation21:07
kfox1111right. s/chart/package/ on the your first line and we're in agreement.21:07
kfox1111how the micropackages are built is my concern. the rest I think are fairlyl standard for helm.21:08
sdakechart package I dunno the difference :)21:08
kfox1111package is a binary tarball looking thing.21:08
kfox1111chart is... source for it I'd say.21:08
kfox1111the dir thing you "helm package foo"21:08
sdakei think we want to work from source to build the packages21:09
portdirectright - so now we are back to templating :)21:09
kfox1111right. templating.21:09
sdakeok lets make sbezverk is in agreement :)21:09
sdakemake sure that is21:09
sdakehe said he would be on a call for a bit21:09
portdirectyeah - untill he is onboard this is just discussion21:09
v1k0d3ni see debates.21:10
v1k0d3n:)21:10
v1k0d3nhi guys21:10
portdirectwell hello.21:10
v1k0d3nlol21:10
sdakev1k0d3n there ya go :)21:10
kfox1111v1k0d3n: hey. :)21:10
kfox1111v1k0d3n: no worries. theres a few details about helm to nail down, and we're working through them. :)21:11
v1k0d3ncool cool21:11
sdakeok kfox1111 this is where we put into the review the log of the irc commentary21:11
sdakehttp://eavesdrop.openstack.org/irclogs/%23openstack-kolla/%23openstack-kolla.2016-11-28.log.html#t2016-11-28T20:13:1321:11
sdakeso people understand the high bandwidth conversation thta just took place21:12
kfox1111sdake: looks good. thanks.21:12
sdakewhen they look at the reviews later21:12
portdirectv1k0d3n: if you have any input it would be appricated21:12
sdakelets wait for sbezverk to get off his call before tackling templating21:12
portdirect+121:13
kfox1111yeah. if it weren't for the templating, he's right, we should just use helm natively to build everyting.21:13
kfox1111so that part needs to be discussed and agreed on.21:13
sdakekfox1111 that doesn't handle your subdeploy case21:13
kfox1111(main deploy! :)21:14
kfox1111well, the overarching packages can be built in the normal helm way I htink. with 'dependencies' on the subpackages.21:15
kfox1111I did it that way when testing the poc.21:15
v1k0d3nhave you guys asked alanmeadows?21:16
*** bmace has quit IRC21:16
kfox1111v1k0d3n: ask what?21:16
*** bmace has joined #openstack-kolla21:16
v1k0d3nhe's been working out some things on our end which we believe are extremely flexible.21:16
v1k0d3nkfox1111 "portdirect: v1k0d3n: if you have any input it would be appricated"21:16
kfox1111ah.21:17
portdirectkfox1111: they have been working with helm for a bit21:17
sdakev1k0d3n i think the best method there is for alanmeadows to read the two reviews, and the irc log i linked above21:17
sdakeso he has context for the conversation21:17
sdakeso we don't have to catch him up again21:18
* alanmeadows scrolls up21:18
sdakealanmeadows it begins here: http://eavesdrop.openstack.org/irclogs/%23openstack-kolla/%23openstack-kolla.2016-11-28.log.html#t2016-11-28T20:13:1321:18
kfox1111yeah. helm's looking pretty nice. I'm totally sold on it for package management/instantiation. I'm just very sceptical of it for orchestrating large apps.21:18
sdakealanmeadows recommend reading reviews first21:18
kfox1111I just want to be able to skip that part when desired.21:18
sdakekfox1111 i'm tired of clicking any chance you can link the two reviews for alanmeadows21:19
portdirecthttps://review.openstack.org/#/c/401394/21:19
alanmeadowsif this re: uber chart, subchart, etc21:19
sdakethanks portdirect21:19
kfox1111https://review.openstack.org/#/c/396296/ is the other.21:19
alanmeadowsand indeed it appears so :)21:19
portdirecthttps://review.openstack.org/#/c/401394/21:19
sdakethanks kfox111121:19
alanmeadowslet me take a look21:19
alanmeadowsbut on the whole one of the challenging things21:20
portdirectlol my bad21:20
alanmeadowsis the SAP approach is if you dont use the uber chart, too bad, you cant play21:20
alanmeadowsand I'm not really sure I like that21:20
kfox1111alanmeadows: exactly. :/21:20
alanmeadowsits especially painful during development21:20
kfox1111thats what I've been trying to avoid. :/21:20
alanmeadowsso the way we're architecting aic-helm PoC is each chart is always install-able individually21:20
alanmeadowsbut this creates some problems21:21
alanmeadows(you can still use the uber chart, but its a convenience, not a requirement)21:21
kfox1111cool. so aic-helm sounds like its going in the same direction I'm wanting to go.21:21
kfox1111what issues have you hit?21:22
*** kristia__ has joined #openstack-kolla21:22
alanmeadowsthis requires the implementation of some kind of helper though which is now not exactly 100% helm, but its a stop gap in my mind21:22
* kfox1111 nods21:22
kfox1111yeah.21:22
*** kristian__ has quit IRC21:22
alanmeadowsimagine something like your FQDN21:22
alanmeadowsyou want that global everywhere21:22
alanmeadowsbut you dont want to repeat it in 30 values.yaml files21:23
sdakeimo lets solve that iteratively21:24
alanmeadowsan uber chart could solve that but again, back to the uber chart being a requirement21:25
alanmeadowsso our stop gap right now until helm can solve this (i've got an issue creation in my todo)21:25
alanmeadowsis Make magic shoving a "_common.tpl" into every subchart21:25
kfox1111alanmeadows: basically what I did too.21:25
alanmeadowsa pre-build step... not 100% native but not sure what else to do right now, I hate to give up on the dream of independent charts21:26
*** fguillot has quit IRC21:26
kfox1111I cheeted a little more, and put it in a subchart. if wrapped in a macro, it shows up in the parent package when instantiated.21:26
portdirectalanmeadows: I'd come to a similar conclusiion that this is our only option right now21:26
kfox1111alanmeadows: so, please review https://review.openstack.org/#/c/396296/ and tell me what you think.21:26
kfox1111it seems similar to your approach.21:26
alanmeadowskfox1111: started there too, I am trying to remember what issue was hit that ended up with the even more hacky _common.tpl insertion....21:27
kfox1111the directory layout is a bit confusing/suboptimal. as its just a poc.21:27
alanmeadowsbut that is definitely more desirable21:27
alanmeadowsas at least its all helm21:27
kfox1111alanmeadows: ah. I'd do the same as you, if I bumped into something.21:27
kfox1111easy enough to just copy it into the package itself rather then a subpackage.21:28
alanmeadowsyeah just removes some of the magic ;-)21:28
*** kristia__ is now known as kristian__21:28
kfox1111we can even change it now to avoid some pain later.21:28
kfox1111I'm not tied to it. was curious if it would wrok at all in helm, and surprisingly it did.21:29
alanmeadowsso if memory serves defines in the "helper" subchart will bubble up21:31
alanmeadowsbut globals wont21:31
kfox1111right.21:31
*** lrensing has quit IRC21:32
alanmeadowswhich was depressing21:32
kfox1111vars defined in the template can't buble back to the calling template. :/21:32
alanmeadowswhich makes sense21:32
kfox1111heh. yeah.21:32
alanmeadowsbut it makes it messy21:32
kfox1111I had the same response. :)21:32
*** schwicht has quit IRC21:32
kfox1111would be nice if you could return a hash.21:32
kfox1111I guess I did end up solving it the same way you did...21:32
kfox1111I have a main that's copied to every package...21:33
*** sdake_ has joined #openstack-kolla21:33
kfox1111https://review.openstack.org/#/c/396296/27/helm/src/common_main.yaml21:33
alanmeadowsI can't find it in my history I think what I hit21:33
kfox1111which then calls the template from the package.21:33
alanmeadowswas literally inconsistency21:33
alanmeadowsin other words a define in a child21:33
alanmeadowswould work *some of the time*21:34
kfox1111yeah. I saw that too.21:34
alanmeadowsbest exampled21:34
alanmeadowsby running dry-run --debug | grep ...21:34
alanmeadowsand it was hit or miss21:34
alanmeadowsso I shelved it21:34
kfox1111if you go up a level, and make super package that depends on 2 subpackages and they both depend on common, and you try and shre some template filenames, it can get ugly. :/21:34
kfox1111sry. macronames21:35
kfox1111so a main macro in subpackage a, and one in b,21:35
kfox1111you might get an instantiation of a & b, a & a, or a & b.21:35
*** Jeffrey4l has quit IRC21:35
*** sdake has quit IRC21:35
kfox1111you might get an instantiation of a & b, a & a, or b & b.21:35
*** YefimG has joined #openstack-kolla21:35
kfox1111they definitely are doing some kind of random hashing on template load ordering.21:36
alanmeadowsso we now have an openstack-base, which I'll be renaming to something more generic as this also applies to ceph, maas, etc. to just "common" or whatever and its only purpose is to be included everywhere for some other helpful things, like ceph secret installation but also _common.tpl generation21:36
kfox1111which they do note in the docs21:36
*** schwicht has joined #openstack-kolla21:36
alanmeadowsI can live with the workaround because I think we can pressure helm to come up with a solution for this21:37
alanmeadowsI have great concerns21:37
kfox1111+121:37
alanmeadowsabout the uber chart and future upgrades21:37
kfox1111me too.21:37
alanmeadowsif you break it or something goes wrong21:38
sdake_you can't pressure upstreams into doing anything they don't want to do fwiw :)21:38
alanmeadowsyou  need to what, redo everything including mariadb, etc?21:38
kfox1111sdake_: as in, add development effort, or shine a light on a use case not covered.21:38
alanmeadowssdake_: you havent seen how fast I can type into separate message windows!21:38
kfox1111alanmeadows: exactly. scares me as an op.21:38
sdake_alanmeadows :)21:38
alanmeadowsalso21:39
alanmeadowsthe uberchart breaks the potential21:39
alanmeadowsfor namespace chunking21:39
kfox1111alanmeadows: seems like we've hit convergent evolution. :)21:39
alanmeadowsas far as I can see at least21:39
kfox1111right.21:39
YefimGHi, guys! Trying to install ceilometer with gnocchi backend, but having Ceilometer bootstrap container failed: Container exited with non-zero return code. Is that even a supported configuration?21:39
sdake_YefimG your looking for jeffrey4l21:40
kfox1111I've got multiple regions and multiple host aggregates too,21:40
sdake_YefimG he is in apac21:40
kfox1111which I think will fit poorly in the uberchart model.21:40
kfox1111(oh, and dvr)21:40
kfox1111multiple types of l3 agents, etc.21:40
kfox1111I don't think helm will support a 1 super package, multiple instances of a sub package use case well, or soon for a while.21:41
kfox1111and openstack really needs it for non trivial clouds from my experience.21:41
sdake_who knows what helm will support21:41
sdake_we follow our upstreams not visa versa :)21:41
akwasniesdake: pbourke: here is the dashboard from summit http://paste.openstack.org/show/590718/21:41
kfox1111sdake_: sure. just saying its hard, so hard things usually take longer then easy things. so unlikely soon. :)21:41
sdake_this is why dependency management is so important :)21:41
jrich523hey guys, i was just using the openstack client, was going to try and look up a user (user ID on an image) and im getting a 401 (needs auth) which is odd because im able to run other things... i was wondering if i needed to set a domain context like i do on the web ui21:42
akwasniesdake: pbourke: sorry for late notice, I forgot I had it on another laptop :)21:42
jrich523i dont know what you're talking about sdake_ but i couldnt agree more :)21:42
sdake_jrich523 did you create your admin-oepnrc.sh via kolla post-deploy?21:43
jrich523yeah the source'ing works fine21:43
YefimGsdake_: Thanks!21:43
jrich523i can query images, vms etc21:43
jrich523its just with Users, which happens to be ldap so i know that also makes things funky21:43
sdake_YefimG no problem - ffwiw it is supported in 3.0.1 - doesn't mean it works for you tho - jeffrey4l can sort out if its been fixed and awaiting a backport tag21:44
jrich523if the user list should just work, (no need for domain context or whatever) then its not the CLI and i'll go deal with that... just wanted to make sure i wasnt missing some basic step21:44
*** hvlad has joined #openstack-kolla21:44
alanmeadowskfox1111: https://github.com/alanmeadows/aic-helm/blob/master/Makefile#L4021:44
alanmeadowsthats my workaround for now21:45
sdake_jrich523 no experience with ldap and kolla21:45
alanmeadowsso here is my thought21:45
jrich523k, thanks21:45
alanmeadowsand I think SAP came to this conclusion too21:45
*** matrohon has quit IRC21:45
kfox1111ah. using make for it.21:45
jrich523so the short of it is, i shouldnt need to do anything like set context, it should just spew all users?21:46
kfox1111nice.21:46
alanmeadowsendpoint URLs get defined in something common like this, ports belong as globals per chart21:46
kfox1111do you template out version numbers too?21:46
alanmeadowsthat way you combine the two you can create an endpoint for any service21:46
alanmeadowsI do not at the moment21:46
kfox1111as an operator, I'd like to be able to build an updated set of containers, then build an updated set of packages, then do a minor rolling upgrade.21:46
kfox1111ah. k.21:46
*** Pavo has quit IRC21:47
* sdake_ dislikes agile fwiw kfox1111 :)21:47
jrich523you're certainly one of the few sdake_ :)21:47
*** Jeffrey4l has joined #openstack-kolla21:47
kfox1111sdake_: I have issues with agile too, btw. :)21:47
sdake_nah i prefer open source method21:47
jrich523Agile is good, agile is not... most places do agile21:47
kfox1111it works great, unless you have a team where your members are operators. ;)21:48
sdake_it works great unless your team is spread across the world21:48
kfox1111where their time is random due to machine uptime obligations.21:48
sdake_Agile requires collocationt o work effectively21:48
kfox1111then you have no idea how long anything will take, and it all falls apart.21:48
sdake_ya agile does provide capacity management - something sorely lacking from open source21:48
kfox1111guess that applies to opensource too. no idea how long anything wil take.21:48
kfox1111but, thats kind of the nature of opensource too.21:49
sdake_however open source is done by volunteers so capacity management in open source is impossible21:49
kfox1111no fixed capacity.21:49
kfox1111right.21:49
jrich523is there an "opensource" workflow?21:49
jrich523guess i've never looked, but would be neat to read about21:49
sdake_jrich523 there is - but it hasn't been written down ;)21:49
kfox1111jrich523: defacto. I don't know if its written down21:49
kfox1111heh21:49
jrich523lol, so no :)21:49
kfox1111:)21:49
*** inc0 has joined #openstack-kolla21:49
sdake_ok well that was entertaining lets wait for sbezverk to catch up with the log posted in each review21:50
inc0hello21:50
sdake_sup inc021:50
*** sdake_ is now known as sdake21:50
jrich523ticket based? open a ticket ("issue") with tasks, someone finishes it, sorta like agile, small sections of well defined work.....21:50
jrich523howdy inc021:50
v1k0d3nportdirect: hey man21:50
sdakejrich523 open source methodology scales to thousands of developers21:51
sdakejrich523 agile does not21:51
sdakethat is my main beef ;)21:51
jrich523lol fair point21:51
jrich523i think the location point is a stronger one21:51
inc0sdake, also doesn't really support any form of concrete goal;) everybody does what the hell they want;)21:51
jrich523its all about working in groups, team dev etc21:51
sdakeya that too :)21:51
*** Pavo has joined #openstack-kolla21:51
inc0which is cool21:51
jrich523sorta21:52
kfox1111jrich523: its desire based.21:52
sdakeinc0 thats called the "Hidden Hand"21:52
kfox1111if someone cares enough about something, they fix it, or they somehow convince a dev to fix it.21:52
jrich523i think thats a category on porn hub sdake lol21:52
kfox1111(monitarily or otherwise)21:52
inc0yeah, but this hidden hand can smack people only so hard21:52
*** jheroux has quit IRC21:52
alanmeadowskfox111: the other thing to develop is consistent values across projects ;-)21:52
alanmeadowsit can quickly get out of hand21:53
sdakeinc0 https://en.wikipedia.org/wiki/Invisible_hand21:53
kfox1111alanmeadows: this looks very similar to where I was going.21:53
alanmeadowshad to nip that early on which even with 4 projects was a big effort21:53
inc0and build culture of quality in community21:53
kfox1111alanmeadows: yeah. and governence. far to often ignored but makes or breaks an effort.21:53
inc0getting good culture = win21:53
inc0governance in OpenStack is very lightweight, probably that's the only reason it still works21:54
kfox1111yeah.21:55
sdakeok well while u guys debate which methodology is best this old man is gonna take a nap21:55
alanmeadowseither something about great minds or fools seldom differ21:55
kfox1111mostlly works... sort of...21:55
alanmeadowsone of those two21:55
sdake(development methodology that is)21:55
kfox1111alanmeadows: hehe. exactly. :)21:55
kfox1111alanmeadows: at least we have company in our madness, if nothing else. :)21:55
jrich523lol21:55
jrich523so you've got that going for you21:55
kfox1111its always conforting to find someone else doing things the same way through experience. it either shows you your are on the right path, or the universe is playing a bigger trick on you then you think. ;)21:56
kfox1111(and you then don't take all the blame. :)21:57
alanmeadowsyou can either take the credit or if it flops21:57
alanmeadowsjust say you copied this guy over here21:57
alanmeadows;-)21:57
* kfox1111 nods :)21:57
jrich523i'll be sure to blame you the next time i can alanmeadows :)21:57
jrich523alan did it!21:57
kfox1111sdake: I gota head out now for a couple hours. :/21:58
kfox1111meetings.21:59
alanmeadowskfox1111: when you get a moment21:59
kfox1111alanmeadows: whats up?22:00
alanmeadowsit might be nice to talk about your thinking wrt neutron components22:00
kfox1111yeah. that would be good. um...22:00
alanmeadowsI don't have a super strong preference but it does present two paths22:00
alanmeadowsdoes the operator control how to configure neutron via helm install <my list of comps>22:00
kfox1111I've got meetings until 4:30 pst though. :/22:00
alanmeadowsor helm install neutron --set ....22:00
alanmeadowsyeah no rush22:00
kfox1111been debating that with sbezverk.22:01
kfox1111I prefer helm install neutron-l3-agent-compute, neutron-l3-agent-network, etc.22:01
kfox1111same package, instantiated multiple times with different flag.22:01
kfox1111rrr, sorry.22:01
alanmeadowsand thats fine its just this is not an area I've put much thought into22:01
kfox1111elm install neutron-l3-agent --set type=compute22:01
alanmeadowsso I think understanding what drove that is good22:02
alanmeadowsI've not put much22:02
alanmeadowsrather22:02
kfox1111yeah. should do the same for host aggretates too.22:02
*** jtriley has joined #openstack-kolla22:02
alanmeadowsthe only thing that complicates that22:02
kfox1111helm install nova-compute --set type=intel-foo22:02
alanmeadowsis the installer has to have more knowledge22:02
kfox1111helm install nova-compute --set type=amd-bar22:02
alanmeadowsi suppose either way its the same22:03
kfox1111yeah.22:03
alanmeadowswhether with sets22:03
alanmeadowsor multiple installs22:03
inc0one with amd won't work22:03
kfox1111ok. really gota go. lets talk soon.22:03
alanmeadowsyeah lates22:03
inc0I22:03
inc0I'm from intel, I think I'm required to say that22:03
alanmeadowsbut that immediately raises suspicion22:03
alanmeadowsabout your amd assertions22:04
alanmeadows;-)22:04
*** lamt has quit IRC22:05
*** TxGirlGeek has quit IRC22:07
*** jtriley has quit IRC22:10
*** lamt has joined #openstack-kolla22:10
sdakeinc0 i agree one with amd wont work and i'm not from intel so not required to say that ;-)22:10
inc0uff thanks sdake you just saved last shreds of my credibility22:11
portdirectmuch better for heating your house though22:11
sdakeinc0 nah you got plenty22:11
*** TxGirlGeek has joined #openstack-kolla22:11
inc0srwilkers, ping22:13
v1k0d3nhey guys...22:16
inc0wassup v1k0d3n22:17
v1k0d3nnot sure it's totally clear to me yet, but does kolla-ansible work just fine on ubuntu 16.04? i'm hearing mixed things...22:17
inc0v1k0d3n, I'm running 16.04 all the time22:17
inc0if it doesn't I'd like to know, it should.22:18
*** goldyfruit has quit IRC22:18
sdakev1k0d3n if you mean 3.0.1 that was definately tested22:18
sdakemaster is passing the gate - as enemic as it is22:18
inc0we run xenial inside container22:18
v1k0d3noh...wonder why i hear mixed things then.22:19
*** TxGirlGeek has quit IRC22:19
v1k0d3nso just follow instructions on ubuntu with systemd.22:19
v1k0d3nok22:19
inc0yeah, 2.0 was based on trusty, if they run mitaka it might be funky22:19
sdakedepends what mixed tings you hear22:19
sdakefor example we think gpu passthrough is busted on ubuntu22:19
inc0v1k0d3n, or use bootstrap-servers22:19
*** TxGirlGeek has joined #openstack-kolla22:20
sdakeJeffrey4l another report of problem with gnocchi and ceilometer22:20
sdakeJeffrey4l i never tested that code - does it work for you?22:20
*** jrist has joined #openstack-kolla22:20
v1k0d3ninc0: good to know. i will play around again some more later and verify this. thanks for the heads up guys.22:21
inc0v1k0d3n, sure, by all means it should be good. let us know what issues people see22:21
inc0also, I have question22:21
inc0since you're around22:21
*** jheroux has joined #openstack-kolla22:22
inc0v1k0d3n, have you guys seen work sbezverk and kfox1111 done with helm?22:22
*** jascott1 has quit IRC22:22
*** jheroux has quit IRC22:24
*** jascott1 has joined #openstack-kolla22:24
v1k0d3nnot really. hard to keep up and just back from vacation.22:25
*** TxGirlGeek has quit IRC22:25
inc0kk, check these out plz22:25
inc0let's get stuff merged and working together in kolla-k8s asap plz22:25
v1k0d3ni can check out.22:25
v1k0d3nbit stretched thin...but first priority for us 100% is our PoC.22:26
v1k0d3nit's been that way all along for us.22:26
*** jascott1 has quit IRC22:26
*** jascott1 has joined #openstack-kolla22:26
inc0fair enough, let's just keep in mind that we'll want to have single project at the end of the day22:27
v1k0d3nboss makes the calls. extremely happy to merge though.22:27
*** sayantan_ has quit IRC22:27
v1k0d3ni find myself in this place a lot.22:27
v1k0d3nthe intent is only to have one project when this is over.....22:28
sdakeportdirect do you have that irc log - it wasn't posted in either review22:28
v1k0d3nsaying this in the clear for everyone to see :)22:28
sdakeportdirect or must i hunt down the irc log myself22:28
sdakeyet again...22:28
inc0cool, that's what I'm saying, you guys should keep track on kolla-k8s community too22:28
portdirectsdake: this one ? http://eavesdrop.openstack.org/irclogs/%23openstack-kolla/%23openstack-kolla.2016-11-28.log.html#t2016-11-28T20:13:1322:29
*** zigo has quit IRC22:29
inc0to make sure that 1. your poc is as good as it gets and 2. we won't close any doors by omission by agreeing on some approach22:29
sdakeportdirect mind posting those in the two reviews in question22:29
portdirectnp22:29
inc0or at least if we do - we do it counciously22:29
*** jascott1 has quit IRC22:31
*** zigo has joined #openstack-kolla22:36
*** kristian__ has quit IRC22:36
* sdake requests the anestaphine from the medlock22:42
sdakealanmeadows fwiw the makefile approach works, but not keen on that - main reason being we shouldn't have to run make in a pip installed package to get things to work22:43
sdakei know your working on a poc so this problem doesn't apply for you22:43
sdakehunting down /usr/share/kolla-kubernetes (which is read only) and running make - not ideal :)22:44
alanmeadowswell the main thing is deciding on whether there needs to be some "helpful" hand in the interim22:44
alanmeadowsMake is just a means to an end22:45
alanmeadowsin our case22:45
sdakeright22:45
sdakejust sharing information :)22:45
sdakeas you did22:45
*** Pavo has quit IRC22:46
*** vhosakot_ has joined #openstack-kolla22:48
*** vhosakot has quit IRC22:50
*** mnasiadka has quit IRC22:51
*** jascott1 has joined #openstack-kolla22:53
*** v1k0d3n has quit IRC22:54
*** v1k0d3n has joined #openstack-kolla22:54
*** Pavo has joined #openstack-kolla22:55
*** hvlad has quit IRC22:55
*** sayantan_ has joined #openstack-kolla22:56
*** sayantan_ has quit IRC22:56
*** sayantan_ has joined #openstack-kolla22:56
*** Pavo has quit IRC22:57
sbezverkalanmeadows: question for you, if helm would have a way to render charts at each level would you use it or still use the workaround way?22:58
alanmeadowswould definitely prefer an entirely native helm approach22:58
*** v1k0d3n has quit IRC22:59
*** Pavo has joined #openstack-kolla22:59
*** TxGirlGeek has joined #openstack-kolla23:00
sbezverkalanmeadows: thank you,in this case, the aproach we choose now should not force us to use it forever and when this feature is available, and I suspect it will be sooner rather than later, it should allow people to use either one depending on their goals/preferences..23:00
*** ipsecguy has quit IRC23:03
alanmeadowssbezverk: yup, agreed23:03
*** sayantan_ has quit IRC23:06
*** jascott1 has quit IRC23:06
*** saneax-_-|AFK is now known as saneax23:09
*** jascott1 has joined #openstack-kolla23:10
*** YefimG has quit IRC23:10
*** sayantan_ has joined #openstack-kolla23:13
kfox1111back for about 5 min.23:13
kfox1111sbezverk: like I said, I'm not aposed to using native helm when it has all the needed features. its just not there yet.23:13
kfox1111I'd prefer it actually.23:13
kfox1111I just don't think we should wait for it.23:14
sbezverkkfox1111: maybe so, but again current approach breaks any possibility of using them in future!23:14
kfox1111sbezverk: I don't think so.23:14
sbezverksince helm has some rules about how things needs to be placed23:14
kfox1111sbezverk: reordering a file system is easy enough.23:14
kfox1111nothing precludes that.23:14
kfox1111git even tracks changes across a rename23:15
sbezverkwhy then use approach which will require in future reordering in stead of designing a solution that does not required it at the first place??23:15
kfox1111because package api trumps filesystem layout.23:15
kfox1111package api is not easily changable.23:16
kfox1111filesystem layout is.23:16
kfox1111and like I said, I don't really care about file system layout,23:16
kfox1111just that we can template all the thinkgs we need to template.23:16
kfox1111which helm doesn't support at build time yet.23:16
sdakei am leary of a huge git mv refactor23:17
sdakei think its better to just c&p all the templates around for now23:17
kfox1111sdake: its like 6 months out I think. :/ not sure its worth stressing about. :/23:17
*** lamt has quit IRC23:18
kfox1111it would happen after helm supports build time templating.23:18
kfox1111which isn't on their roadmap currently.23:18
kfox1111but we can work to get it there.23:18
kfox11112 min. :/23:18
portdirectonce we have a working implementation then we can make it DRY, I'm also wary of code restructuring23:20
kfox1111+1. really we could mirror our current existing structure as is, and restructure not when we go to helm, but when heml supports build templating. then move it only then.23:21
kfox1111might make it a little easier to review too, as you can just compare the two directory structures.23:22
kfox1111diff -ruN services tempates23:22
kfox1111ok. times up. be back later. :/23:23
jascott1sbezverk wrt to your helm plugin, why cant we just take the globals and subtree we want and pass to child chart? seems like a simple stop gap script until helm plugins are baked. is that what you are thiking?23:28
jascott1thinking even23:28
sbezverkjascott1: that is what I use currently23:28
sdakeportdirect ++23:29
sbezverkthere is just one thing missing is to build micro-service with global vaules defined in paert23:29
sbezverkoops parent23:29
sbezverkjascott1: I was told that this plugin will add this missing piece and then you should be able to build parent/service or micro-service independently but considering values.yaml from higher levels too.23:30
sdakesbezverk right - thats why i dont care if values.yaml is copied all over the place as long as each microservice is in a helm structured format23:31
sdakeif that feature comes, fantastic23:31
sdakeif ti doesn't, plan B23:31
sbezverksdake: agree as long as template format will be presereved and helm will be able to use them without any modification I am all up for plan B23:33
sdakeportdirect wfu?23:33
sdakewe have 150 microservices in the long haul to convert over23:33
sdakelets focus on the ones that matter first - and make them helm native23:34
jascott1jeebus!23:34
sdakejascott1 had a look at kolla container repo lately? :)23:34
jascott1i havent23:35
jascott1oh23:35
jascott1that huge list23:35
sdakeSTDAKE-M-26CJ:docker sdake$ find . -name \*.j2 -ls | wc -l23:35
sdake     18023:35
sdakerather 180 ;)23:35
portdirectsdake: wfm23:35
*** inc0 has quit IRC23:35
* jascott1 questions life choices23:36
*** JRobinson__ has joined #openstack-kolla23:36
jascott1starting to think I need k8s networking or complete idiots book23:38
jascott1or/for23:38
portdirectjascott1: AMA (there are no stupid questions..)23:38
*** saneax is now known as saneax-_-|AFK23:39
jascott1well I was going to try v1k0d3n and crew's work so I nuked my kargo based k8s deployment and tried to use kubadm so I could get k8s 1.5beta1 as they use. got the master and a node but cant even run a busybox image wihtout crash looping23:40
jascott1pretty sure its dns related, i see kube dns pod23:40
portdirectok - what vagrant backend are you using?23:41
jascott1i say its dns related because it was getting an error about not finding node-0 (worker)23:41
jascott1although its in the node list and everything23:41
portdirectare you using halycon - or kubeadm on nodes you've set up yourself?23:42
jascott1terraform for node deployment (just vms) and then kubeadm23:42
sdakeportdirect perhaps you can walk jascott1 thru a kubernetes vagrant deploy23:42
jascott1well I need to use kubeadm with 1.5beta123:43
portdirectjascott1: is that possible (vagrant)? or I can try and see whats going on with your current setup if you'd prefer23:43
portdirecteither way you should be able to install whatever version ok k8s you want23:44
*** sayantan_ has quit IRC23:44
sbezverkjascott1: can you show how you kubectl get pods --namespace=kube-system23:44
sbezverklooks like23:44
jascott1sure thanks for the help guys23:44
jascott1oh wait I was re-nuking, maybe someone can help with kubeadm install ? I can paste what I am doing23:45
sdakejascott1 why not use vagrant?23:45
sdakejascott1 thats the dev env we are staandaridzing on23:45
sbezverkjascott1: sure thing I really like kubeadm, it behaves nicely when supplied with right parameters :-)23:46
jascott1i was trying to go more prod-like basically23:46
sdakeare you dev or qa :)23:46
portdirectjascott1: I wouldnt use use kubeadm in prod23:46
jascott1before I did sw, now I just swing a sword at k8s it seems :P23:47
sbezverkportdirect: why, you hit any issues?23:47
jascott1it doesnt do some HA some missing logging?23:47
jascott1i read those shortcommings23:48
portdirectsbezverk: no HA, setting cluster domain doesnt work and I'd prefer to use a PKI i have control over.23:48
sbezverkjascott1: you can add to kubeadm built cluster by labeling additional nodes23:48
jascott1portdirect I thought it would be the new officially recommended way23:48
sbezverkportdirect: you can easily add HA after it is installed, dns to work require 1 change23:49
jascott1eventually being good enough for PR23:49
jascott1PRD23:49
portdirectjascott1: I think that it will be ready soon, but not yet i dont think23:49
sdakeok folks so we have agreement on the fact that we need individual microservices23:50
sbezverkportdirect: the only thing I did not try is to use custom PKI23:50
sdakenow on to the next thing, which is implementing all of those23:50
portdirectsbezverk: PKI is the big one for me - but then I do wear a tinfoil hat23:51
sdakewe can sort out the service level packaging and metapackage stuff next once we have a working compute kit in helm we can manually deploy as kfox1111 intends to do23:51
sbezverkportdirect: lol23:51
sdakesbezverk can you break out your patch into the microservice model - its closest to hte dir structure i think we desire23:52
portdirectsdake: I'm working on ceph right now, and then will hit keystone soon as we have a db23:52
sdakeso one patch on top of another23:52
jascott1http://paste.openstack.org/show/590729/23:52
sdakeportdirect db coming - been busy today - monday funday23:52
sbezverksdake: it is based on microservice model already. All I need is to get rid on common library and then each microservice can be deployed individually23:53
sdakecool so do each one as separate patches and remove the common lib23:54
sdakecommon stuff is not essential at this point in time23:54
sdakeif its common it can be copied around23:54
sdakeannoying but workable23:54
sdakethis is how docker repo was run as well :)23:54
sdakerather kolla's docker dir23:54
sdakewe went through about 6 iterations before we had it right ;)23:55
jascott1understandable goes a long way :)23:55
sdakewe can tckle common on o323:56
sbezverksdake: is it really required to do PS per microservice? I mean when common staff is gone then PS will be very small23:57
sdakemany of those iteratations were git mv23:57
sdakeyup that will maek it easier to review23:57
sdakefill review queue up ;)23:57
jascott1yay!23:58
« Sunday, 2016-11-27 Index Tuesday, 2016-11-29 »

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