sdake | like I said, one of our approaches was just like entrypoint except more crude | 00:00 |
---|---|---|
v1k0d3n | because in different ways, they solve the same issue. | 00:00 |
sdake | if you break down entrypoint it doe sthe following | 00:00 |
sdake | 1. contacts apis to make sure they are available | 00:00 |
sdake | 2. contacts openstack to make sure stuff that should be created has been | 00:00 |
sdake | 3. contacts other infrastructure to make sure stuff that should be created has been | 00:01 |
sdake | 4. blocks on 2 and 3 creating a completely self-orchestrating system | 00:01 |
sdake | there is still orchestration | 00:01 |
sdake | it is just different then the orchestration provided by ansible | 00:01 |
sdake | its decentralized | 00:01 |
sdake | when we implemented compose (AIO) we did 1, 2, 3, and 4 | 00:02 |
sdake | and it worked 19 times out of 20 | 00:02 |
sdake | the other 5% of the time it completely failed | 00:02 |
sdake | if you distill kubernetes to a single node and get rid of the dynamic scheduling part (because its AIO) we tried this model | 00:02 |
*** jtriley has joined #openstack-kolla | 00:02 | |
sdake | so i am not totally all that keen on entrypoint | 00:03 |
sdake | can you tell me what additional steps an operator takes? | 00:03 |
sdake | v1k0d3n the reason I ask is we write down specs for this stuff | 00:07 |
sdake | v1k0d3n i was planning to write a spec since once has been written for helm | 00:07 |
sdake | if an operator would solve the problem instead of helm, no need for a helm spec | 00:07 |
*** jtriley has quit IRC | 00:08 | |
*** jtriley has joined #openstack-kolla | 00:08 | |
v1k0d3n | there is still a need for helm | 00:11 |
v1k0d3n | helm is an installer | 00:12 |
v1k0d3n | i can explain a bit later | 00:12 |
*** jtriley has quit IRC | 00:13 | |
sdake | v1k0d3n ok i can buy that | 00:14 |
sdake | can you answer me a Q, does the etcd container in some way integrate with the operator? | 00:14 |
sdake | reading the etcd-operator code, I see a bit how it works | 00:14 |
sdake | but it is unclear what changes are needed to containers | 00:14 |
sdake | that is the crux of the information I want from you :) | 00:15 |
sdake | etcd-operator is an orchestration engine for etcd | 00:15 |
v1k0d3n | watch this video from lachie: https://www.youtube.com/watch?v=n4GYyo1V3wY | 00:15 |
sdake | custom orchestration | 00:15 |
sdake | ok will watch - does it answer my Q about changes needed to containers? | 00:15 |
sdake | i'd be somewhat interested in the custom orchestration model for kolla-kubernetes | 00:16 |
sdake | that is what an operator is imo | 00:16 |
sdake | "operationally manage" | 00:18 |
sdake | application speciic controller | 00:18 |
sdake | unclear if etcd container requires any changes | 00:18 |
sdake | v1k0d3n i've looked at that video now | 00:26 |
sdake | v1k0d3n it really looks to me like a completely customized orchestration engine specifically for etcd | 00:26 |
sdake | and looking at the codebase, looks like same story | 00:26 |
sdake | so the answer is no changes are required in the containers themselves nor the API | 00:27 |
*** gfidente has joined #openstack-kolla | 00:33 | |
*** gfidente has joined #openstack-kolla | 00:33 | |
*** yingjun has joined #openstack-kolla | 00:38 | |
v1k0d3n | sdake: that's right. it's an operator for etcd. | 00:59 |
*** HyperJohnGraham has joined #openstack-kolla | 01:00 | |
v1k0d3n | the framework could be the same for any of the openstack services that require the same level of "operator" support. | 01:00 |
*** tovin07 has joined #openstack-kolla | 01:00 | |
v1k0d3n | and no codebase or container changes needed in this case. | 01:00 |
sdake | which framework, etcd? | 01:00 |
v1k0d3n | operator for etcd. | 01:00 |
sdake | the operator for etcd wont just work for kolla ;) | 01:00 |
v1k0d3n | there is one for prometheus as well, that shows a general architecture. | 01:00 |
sdake | its tuned specficially to etcd | 01:00 |
v1k0d3n | right...we'd have to write it. | 01:00 |
v1k0d3n | but only for a few services. | 01:01 |
v1k0d3n | let me put it this way... | 01:01 |
v1k0d3n | for what we need...at our scale...we're going to have to do this. | 01:01 |
sdake | right - so this is a completely different approach in comparison to entrypoint | 01:01 |
v1k0d3n | if kolla uses or not, we need it. | 01:01 |
v1k0d3n | it is. | 01:01 |
v1k0d3n | not saying they are the same; i think that's causing confusion. | 01:01 |
sdake | why only for a few services? | 01:01 |
sdake | the confusion is cleared up now | 01:01 |
*** duonghq has joined #openstack-kolla | 01:02 | |
sdake | ou were using them interchangably | 01:02 |
sdake | but I read the code of both and understand how both work now | 01:02 |
sdake | one is container internal orchestraiton, the other is container external orchestration | 01:02 |
v1k0d3n | no. just different solutions that solve what results in one aspect....dependency handling. | 01:02 |
v1k0d3n | right | 01:02 |
sdake | a kolla-operator tool would be very similar in nature to ansible just tuned around openstack | 01:03 |
v1k0d3n | turned around openstack = correct | 01:03 |
v1k0d3n | that's what i am thinking too | 01:03 |
v1k0d3n | same page | 01:03 |
v1k0d3n | but decoupled, just to be clear | 01:03 |
v1k0d3n | so operator per service. | 01:03 |
sdake | operator for nova, operator for heat, etc? | 01:04 |
v1k0d3n | no, those are api based and can pretty much be plugged in at any point. | 01:04 |
v1k0d3n | rabbit, db....those are the first to get an operator. | 01:04 |
sdake | what operators need be written | 01:04 |
v1k0d3n | memcached | 01:04 |
v1k0d3n | those three... | 01:04 |
v1k0d3n | i think the rest is ok...but need to explore. | 01:05 |
sdake | what about openstack itself? | 01:05 |
v1k0d3n | microservices-based. i would not write for whole stack. just components that can be reused. | 01:05 |
v1k0d3n | operator should be service-specific not end-state application specific i believe... | 01:06 |
sdake | what operates openstack? | 01:06 |
v1k0d3n | i was going to talk to coreos about this at kubecon. | 01:06 |
v1k0d3n | get their thoughts | 01:06 |
sdake | the thing that operates openstack is kolla-operator | 01:06 |
sdake | if we want to make infra a separate thing, that seems reasonable | 01:06 |
v1k0d3n | well, perhaps...but from what i have been gathered from others so far....service-based is the way to go. | 01:06 |
sdake | openstack services depend on each other | 01:07 |
sdake | for example, nova depends on glance and keystone | 01:07 |
v1k0d3n | then there would probably be a good usecase for it. | 01:08 |
v1k0d3n | you're right. | 01:08 |
sdake | usecase for what? | 01:08 |
sdake | my point is all of openstack needs an operator - its not like we can just have yaml files kicking around | 01:08 |
sdake | i mean thats helpful | 01:08 |
sdake | but not the end state | 01:08 |
v1k0d3n | for a nova or glance or openstack controller is what i'm saiyng | 01:08 |
sdake | is a controller the same thing as an operator? | 01:09 |
Pavo | now chat is active | 01:09 |
Pavo | lol | 01:09 |
v1k0d3n | yeah, but then the operator just becames openstack and not individual services. | 01:09 |
v1k0d3n | for instance, etcd operator just operates etcd. | 01:09 |
*** jtriley has joined #openstack-kolla | 01:09 | |
v1k0d3n | from what i have been hearing smallest atomic unit is good...repeatable. | 01:09 |
sdake | i hear where your coming from | 01:10 |
v1k0d3n | Pavo: only two folks lighting it up really :) | 01:10 |
sdake | this is what i thought helm would solve for us - operating nova for example ;) | 01:10 |
v1k0d3n | lol | 01:10 |
v1k0d3n | we need to talk | 01:10 |
Pavo | lol yeah well it was absolutely dead earlier | 01:10 |
v1k0d3n | helm is somewhat different | 01:10 |
sdake | we are talking | 01:10 |
v1k0d3n | together | 01:10 |
sdake | we are talking together | 01:10 |
v1k0d3n | need them together. | 01:10 |
v1k0d3n | helm is a catalog and some dependancy. operator is dependency after lifecycle as well...which helm does not do. | 01:11 |
sdake | right | 01:11 |
sdake | our option #1 (helm) is driving the ideas of operator into helm directly | 01:11 |
sdake | so instead of a million custom operators there is one - helm | 01:12 |
v1k0d3n | yes | 01:12 |
sdake | i have a hard time understanding why to break apart the operational dependencies of openstack (mariadb, rabbitmq, memcache) from the rest of openstack | 01:13 |
v1k0d3n | and helm can also do upgrade, rollback, keep (meaning a lock...can't be accidentally deleted), dependency for initial deployment...lot of good things. | 01:13 |
Pavo | what is this helm you are speaking of? | 01:13 |
sdake | pavo we are figuring that out as we speak ;) | 01:14 |
Pavo | lol | 01:14 |
*** jtriley has quit IRC | 01:14 | |
sdake | v1k0d3n you do see how operator is just a slimmed down helm, no? | 01:14 |
sdake | slimmed down in that it is a single use case | 01:15 |
v1k0d3n | i see them differently | 01:15 |
sdake | expand? | 01:16 |
v1k0d3n | i have a meeting with coreos during kubecon....want to join? | 01:16 |
sdake | to listen in sure | 01:16 |
v1k0d3n | it's 8....my wife is goig to kill me if i don't get off here, since i'm going to be gone all week. movie night :) | 01:16 |
sdake | i dont hvae much to contribute at this point | 01:16 |
sdake | roger | 01:16 |
sdake | ttyl | 01:16 |
sdake | send me invite pls | 01:16 |
sdake | i'd like to get to the bottom of where we start ;-) | 01:17 |
sdake | we got something already | 01:17 |
sdake | and it sort of works | 01:17 |
v1k0d3n | sdake: want to use the whatsapp chat room to meet up during the conference? | 01:17 |
sdake | but it doesn't actually provide operational features | 01:17 |
v1k0d3n | i understand | 01:17 |
sdake | which is not what kolla's mission states :) | 01:17 |
*** duonghq_ has joined #openstack-kolla | 01:19 | |
*** neilus has joined #openstack-kolla | 01:26 | |
*** zhurong has joined #openstack-kolla | 01:27 | |
*** neilus has quit IRC | 01:32 | |
*** zhurong has quit IRC | 01:34 | |
*** zhurong has joined #openstack-kolla | 01:34 | |
openstackgerrit | Zhijiang Hu proposed openstack/kolla: Uses tunnel_interface as ovs tunnel instead of api_interface https://review.openstack.org/392094 | 01:41 |
openstackgerrit | Zhijiang Hu proposed openstack/kolla: Uses tunnel_interface as ovs tunnel instead of api_interface https://review.openstack.org/392094 | 01:43 |
duonghq | sdake, are you free right now? I have some question | 01:58 |
sdake | duonghq battery at 2% | 01:58 |
sdake | shoot | 01:58 |
sdake | if it dies i'll be back online in 5-10 mins | 01:58 |
duonghq | Do we support changing number of compute nodes? i.e. add or remove compute nodes | 01:58 |
sdake | think so | 01:58 |
sdake | add a compute node | 01:59 |
sdake | run deploy | 01:59 |
duonghq | w/o need to restart instance | 01:59 |
sdake | remove not sure about | 01:59 |
sdake | i personally believe for remove we would need a remove-compute api or sosmething similar in kolla-ansible | 01:59 |
duonghq | iirc, I tried it before but every instance (nova instance) is power-off | 01:59 |
duonghq | do you mean nova-compute? | 02:00 |
sdake | you added one node, and ran deploy and everything was powered off? | 02:00 |
duonghq | iirc | 02:00 |
sdake | hmm - pretty sure i have tested that and it wfm at one point | 02:00 |
sdake | but the one point i dont recall | 02:00 |
sdake | jeffrey4l may have more up to date testing info on that | 02:00 |
duonghq | sdake, I'll re-check latter when I can build the image, | 02:01 |
Pavo | couldn't you add an option to destroy something like -n computename | 02:01 |
duonghq | proxy virus checking mess up with Kolla building process | 02:01 |
sdake | pavo yes - its been proposed but nobody has done the ansible code | 02:01 |
sdake | duonghq ya - proxies = annoying | 02:01 |
*** jtriley has joined #openstack-kolla | 02:03 | |
Pavo | also was looking at some ansible code and was thinking that you could use Passlib for passwords/yml to specify actual encyrption for passwords | 02:03 |
Pavo | http://docs.ansible.com/ansible/playbooks_prompts.html | 02:03 |
duonghq | Pavo, how about Ansible vault? | 02:04 |
duonghq | sorry but I have not catch up yet | 02:04 |
Pavo | interesting | 02:05 |
Pavo | never seen ansible vault | 02:05 |
*** jtriley has quit IRC | 02:08 | |
*** sdake has quit IRC | 02:08 | |
*** tovin07 is now known as tovin07_ | 02:08 | |
*** tovin07 has joined #openstack-kolla | 02:09 | |
*** tovin07_ has quit IRC | 02:10 | |
*** tovin07 has quit IRC | 02:10 | |
*** tovin07 has joined #openstack-kolla | 02:11 | |
*** tovin07_ has joined #openstack-kolla | 02:11 | |
*** sdake has joined #openstack-kolla | 02:12 | |
sdake | pavo we use ansible vault for your use case of ondisk encrytpion of passwords.yml | 02:12 |
*** tovin07__ has joined #openstack-kolla | 02:12 | |
Pavo | ah ok | 02:12 |
*** tovin07__ has quit IRC | 02:13 | |
*** tovin07_ has quit IRC | 02:13 | |
*** tovin07_ has joined #openstack-kolla | 02:13 | |
*** tovin07__ has joined #openstack-kolla | 02:13 | |
*** tovin07__ has quit IRC | 02:13 | |
openstackgerrit | Duong Ha-Quang proposed openstack/kolla: Specify 'become' to neccesary tasks (general roles) https://review.openstack.org/358539 | 02:13 |
*** tovin07_ has quit IRC | 02:13 | |
*** tovin07_ has joined #openstack-kolla | 02:14 | |
*** tovin07_ has quit IRC | 02:15 | |
*** tovin07_ has joined #openstack-kolla | 02:16 | |
*** tovin07 has quit IRC | 02:16 | |
*** tovin07_ is now known as tovin07 | 02:16 | |
*** tovin07 is now known as tovin07_ | 02:17 | |
Pavo | you setup your guys way of testing all your code and stuff? | 02:18 |
*** tovin07_ is now known as tovin07 | 02:18 | |
Pavo | would love to learn how to do that | 02:18 |
*** tovin07 has quit IRC | 02:18 | |
*** tovin07 has joined #openstack-kolla | 02:18 | |
*** tovin07_ has joined #openstack-kolla | 02:19 | |
openstackgerrit | Duong Ha-Quang proposed openstack/kolla: Specify 'become' for only neccesary tasks (default roles) https://review.openstack.org/359031 | 02:19 |
openstackgerrit | zhongshengping proposed openstack/kolla: baremetal group not defined in all-in-one inventory file https://review.openstack.org/394233 | 02:36 |
sdake | pavo huh? | 02:36 |
Pavo | I was asking who setup your guys way of testing your code | 02:37 |
Pavo | without phyically testing | 02:37 |
openstackgerrit | zhongshengping proposed openstack/kolla: baremetal group not defined in all-in-one inventory file https://review.openstack.org/394233 | 02:38 |
Pavo | I guess its Jenkins | 02:38 |
sdake | no jenkins isn't use | 02:42 |
sdake | d | 02:42 |
sdake | I did most of the initial setup | 02:42 |
Pavo | so what are you using to test your code | 02:42 |
sdake | SamYaple later extended it to be far better | 02:42 |
sdake | our testing infrastructure is provided by OpenStack | 02:43 |
sdake | in a tool called "zuul" | 02:43 |
Pavo | I see some reviews showing test where successful and some failed | 02:43 |
sdake | zuul.openstack.org | 02:43 |
Pavo | ah | 02:43 |
Pavo | ok | 02:43 |
Pavo | ty | 02:43 |
sdake | zuul schedules jobs via nodepool | 02:43 |
sdake | nodepool finds a vm and starts it up | 02:43 |
sdake | and runs a test script via shell | 02:43 |
sdake | rather via tox | 02:43 |
sdake | tox runs all of our testing | 02:43 |
sdake | if you look in tox.ini you can see the tests we run | 02:44 |
Pavo | hmm what is tox | 02:44 |
sdake | tox in a python infrastructure thing | 02:44 |
sdake | used to run jobs in a well-defined manner | 02:44 |
Pavo | damn I have some much to learn | 02:44 |
sdake | check out tox.ini - should be obvious ;) | 02:44 |
sdake | you heard this one before - never stop learning? :) | 02:44 |
Pavo | my brain is melting | 02:45 |
Pavo | or over flowing | 02:45 |
Pavo | lol | 02:45 |
sdake | our tests could be better | 02:45 |
sdake | we theoretically will work on that some day | 02:46 |
Pavo | I have some ideas for kolla, just gotta learn how to code it first | 02:47 |
sdake | kolla is relatively e asy to write code for | 02:48 |
sdake | its a good starter project :) | 02:48 |
Pavo | true its mostly ansible code but I am just learning ansible | 02:48 |
sdake | it was designed with 2 8 hour blocks of time required | 02:48 |
sdake | so you can learn how to contribute to kolla in about 16 hrs of playing around with it | 02:49 |
sdake | thats the theory anyway | 02:49 |
Pavo | we;; after you learn ansible lol | 02:49 |
Pavo | wll | 02:49 |
Pavo | grrr | 02:49 |
Pavo | I can not type tonight | 02:49 |
sdake | right, it helps to know anisble up front | 02:49 |
Pavo | well is what I meant | 02:49 |
sdake | even if you dont, I think you can be effective contributor in under 20 hours of code exploration | 02:50 |
sdake | all the code is very similar in nature | 02:50 |
sdake | lots of the same patterns used throughout | 02:50 |
Pavo | true | 02:50 |
sdake | kolla-kubernetes a different beast | 02:51 |
sdake | it will require much learning | 02:51 |
Pavo | but I do wonder why you guys use a specific tenant network type instead of using all types of front and if users need to use specific ones then let them change it | 02:51 |
sdake | pavo the defaults are chosen for maximum flexibility | 02:52 |
sdake | I'm not saying we have done this perfectly | 02:52 |
sdake | the example you give - i am not a networking dude, so I'm not sure what the complaint it :) | 02:52 |
sdake | it/is | 02:52 |
*** jtriley has joined #openstack-kolla | 02:57 | |
*** jtriley has quit IRC | 03:02 | |
*** jtriley has joined #openstack-kolla | 03:02 | |
*** jtriley has quit IRC | 03:07 | |
*** dave-mccowan has quit IRC | 03:09 | |
*** lrensing has joined #openstack-kolla | 03:14 | |
*** neilus has joined #openstack-kolla | 03:15 | |
duonghq | sdake, ping | 03:19 |
sdake | sup duonghq | 03:19 |
duonghq | I see we add custom policy.json recently? | 03:19 |
*** neilus has quit IRC | 03:19 | |
duonghq | and we put it in {{ node_config_directory }}/nova/ which has not been created | 03:20 |
duonghq | or I missed something? | 03:21 |
*** jrich523_ has joined #openstack-kolla | 03:23 | |
duonghq | we also do not mount nova (for ex) to container | 03:24 |
duonghq | any idea, sdake? | 03:24 |
sdake | duonghq right | 03:25 |
*** jrich523 has quit IRC | 03:25 | |
sdake | moment | 03:25 |
openstackgerrit | Duong Ha-Quang proposed openstack/kolla: Specify 'become' for only neccesary tasks (all other roles) https://review.openstack.org/359096 | 03:32 |
*** yingjun has quit IRC | 03:36 | |
*** yingjun has joined #openstack-kolla | 03:38 | |
sdake | duonghq so custom policy.json is only in masater | 03:47 |
sdake | and only a few patches have merged so far | 03:47 |
sdake | the directories are not created by default | 03:48 |
sdake | if you need them you have to create them | 03:48 |
sdake | but in newotn, policy.json is not customizable | 03:48 |
duonghq | sdake, sure, but iirc, template does not create directory on target nodes automatically | 03:48 |
duonghq | and we also do not create this | 03:49 |
sdake | duonghq those are source addresses | 03:49 |
sdake | source files rather | 03:49 |
sdake | oh wait moment | 03:49 |
sdake | ya i'm not sure what creates the target dirs, if its template or what, but they are definately created | 03:49 |
sdake | node_config_directory is /etc/kolla | 03:50 |
sdake | and in a deployent, /etc/kolla/nova will always exist since liberty | 03:50 |
duonghq | hmm, okay, even in this case, seem that nothing mount /etc/kolla/nova to container? | 03:50 |
duonghq | I mean start_contr..... yml | 03:51 |
*** jtriley has joined #openstack-kolla | 03:57 | |
sdake | moment | 03:57 |
*** CBR09 has joined #openstack-kolla | 03:58 | |
sdake | duonghq here is where it is loaded: https://github.com/openstack/kolla/blob/master/ansible/roles/nova/tasks/start_controllers.yml#L10 | 03:59 |
duonghq | sdake, it's loading nova-api one, not nova one | 04:00 |
sdake | duonghq here is where it is written: https://github.com/openstack/kolla/blob/master/ansible/roles/nova/tasks/config.yml#L58 | 04:00 |
sdake | there is no nova-api one... | 04:00 |
sdake | its only nova | 04:01 |
duonghq | it's merged and copied to {{ node_config_directory }}/{{ item }} in nova.conf, right? | 04:01 |
duonghq | for e.x. {{ node_config_directory }}/nova-api/nova.conf | 04:01 |
*** jtriley has quit IRC | 04:02 | |
*** zhubingbing_ has joined #openstack-kolla | 04:05 | |
sdake | duonghq check ou tthis code: https://github.com/openstack/kolla/blob/master/ansible/roles/nova/tasks/config.yml#L58-L59 | 04:07 |
sdake | duonghq do you ahve a specific bug to report? | 04:07 |
sdake | duonghq the code looks correct to me | 04:07 |
duonghq | This code is totally ok, I do not blame this piece of code | 04:08 |
duonghq | but I want to make sure I interpret correctly | 04:09 |
*** neilus has joined #openstack-kolla | 04:09 | |
*** coolsvap has joined #openstack-kolla | 04:10 | |
duonghq | As I understand, this code merge above configs, put in nova.conf and deploy to {{ node_config_directory }}/{{ item }} directory under nova.conf file? | 04:10 |
duonghq | sdake, correct me if I wrong | 04:10 |
sdake | right | 04:11 |
sdake | nova-api/nova.conf | 04:11 |
duonghq | so, nova-api is mounted into container in https://github.com/openstack/kolla/blob/master/ansible/roles/nova/tasks/start_controllers.yml#L10 | 04:12 |
sdake | item in that example = nova-api | 04:12 |
duonghq | you can see we mount nova-.... to container | 04:12 |
sdake | right | 04:12 |
sdake | do you see the : inbetween the mount points | 04:12 |
duonghq | but the policy.json is place in "{{ node_config_directory }}/nova/policy.json" | 04:13 |
sdake | the first is the host mount, the second is the container mount point | 04:13 |
sdake | got a link to that? | 04:13 |
duonghq | https://github.com/openstack/kolla/blob/master/ansible/roles/nova/tasks/config.yml#L94 | 04:13 |
sdake | that is an error if thats the case | 04:13 |
duonghq | every policy.json is deploy by this manner | 04:13 |
sdake | yup that code is busted | 04:13 |
sdake | needs a with_item: | 04:14 |
*** neilus has quit IRC | 04:14 | |
sdake | policy.json patches are just merging now iiuc | 04:14 |
sdake | probably untested, and reviewers failed to detect the issue | 04:14 |
duonghq | so I'll file a bug | 04:15 |
duonghq | thank you sdake | 04:16 |
duonghq | sdake, fyi https://bugs.launchpad.net/kolla/+bug/1639686 | 04:20 |
openstack | Launchpad bug 1639686 in kolla "Wrong placement of policy.json " [Undecided,New] | 04:20 |
*** sdake has quit IRC | 04:21 | |
*** ayoung has quit IRC | 04:25 | |
*** yingjun has quit IRC | 04:50 | |
*** jtriley has joined #openstack-kolla | 04:51 | |
*** yingjun has joined #openstack-kolla | 04:53 | |
*** jtriley has quit IRC | 04:56 | |
*** jtriley has joined #openstack-kolla | 04:56 | |
*** jtriley has quit IRC | 05:01 | |
openstackgerrit | Duong Ha-Quang proposed openstack/kolla: Fix placement of policy.json https://review.openstack.org/394260 | 05:05 |
*** skramaja has joined #openstack-kolla | 05:10 | |
*** f13o_ has quit IRC | 05:10 | |
*** jtriley has joined #openstack-kolla | 05:18 | |
*** Jeffrey4l has joined #openstack-kolla | 05:21 | |
openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 05:28 |
*** jtriley has quit IRC | 05:34 | |
*** CBR09 has quit IRC | 05:43 | |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Fix plugin parsing for base images https://review.openstack.org/394267 | 05:43 |
*** neilus has joined #openstack-kolla | 05:57 | |
openstackgerrit | Duong Ha-Quang proposed openstack/kolla: Fix placement of policy.json https://review.openstack.org/394260 | 05:59 |
*** neilus has quit IRC | 06:02 | |
zhubingbing_ | sdake | 06:03 |
duonghq | zhubingbing_, he does not online now | 06:04 |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Bump OpenStack service version https://review.openstack.org/394271 | 06:04 |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Bump OpenStack service version https://review.openstack.org/394277 | 06:13 |
bjolo | morning | 06:23 |
duonghq | morning bjolo | 06:24 |
bjolo | duonghq, just read your conversation with sdake | 06:24 |
bjolo | let me go through that | 06:24 |
bjolo | :) | 06:24 |
duonghq | bjolo, nice, just the policy.json placement, in Kolla, we merge all config file and put in nova.conf (for example) | 06:26 |
duonghq | continue with nova example: | 06:26 |
duonghq | this config is distribute to nova component by mounting config folder in container | 06:27 |
zhubingbing_ | duonghq | 06:27 |
bjolo | yes | 06:27 |
openstackgerrit | Merged openstack/kolla: Install MySQL-python with pip in openstack-base container (type source) https://review.openstack.org/384938 | 06:27 |
zhubingbing_ | can u help me review it ?https://review.openstack.org/#/c/387232/ | 06:27 |
duonghq | somebody put policy.json into "nova" one, rendering it is useless due to nobody mount it into container, (and it does not meant is, also) | 06:30 |
duonghq | zhubingbing_, moment | 06:30 |
zhubingbing_ | 3 | 06:31 |
zhubingbing_ | thinks | 06:32 |
zhubingbing_ | thanks | 06:32 |
zhubingbing_ | -) | 06:32 |
duonghq | bjolo, sorry for bad explanation :( | 06:34 |
bjolo | no worries | 06:34 |
bjolo | tracing through the code now | 06:34 |
duonghq | sorry bjolo :( | 06:35 |
duonghq | Jeffrey4l, are you there? | 06:40 |
Jeffrey4l | duonghq, yep. | 06:41 |
duonghq | you said that deploy action can handle the start one service or only one host, how can I do that? | 06:41 |
Jeffrey4l | for one service, using ` --tag nova` | 06:41 |
Jeffrey4l | for one host using `--limit node1 ` | 06:41 |
duonghq | understood, forgot that we passthrough args to ansible | 06:42 |
duonghq | thank you | 06:42 |
Jeffrey4l | np ;) | 06:42 |
*** tonanhngo has joined #openstack-kolla | 06:44 | |
*** bdaca has quit IRC | 06:45 | |
*** tonanhngo has quit IRC | 06:46 | |
*** tonanhngo has joined #openstack-kolla | 06:46 | |
duonghq | zhubingbing_, I see egonzalez comment something about Fedora deprecation, do you have final decision about it? | 06:48 |
zhubingbing_ | Fedora?sorry | 06:49 |
zhubingbing_ | i should remove it | 06:49 |
zhubingbing_ | i fix it now | 06:49 |
*** berendt has joined #openstack-kolla | 07:00 | |
*** berendt has quit IRC | 07:00 | |
*** berendt has joined #openstack-kolla | 07:01 | |
*** tonanhngo has quit IRC | 07:01 | |
*** Jeffrey4l has quit IRC | 07:18 | |
openstackgerrit | Wei Cao proposed openstack/kolla: Add Solum doc link in README.rst https://review.openstack.org/394296 | 07:30 |
openstackgerrit | zhubingbing proposed openstack/kolla: add panko role https://review.openstack.org/394298 | 07:33 |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Install all python packages in global-requirements.txt file https://review.openstack.org/394150 | 07:37 |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Install all python packages in global-requirements.txt file https://review.openstack.org/394150 | 07:39 |
*** khamtamtun has joined #openstack-kolla | 07:41 | |
*** khamtamtun has quit IRC | 07:43 | |
*** asalkeld has joined #openstack-kolla | 07:43 | |
*** asalkeld has quit IRC | 07:47 | |
openstackgerrit | Merged openstack/kolla: Add Solum doc link in README.rst https://review.openstack.org/394296 | 07:48 |
openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 07:51 |
*** athomas has joined #openstack-kolla | 07:53 | |
*** tonanhngo has joined #openstack-kolla | 08:00 | |
*** asalkeld has joined #openstack-kolla | 08:01 | |
*** tonanhngo has quit IRC | 08:01 | |
*** asalkeld has quit IRC | 08:06 | |
openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 08:08 |
duonghq | coolsvap, if you have space time, can you give me a comment about my ps: https://review.openstack.org/#/c/394260/2 | 08:08 |
duonghq | *spare | 08:08 |
* coolsvap checks | 08:12 | |
coolsvap | duonghq: it will take some time need to get the context of your conversation | 08:14 |
duonghq | coolsvap, basically, I try to get policy.json in right place, | 08:15 |
*** hyakuhei has quit IRC | 08:15 | |
duonghq | but I'm not sure if the file need to be placed in every service or only api ones | 08:15 |
*** Jeffrey4l has joined #openstack-kolla | 08:17 | |
*** asalkeld has joined #openstack-kolla | 08:19 | |
*** asalkeld has quit IRC | 08:19 | |
*** matrohon has joined #openstack-kolla | 08:22 | |
*** hyakuhei has joined #openstack-kolla | 08:24 | |
openstackgerrit | zhubingbing proposed openstack/kolla: fix collectd dockerfile https://review.openstack.org/383997 | 08:25 |
*** tonanhngo has joined #openstack-kolla | 08:26 | |
coolsvap | duonghq: I kinda agree with egongalez | 08:26 |
coolsvap | Need to verify how it works when you change policies | 08:26 |
*** tonanhngo has quit IRC | 08:27 | |
coolsvap | Should work fine I suppose | 08:27 |
zhubingbing_ | around Eduardo Gonzalez | 08:28 |
duonghq | coolsvap, I also think so, but I have not much experience with OpenStack in general | 08:28 |
duonghq | thank you coolsvap | 08:28 |
coolsvap | I can check but not right now | 08:29 |
duonghq | thank you very much | 08:29 |
*** mkoderer has joined #openstack-kolla | 08:31 | |
*** zhubingbing_ has quit IRC | 08:35 | |
-openstackstatus- NOTICE: Gerrit is going to be restarted due to slowness and proxy errors | 08:46 | |
*** openstackgerrit has quit IRC | 08:48 | |
*** openstackgerrit has joined #openstack-kolla | 08:48 | |
*** tonanhngo has joined #openstack-kolla | 08:54 | |
*** tonanhngo has quit IRC | 08:55 | |
*** hyakuhei has quit IRC | 08:59 | |
*** hyakuhei has joined #openstack-kolla | 08:59 | |
*** hyakuhei has quit IRC | 08:59 | |
*** hyakuhei has joined #openstack-kolla | 08:59 | |
*** shardy has joined #openstack-kolla | 09:00 | |
*** Serlex has joined #openstack-kolla | 09:06 | |
*** tonanhngo has joined #openstack-kolla | 09:15 | |
*** tonanhngo has quit IRC | 09:17 | |
*** pbourke has joined #openstack-kolla | 09:19 | |
*** Serlex has quit IRC | 09:23 | |
sp_ | Jeffrey4l : I was exploring about some use cases of kolla-kubernetes, i just need to know, if we compare kolla-kubernetes with fuel then in which way it would be more effective. | 09:25 |
sp_ | Jeffrey4l : sdake it would be very helpful, if anyone provide your views or opinion for the same | 09:26 |
sp_ | I am just a new folks to this arena | 09:27 |
Jeffrey4l | sp_, ignore the k8s. compare container with fuel, the benefit are: 1. easy to deploy 2. unified running environment, | 09:27 |
Jeffrey4l | 3. possible to upgrade ( without care about the host OS ) | 09:27 |
Jeffrey4l | kolla-k8s bring application ha which come from k8s | 09:28 |
*** msimonin has joined #openstack-kolla | 09:32 | |
sp_ | Jeffrey4l : thanks for your kind information, I will come up with some more points during my investigation. I will relate your points with my further exploration. | 09:33 |
*** zhurong_ has joined #openstack-kolla | 09:34 | |
*** zhurong has quit IRC | 09:35 | |
*** yingjun has quit IRC | 09:37 | |
*** yingjun has joined #openstack-kolla | 09:37 | |
*** yingjun has quit IRC | 09:37 | |
*** yingjun has joined #openstack-kolla | 09:38 | |
*** yingjun has quit IRC | 09:38 | |
*** yingjun has joined #openstack-kolla | 09:38 | |
*** mgoddard has joined #openstack-kolla | 09:39 | |
*** yingjun has quit IRC | 09:39 | |
*** yingjun has joined #openstack-kolla | 09:39 | |
*** yingjun has quit IRC | 09:40 | |
*** yingjun has joined #openstack-kolla | 09:40 | |
Jeffrey4l | you are welcome ;) | 09:40 |
*** yingjun has quit IRC | 09:44 | |
duonghq | Jeffrey4l, can you confirm egonzalez point in https://review.openstack.org/#/c/394260/ ? | 09:45 |
*** pbourke has quit IRC | 09:48 | |
Jeffrey4l | duonghq, i am not sure about this. especially in current lots of openstack project. | 09:48 |
duonghq | Jeffrey4l, me too | 09:48 |
Jeffrey4l | nova moved all policy check into api level in liberty?(not sure) . | 09:48 |
duonghq | the point is true for nova, keystone, but for other services, I'm not sure | 09:49 |
*** pbourke has joined #openstack-kolla | 09:50 | |
Jeffrey4l | yes. so i think the right solution is: add policy into api service only when u are pretty sure. | 09:50 |
duonghq | okay, so I think I'll wait until more guys can confirm | 09:50 |
Jeffrey4l | otherwise, when end-user add custom policy, it may not work as expect. | 09:51 |
*** tovin07 has quit IRC | 09:51 | |
duonghq | according the link I included in my comment, the policy should be used only by api service. Do I miss understand it? | 09:52 |
duonghq | pbourke, ping | 09:53 |
pbourke | duonghq: hi | 09:53 |
duonghq | nice to see you | 09:53 |
pbourke | morning :) | 09:53 |
Jeffrey4l | oh. i am wrong about the nova policy. it is move from db level to api level. not move from other module to api. | 09:53 |
Jeffrey4l | maybe it is correct, policy file is only used by api. | 09:53 |
*** tovin07_ has quit IRC | 09:53 | |
*** pprokop has joined #openstack-kolla | 09:53 | |
duonghq | pbourke, I rebased my ansible become bp, so if you can get some spare time, please review it, thank you | 09:54 |
pbourke | will do | 09:54 |
duonghq | Jeffrey4l, if the link implied every service should use policy.json only in API part, so I think we can enforce it in Kolla, | 09:55 |
*** shardy has quit IRC | 09:55 | |
*** sdake has joined #openstack-kolla | 09:55 | |
duonghq | pbourke, sdake: if you have spare time, can you give me some comment about policy.json placement in https://review.openstack.org/#/c/394260/ | 09:56 |
Jeffrey4l | yep, | 09:56 |
*** shardy has joined #openstack-kolla | 09:57 | |
duonghq | I asked some people and it seems that policy.json should only be used by api service, if I can get 1-2 more guys agreed with this, I'll update by ps | 09:58 |
*** sdake_ has joined #openstack-kolla | 10:00 | |
*** Serlex has joined #openstack-kolla | 10:02 | |
*** sdake has quit IRC | 10:03 | |
duonghq | sdake_, ping | 10:03 |
sdake_ | sup | 10:04 |
sdake_ | hey Jeffrey4l | 10:04 |
duonghq | if you have spare time, can you give me some comment about policy.json placement in https://review.openstack.org/#/c/394260/ | 10:04 |
sdake_ | just woke up | 10:04 |
sdake_ | prefer not to review right when i wake up | 10:04 |
Jeffrey4l | sdake_, hi | 10:04 |
duonghq | oh, sorry, always forgot you in America | 10:05 |
pbourke | duonghq: what if I want to use the same policy for all aodh services for example | 10:05 |
pbourke | duonghq: I would have to copy the same file 4 times | 10:05 |
sdake_ | ya i'm getting older and only need about 6 hrss of sleep | 10:05 |
*** zhurong_ has quit IRC | 10:05 | |
sdake_ | wen to bed around 10pm last night so 5 hrs will hae to do :( | 10:05 |
pbourke | duonghq: so perhaps, using this example (https://review.openstack.org/#/c/394260/2/ansible/roles/aodh/tasks/config.yml) we should have an entry for just "aodh" also | 10:05 |
pbourke | does that make sense? | 10:06 |
duonghq | pbourke, as I understand only API utilize the policy? | 10:06 |
*** shardy_ has joined #openstack-kolla | 10:07 | |
sdake_ | duonghq the with_items for config.yml isn't quite right | 10:07 |
sdake_ | ya what pbourke said :) | 10:07 |
pbourke | :p | 10:07 |
sdake_ | you are missing two entries for each of those | 10:07 |
sdake_ | but i'm not going to finish the review - don't want to break it :) | 10:08 |
sdake_ | was that policy.json code merged prior to newton? | 10:08 |
sdake_ | its completely wrong | 10:08 |
duonghq | pbourke, sdake_, so which service should use the policy.json? | 10:08 |
sdake_ | duonghq the issue is your not copying from /etc/kolla/config/nova/policy.json | 10:09 |
*** duonghq_ has quit IRC | 10:09 | |
sdake_ | duonghq another issue is your not copying from /etc/kolla/config/nova-10.10.10.1/nova/policy.son | 10:09 |
pbourke | i also cant see where the services are copying in the policy.json file - there's no entry in the json | 10:09 |
*** egonzalez90 has joined #openstack-kolla | 10:09 | |
duonghq | sure | 10:09 |
*** shardy has quit IRC | 10:10 | |
egonzalez90 | Morning | 10:10 |
sdake_ | sup egonzalez90 | 10:10 |
duonghq | I'll add this | 10:10 |
duonghq | pbourke, I'm sorry but you mean which json? | 10:12 |
duonghq | sdake_, when the /etc/kolla/config/nova-10.10.10.1/nova/policy.son is generated? | 10:13 |
duonghq | morning egonzalez90 | 10:14 |
*** shardy_ is now known as shardy | 10:14 | |
duonghq | pbourke, you mean the template files? | 10:16 |
pbourke | duonghq: yes, i'll add more detailed info to the review in a moment | 10:16 |
duonghq | pbourke, sdake_ thank you, I need leaving now, will read the IRC log later | 10:17 |
pbourke | duonghq: ok thanks chat later :) | 10:17 |
*** duonghq has quit IRC | 10:18 | |
openstackgerrit | Paul Bourke (pbourke) proposed openstack/kolla: Install neutron-lbaas in neutron-server https://review.openstack.org/393435 | 10:22 |
openstackgerrit | Paul Bourke (pbourke) proposed openstack/kolla: Allow neutron-fwaas to be enabled in Neutron https://review.openstack.org/393409 | 10:22 |
pbourke | egonzalez90: ping | 10:24 |
egonzalez90 | pbourke pong | 10:24 |
pbourke | hey man | 10:25 |
pbourke | so for your patch around custom policy files - I think we're missing corresponding json template changes for the service to actually copy them in on start up? | 10:25 |
*** sdake_ is now known as sdake | 10:25 | |
pbourke | egonzalez90: e.g. for keystone we have https://github.com/openstack/kolla/blob/master/ansible/roles/keystone/templates/keystone.json.j2#L25-L31 | 10:26 |
egonzalez90 | Is under review that PS | 10:26 |
pbourke | but not for aodh https://github.com/openstack/kolla/blob/master/ansible/roles/aodh/templates/aodh-api.json.j2 | 10:26 |
egonzalez90 | Let me find it | 10:26 |
pbourke | oh there's another PS? | 10:26 |
egonzalez90 | https://review.openstack.org/#/c/394177/ | 10:26 |
pbourke | ah | 10:26 |
pbourke | thanks | 10:26 |
egonzalez90 | But we are not sure which services need to be in an API level or in every service | 10:27 |
egonzalez90 | Im digging into that | 10:27 |
egonzalez90 | Mostly only on API servers | 10:27 |
pbourke | ok sounds good | 10:28 |
*** b_bezak has joined #openstack-kolla | 10:40 | |
*** ppalacios has joined #openstack-kolla | 10:43 | |
*** pprokop has quit IRC | 10:47 | |
*** pprokop has joined #openstack-kolla | 10:49 | |
*** egonzalez90 has quit IRC | 10:49 | |
*** egonzalez90 has joined #openstack-kolla | 10:49 | |
*** csj_ has joined #openstack-kolla | 10:52 | |
*** csj has quit IRC | 10:55 | |
sdake | ok - there is another issue egonzalez90 | 11:04 |
sdake | the patches as is are wrong | 11:04 |
sdake | i merged a couple of them until Duong pointed out the problem | 11:04 |
egonzalez90 | Yes, i noticed | 11:04 |
egonzalez90 | Dest path is completely wrong | 11:04 |
sdake | ok cool | 11:04 |
egonzalez90 | My fault, sorry | 11:04 |
sdake | i was getting ready to beat you up with git links :) | 11:05 |
sdake | i may have guided duong wrong | 11:05 |
sdake | i told him we needed a with_items | 11:05 |
sdake | i didn't realize policy only applies to api containers | 11:05 |
egonzalez90 | We are not really sure which services uses policy on api and which in other services | 11:06 |
egonzalez90 | Im digging into this | 11:06 |
sdake | when in doubt - copy em all :) | 11:06 |
egonzalez90 | Most of them only in API | 11:06 |
sdake | ok | 11:06 |
egonzalez90 | Yup, i was thinking on that, will not harm having all | 11:06 |
sdake | little sloppy - but whatever | 11:07 |
sdake | if you think its relaitvely easy to understand whatneeds a policy.json then my rec would be to proceed with optimizing the list | 11:07 |
*** g3ek has quit IRC | 11:09 | |
egonzalez90 | There is a lack of docs about policies, but most services follows the API, however we should test every service for this | 11:10 |
egonzalez90 | pbourke, kolla meetings are at 5pm in Ireland time right? | 11:12 |
*** haplo37 has quit IRC | 11:13 | |
sdake | egonzalez90 docs always weak :( | 11:13 |
pbourke | egonzalez90: we're now on UTC since last weekend so meetings are 4pm in Ireland | 11:14 |
egonzalez90 | pbourke, thanks | 11:14 |
*** haplo37 has joined #openstack-kolla | 11:15 | |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Wait the wsrep_ready to be ON in mariadb https://review.openstack.org/394368 | 11:15 |
pbourke | egonzalez90: so you've made the move? | 11:15 |
egonzalez90 | Yup | 11:16 |
pbourke | cool welcome :) | 11:17 |
egonzalez90 | Welcomed with cold weather ;) | 11:17 |
*** g3ek has joined #openstack-kolla | 11:18 | |
pbourke | cold but nice | 11:19 |
sdake | ok guys, sdake trying to learn here not being a jerk | 11:19 |
sdake | could folks define for me "cloud native" | 11:19 |
pbourke | written in go | 11:20 |
pbourke | of course | 11:20 |
sdake | groan | 11:20 |
sdake | written in go is a big hurdle for openstack | 11:22 |
pbourke | for me it generally means written in a loosely coupled way, microservices etc. Stateless where possible and doesn't have hard dependencies on the underlying infrastructure | 11:22 |
sdake | so that lat part | 11:23 |
sdake | the hard dependencies | 11:23 |
pbourke | lots of link online explain it better | 11:23 |
*** pc_m has joined #openstack-kolla | 11:23 | |
sdake | yes, but I want to learn from you, not from links | 11:23 |
*** rhallisey has joined #openstack-kolla | 11:23 | |
sdake | the hard dependencies, that is only possible if the application has NO dependencies | 11:24 |
sdake | would you argue otherwise? | 11:24 |
rhallisey | sdake, morning | 11:24 |
sdake | morning rhallisey | 11:24 |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Move glance precheck into its own role https://review.openstack.org/394371 | 11:24 |
sdake | rhallisey spoke with v1k0d3n for several hours yesterday | 11:24 |
rhallisey | ya | 11:24 |
*** athomas has quit IRC | 11:25 | |
rhallisey | just starting to read it | 11:25 |
*** tonanhngo has joined #openstack-kolla | 11:25 | |
sdake | rhallisey he seems to think an "operator" is the way to go | 11:25 |
sdake | see what I did there | 11:25 |
sdake | I suspect the barrier to an operator is pretty low for us, assuming we write the code in python | 11:25 |
rhallisey | explain | 11:26 |
*** tonanhngo has quit IRC | 11:26 | |
rhallisey | or I'll read the logs | 11:26 |
sdake | which part | 11:26 |
sdake | what an operator is? | 11:26 |
rhallisey | idk what you mean by an operator | 11:26 |
sdake | right | 11:26 |
sdake | moment | 11:26 |
sdake | read this please https://coreos.com/blog/introducing-the-etcd-operator.html | 11:26 |
sdake | then grill me | 11:26 |
sdake | i want to have my story straight for today :) | 11:27 |
rhallisey | roger that | 11:27 |
sdake | so i need that to happen in the next 20 mins :) | 11:27 |
sdake | followup material: https://coreos.com/blog/introducing-operators.html | 11:31 |
sdake | (the first link was a real world example) | 11:31 |
sdake | so pbourke - re your definition of cloud native, the hard dep thing - any thoughts? | 11:31 |
pbourke | sdake: trying to think how I can explain in a way that makes sense | 11:32 |
*** athomas has joined #openstack-kolla | 11:32 | |
sdake | oh lagging, got i t ;) | 11:32 |
sdake | thoguth you were ignoring me | 11:32 |
sdake | not trying to be a jerk, remember ;) | 11:32 |
pbourke | sdake: think of something like rabbitmq, it needs the IPs of each in it's cluster hardcoded into it's config file | 11:32 |
pbourke | also it depends on the hostnames being resolvable | 11:32 |
pbourke | if one of those needs to be changed there are multiple steps | 11:33 |
sdake | do the ips really need to be hardcoded, or would dns names be enough? | 11:33 |
pbourke | where as a more cloud native app will respond easily to a node dropping out and a new one coming online. pull its configs from etcd or similar | 11:33 |
*** caowei has joined #openstack-kolla | 11:33 | |
pbourke | I think DNS is enough in that case | 11:33 |
sdake | so *DNS* needs to be changed and is the hard dependency in your example? | 11:34 |
pbourke | yes | 11:34 |
sdake | there is no way of getting away from hard dependencies in software development | 11:34 |
pbourke | im not an expert | 11:34 |
sdake | i understand | 11:34 |
pbourke | right but I think its how those dependencies are managed | 11:35 |
sdake | ok, so hard dependencies *MANAGED* is what you meant then? | 11:35 |
pbourke | inability to horizontally scale seems to be a big feature of non cloud native apps also | 11:35 |
rhallisey | sdake, do we need to use helm anymore? | 11:35 |
pbourke | as in they depend on a very specific hardware setup, need a lot of memory etc | 11:35 |
rhallisey | seems like | 11:35 |
pbourke | rhallisey might have a good take on it | 11:35 |
sdake | rhallisey v1k0d3n says yes, I am on th fence | 11:36 |
rhallisey | seems like an operator can handle the deps | 11:36 |
sdake | rhallisey would you mind defining cloud native for me | 11:36 |
rhallisey | ok | 11:36 |
pbourke | rhallisey: so according to the coreos blog helm is still necessary | 11:36 |
*** mliima has joined #openstack-kolla | 11:36 | |
sdake | pbourke which keywords should I search for | 11:36 |
sdake | pbourke i missed that | 11:36 |
pbourke | sdake: on the coreos blog? it's in the faq, I guess search for helm | 11:37 |
rhallisey | cloud native - a service that has the ability to be destroyed without fear of destorying the application | 11:37 |
sdake | Q: How is this different than Helm? | 11:37 |
sdake | A: Helm is a tool for packaging multiple Kubernetes resources into a single package. The concept of packaging up multiple applications together and using Operators that actively manage applications are complementary. For example, traefik is a load balancer that can use etcd as its backend database. You could create a Helm Chart that deploys a traefik Deployment and etcd cluster instance together. The etcd cluster would then be | 11:37 |
sdake | deployed and managed by the etcd Operator. | 11:37 |
sdake | is that what your talking about? | 11:38 |
pbourke | that's the one | 11:38 |
sdake | i think its pretty cool to see all the different definitions of cloud native | 11:38 |
rhallisey | let me see the offical def... | 11:38 |
rhallisey | if there is one | 11:38 |
sdake | dont corrupt my thinking w ith marketing spam | 11:39 |
sdake | pls | 11:39 |
sdake | think for yourselves guys :) | 11:39 |
sdake | here is what I dont get about helm atm - why doesn't it just become a global operator? | 11:39 |
rhallisey | sdake, yes | 11:40 |
sdake | rather then have tens of thousands of operators for each service | 11:40 |
rhallisey | imo right now, helm hardly gives us much | 11:40 |
sdake | rather for all ervices | 11:40 |
sdake | rhallisey i haven't evaled helm | 11:40 |
sdake | have you? | 11:40 |
rhallisey | from what I can tell so far | 11:40 |
rhallisey | it's a bunch of charts | 11:40 |
rhallisey | essentially, it's what we already have in our repo | 11:40 |
rhallisey | but it's the official way of doing things | 11:41 |
sdake | we have = which part? | 11:41 |
sdake | you mean the ansible part? | 11:41 |
rhallisey | no | 11:41 |
rhallisey | we have templates defining config maps, pods, RCs, deployments, ect... | 11:41 |
sdake | mind throwing some git links at me so i can understand | 11:41 |
sdake | need rapid learning | 11:42 |
sdake | i depart for cncf in 8 hrs | 11:42 |
rhallisey | ah | 11:42 |
sdake | this is prep for that | 11:42 |
rhallisey | sdake, what's your definition of cloud native? | 11:42 |
rhallisey | https://github.com/openstack/kolla-kubernetes/tree/master/services/mariadb | 11:43 |
sdake | cloud native is an application born in the cloud meant to use cloud services. Just like I am a native Arizonan of the United States using the services available to me of the city, state, and country I live in. | 11:43 |
rhallisey | ^ those are the mariadb definitions | 11:43 |
sdake | and a chart is a collection of the j2's converted to just json? | 11:44 |
*** tonanhngo has joined #openstack-kolla | 11:45 | |
rhallisey | https://github.com/kubernetes/charts/tree/master/stable/mariadb/templates | 11:45 |
*** tonanhngo has quit IRC | 11:46 | |
rhallisey | sdake, I like that definition too | 11:47 |
sdake | imy definition has nothing to do with | 11:47 |
sdake | 1. autohealing | 11:47 |
sdake | 2. self-recovery | 11:47 |
rhallisey | ya | 11:47 |
sdake | 3. dynamic scheduling | 11:47 |
sdake | 4. microservices architecture | 11:48 |
sdake | 5. loosely coupled | 11:48 |
sdake | 6. dependency management | 11:48 |
sdake | 7. containers | 11:48 |
sdake | or | 11:48 |
sdake | as we all know | 11:48 |
sdake | go | 11:48 |
rhallisey | thoughts on the link? | 11:48 |
sdake | rhallisey processing | 11:49 |
rhallisey | k | 11:49 |
pbourke | sdake: good list | 11:49 |
sdake | https://github.com/kubernetes/charts/blob/master/stable/mariadb/templates/pvc.yaml#L1 | 11:49 |
sdake | pbourke those are implementation details imo :) | 11:49 |
rhallisey | sdake, ya I derive my definition from those items | 11:51 |
sdake | so that line 1 - where does sthe {{ }} come from | 11:51 |
rhallisey | you definition is neat because it takes into account the meaning of the two words | 11:51 |
rhallisey | cloud & native | 11:51 |
sdake | thanks my wife actually came up with it | 11:55 |
sdake | rbergeron gets the credit | 11:55 |
sdake | although I came up with the list on my own :) | 11:58 |
*** eaguilar has joined #openstack-kolla | 12:03 | |
*** tonanhngo has joined #openstack-kolla | 12:04 | |
*** b_bezak has quit IRC | 12:05 | |
*** tonanhngo has quit IRC | 12:05 | |
sdake | so rhallisey i think you are lagging, what is https://github.com/kubernetes/charts/blob/master/stable/mariadb/templates/pvc.yaml#L1 fed by? | 12:09 |
sdake | [04:52:20] <sdake>so that line 1 - where does sthe {{ }} come from | 12:10 |
sdake | that is a variable substitution, but where is the variable substituted from? | 12:10 |
rhallisey | https://github.com/kubernetes/charts/blob/master/stable/mariadb/values.yaml | 12:10 |
sdake | values.yaml is a built-in in helm charts? | 12:11 |
rhallisey | what do you mean by build-in? | 12:11 |
rhallisey | built-in* | 12:12 |
sdake | i mean like all.yaml | 12:12 |
sdake | expected | 12:12 |
sdake | required | 12:12 |
rhallisey | yes | 12:12 |
sdake | its part of every chart | 12:12 |
rhallisey | idk if you need that exact name | 12:12 |
rhallisey | probably tho | 12:12 |
sdake | ok - how to override values.yaml? | 12:12 |
sdake | perhaps I lack vision (or helm lacks a roadmap of one), but what precise value does helm add to kolla-kubernetess? | 12:14 |
sdake | rhallisey ^^ | 12:16 |
sdake | if helm doesn't provide an operational experience, what is its purpose in life? | 12:16 |
sdake | rhallisey ^^ | 12:17 |
openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 12:18 |
rhallisey | it's the kubernetes native way of handleing an application | 12:19 |
rhallisey | rpm | 12:19 |
rhallisey | it doesn't fill a gap, but dresses up what we already have in templates | 12:20 |
sdake | ok, so work for later then? | 12:20 |
sdake | to me, the biggest problem we hae today is lack of orchestration | 12:20 |
rhallisey | yes | 12:20 |
sdake | I think of an operator as a custom orchestrator | 12:20 |
rhallisey | I thought helm could've helped us there | 12:20 |
sdake | because that is precisely what it is | 12:20 |
rhallisey | but it won't | 12:20 |
sdake | it won't why? | 12:21 |
sdake | you mean it doesn't? | 12:21 |
sdake | or won't | 12:21 |
rhallisey | doesn't | 12:21 |
sdake | two totally differnet words :) | 12:21 |
sdake | ok | 12:21 |
rhallisey | brr | 12:21 |
rhallisey | 36 degrees in boston | 12:21 |
sdake | also I think the idea of the entrypoint while intriguing, didnt' work for us twice in the past | 12:22 |
sdake | and requires significant rework to the container framework we have in place | 12:22 |
rhallisey | the entrypoint piece is challenging | 12:24 |
v1k0d3n | morning guyes | 12:24 |
sdake | sup v1k0d3n | 12:24 |
rhallisey | v1k0d3n, yo | 12:24 |
v1k0d3n | sdake: you never sleep man :) haha | 12:24 |
sdake | rhallisey you remember compsoe worked 19 times out of 20 | 12:24 |
sdake | consistently | 12:24 |
sdake | v1k0d3n 6 hrs a day typically :) | 12:24 |
rhallisey | yes | 12:24 |
v1k0d3n | are you really an intelligent bot on IRC? | 12:24 |
*** zhurong has joined #openstack-kolla | 12:24 | |
v1k0d3n | yo rhallisey | 12:24 |
sdake | and that was a crude implementation of entrypoint | 12:24 |
rhallisey | that was the baked in | 12:25 |
v1k0d3n | haha | 12:25 |
rhallisey | config internal | 12:25 |
sdake | v1k0d3n whoever invented me should earn a turing award ;) | 12:25 |
*** tonanhngo has joined #openstack-kolla | 12:25 | |
v1k0d3n | reading through notes above...looks like you discussed some about the operators though | 12:25 |
sdake | v1k0d3n yup | 12:25 |
*** tonanhngo has quit IRC | 12:25 | |
sdake | v1k0d3n i also synced up with teh tc on what it would take to make go happen in the openstack ecosystem | 12:25 |
sdake | since go inevitbly is a thing that people will want for the wrong reasons | 12:26 |
rhallisey | what did they say? | 12:26 |
sdake | the vote for swift on go was 6/7 | 12:26 |
sdake | needs to be 7/6 | 12:26 |
sdake | in order to be 7/6, openstack needs to grow | 12:26 |
sdake | 1. dependency management system | 12:26 |
*** f13o has joined #openstack-kolla | 12:27 | |
sdake | 2. release management system | 12:27 |
v1k0d3n | so confusion on helm...did that get cleared up? i'm still reading. | 12:27 |
*** neilus has joined #openstack-kolla | 12:27 | |
sdake | 3. replacements for oslo.db and oslo.log and oslo.messaging | 12:27 |
v1k0d3n | operator manages dependancies and complex logic (where hard ip's are needed etc). helm is the deployment tool and catalog search for these items with initial deployment logic. both are important steps. | 12:28 |
sdake | 4. some form of automated testing framework | 12:28 |
sdake | v1k0d3n in that video of etcd_operator, helm was not used | 12:28 |
sdake | instead etcd_oeprator was used | 12:28 |
sdake | so rhallisey - as you can see, its an uphill battle to implement anything in go native without more champions besides just kolla | 12:29 |
rhallisey | wel | 12:29 |
rhallisey | new project in kubernetes? | 12:29 |
sdake | I was thinking we should be prototyping and implementing an operator in ansible | 12:29 |
rhallisey | that's what I was thinking :) | 12:29 |
sdake | sorry ansible/python | 12:30 |
sdake | i recommend not thinking about the end state guys, everyone so wrapped up in the end state, we are not workign on an iterative approach | 12:30 |
sdake | kfox1111 and sbezverk working on an iterative approach | 12:30 |
sdake | big bang software engineering always always fails | 12:30 |
*** b_bezak_ has joined #openstack-kolla | 12:30 | |
rhallisey | agree with an iterative approach, but want to pick the right one | 12:31 |
rhallisey | since we have a bunch of approaches here | 12:31 |
sdake | rhallisey because of openstack CLA, we can't take the code from kolla-kubernetes and place it under some different organization such as CNCF | 12:31 |
rhallisey | we don't take the code | 12:31 |
sdake | i think helm (short term) is an optimization | 12:31 |
*** neilus has quit IRC | 12:31 | |
rhallisey | we write the operators under cncf | 12:31 |
rhallisey | or we wait | 12:31 |
sdake | rhallisey that was step B | 12:32 |
sdake | letsget step A out of the way first | 12:32 |
rhallisey | that's if we go with operators | 12:32 |
sdake | we can do everything in python very easily and very fast | 12:32 |
sdake | without having to ramp up our community on go as a language | 12:32 |
rhallisey | hmm | 12:32 |
sdake | also, for true interop, we need to solve the problems 1-4 above I talked about | 12:32 |
rhallisey | we as in kolla? | 12:33 |
*** zhubingbing has joined #openstack-kolla | 12:33 | |
sdake | kolla may be the best place to take a stab at solving them post ocata | 12:33 |
rhallisey | or openstack? | 12:33 |
v1k0d3n | ok finally caught up. | 12:33 |
sdake | no, we as in openstack | 12:33 |
sdake | but yes, we as in kolla ;) | 12:33 |
rhallisey | erm | 12:33 |
v1k0d3n | so where are we at? | 12:33 |
rhallisey | let's put that on the ml | 12:33 |
sdake | put which on the ml? | 12:34 |
rhallisey | 4 steps before we can achieve interop | 12:34 |
v1k0d3n | not sure exactly why operator is considered as a replacement for helm. they accomplish different things. | 12:34 |
sdake | lets not and say we did | 12:34 |
rhallisey | k | 12:34 |
rhallisey | :/ | 12:34 |
rhallisey | dunno why not | 12:34 |
rhallisey | moving on then | 12:34 |
sdake | v1k0d3n what I'm getting at is the following | 12:35 |
sdake | we dont' immediately need helm | 12:35 |
zhubingbing | hey guys | 12:35 |
sdake | we immediately need an operator | 12:35 |
sdake | hey zhubingbing | 12:35 |
rhallisey | v1k0d3n, not as a replacement, but idk if we need helm yet | 12:35 |
sdake | zhubingbing start learning go plz | 12:35 |
zhubingbing | ok~ | 12:35 |
rhallisey | lol | 12:35 |
zhubingbing | plz ? | 12:35 |
v1k0d3n | well, if you don't have helm...what's the deployment logic? | 12:35 |
sdake | plz = shortcut for please | 12:35 |
zhubingbing | sure | 12:35 |
sdake | v1k0d3n the operator does deployment for etcd | 12:35 |
v1k0d3n | or i should say mechanism | 12:35 |
*** eaguilar has quit IRC | 12:36 | |
rhallisey | we already have the logic with the existing templates | 12:36 |
v1k0d3n | right, but an operator shouldn't deploy every single project that kolla has to offer. | 12:36 |
v1k0d3n | how is an operator going to choose? | 12:36 |
zhubingbing | Go language I will point | 12:36 |
sdake | v1k0d3n via yaml config file? | 12:36 |
zhubingbing | -) | 12:37 |
sdake | zhubingbing in your spare time | 12:37 |
sdake | v1k0d3n we have one of those alreayd called globals.yaml :) | 12:37 |
zhubingbing | roger | 12:37 |
v1k0d3n | can i just for one say that i don't like that approach? | 12:37 |
rhallisey | sure | 12:37 |
v1k0d3n | for kolla-kubernetes... | 12:37 |
v1k0d3n | ansible totally fine. | 12:37 |
sdake | v1k0d3n which approach, reading globals.yaml? | 12:37 |
v1k0d3n | yes. | 12:38 |
rhallisey | no ansible | 12:38 |
rhallisey | ugh | 12:38 |
sdake | which is it | 12:38 |
openstackgerrit | zhubingbing proposed openstack/kolla: add panko role https://review.openstack.org/394298 | 12:38 |
rhallisey | let me just delete that repo | 12:38 |
sdake | v1k0d3n could you expand on why you think a yaml file for configuration is the wrong approach? | 12:38 |
v1k0d3n | we're going full circle on this a lot | 12:39 |
sdake | http://files.enjin.com.s3.amazonaws.com/381916/modules/forum/attachments/squirrelarmy_1401683593.jpg | 12:39 |
v1k0d3n | if helm is used, audience grows beyond kolla. | 12:39 |
sdake | v1k0d3n ok before we get back into helm | 12:39 |
sdake | lets tlak about globals.yaml | 12:39 |
v1k0d3n | the operator model is great....you know i'm a huge fan | 12:39 |
rhallisey | true | 12:40 |
sdake | what part precisely do you not like about globals.yaml as a config file? | 12:40 |
rhallisey | v1k0d3n, btw I'm not saying no to helm | 12:40 |
rhallisey | don't think sdake is either | 12:40 |
*** sdake has quit IRC | 12:40 | |
*** sdake has joined #openstack-kolla | 12:40 | |
rhallisey | just deployment #1 here | 12:40 |
v1k0d3n | just split out some of the work | 12:41 |
v1k0d3n | SAP is already refining a helm deployment. use as a template. | 12:41 |
v1k0d3n | that path is pretty quick. lots interested in that. | 12:41 |
*** berendt has quit IRC | 12:41 | |
*** shardy has quit IRC | 12:41 | |
v1k0d3n | they just pushed a lot of changes today...all the make/precompile logic. | 12:41 |
*** b_bezak_ has quit IRC | 12:41 | |
v1k0d3n | just generalize that. and others who are more python savvy can work on the operators. | 12:42 |
sdake | v1k0d3n can you answer my question about what is wrong with one config file in /etc/kolla? | 12:42 |
*** b_bezak has joined #openstack-kolla | 12:42 | |
sdake | if we can't get the basic problems solved, we are going to have a terrible time collaborating :) | 12:42 |
v1k0d3n | because helm can address that. | 12:42 |
v1k0d3n | has anyone else tried using helm first? | 12:42 |
rhallisey | v1k0d3n, more exposure in helm | 12:42 |
sdake | do you have an example | 12:42 |
v1k0d3n | just an ask... | 12:42 |
rhallisey | I mean sdake * | 12:43 |
sdake | how is there more exposure in helm | 12:43 |
sdake | comon guys feed me :) | 12:43 |
zhubingbing | who can u help me review panko | 12:43 |
sdake | my AI becomes sharper by the moment :) | 12:43 |
rhallisey | helm is the place you get your app in kubernetes | 12:44 |
*** tonanhngo has joined #openstack-kolla | 12:44 | |
sdake | rhallisey is it like pypi? | 12:44 |
*** tonanhngo has quit IRC | 12:44 | |
zhubingbing | https://review.openstack.org/#/c/387232/ | 12:44 |
zhubingbing | https://review.openstack.org/394298 | 12:44 |
v1k0d3n | https://github.com/kubernetes/helm/blob/master/docs/using_helm.md | 12:45 |
sdake | so muc h documentation | 12:45 |
v1k0d3n | the docs need a bit more imo, but right now it's a really good start | 12:45 |
sdake | whisky tango foxtrot tiller | 12:45 |
v1k0d3n | rhallisey: think a bit beyond that though. | 12:46 |
*** b_bezak has quit IRC | 12:46 | |
v1k0d3n | helm gives operators a catalog lookup with initial deploy logic. | 12:47 |
v1k0d3n | ....and.....icing on the cake, the audience widens. | 12:47 |
openstackgerrit | Ryan Hallisey proposed openstack/kolla-kubernetes: Remove ansible dir from kolla-kubernetes https://review.openstack.org/394405 | 12:47 |
zhubingbing | ping Eduardo Gonzalez | 12:47 |
zhubingbing | -) | 12:47 |
v1k0d3n | ultimately, the deployment that makes operators life simple will be the tool that people gravitate to. | 12:48 |
sdake | A Repository is the place where charts can be collected and shared. It's like Perl's CPAN archive or the Fedora Package Database, but for Kubernetes packages. | 12:48 |
sdake | v1k0d3n ya we already figured that out with kolla-ansible ;) | 12:48 |
v1k0d3n | and if you remember, what i said when we were at the summit is that the community is fractured when it comes to container deployments of openstack. this takes back that command. | 12:48 |
sdake | community is not fractured | 12:49 |
*** egonzalez90 has quit IRC | 12:49 | |
sdake | there are 4-5 cats working in silos | 12:49 |
sdake | in 2-3 silos | 12:49 |
v1k0d3n | sdake: kolla-ansible is why there are 3 deployments of openstack/kubernetes. | 12:49 |
*** egonzalez90 has joined #openstack-kolla | 12:49 | |
sdake | and 500 people working on kolla | 12:49 |
v1k0d3n | but those 2 out of the 3 are trying to be solutions at corp scale. | 12:49 |
sdake | v1k0d3n you are wrong | 12:49 |
sdake | kolla-ansible is not why there are 3 deployments of opensstack/kubernetes | 12:49 |
v1k0d3n | no. i'm not. because i'm sitting at the other end of these sales pitches. | 12:49 |
sdake | dont confuse marketing with the truth | 12:50 |
sdake | the reality is this: | 12:50 |
sdake | kolla-ansible trounced other openstack deployment tools | 12:50 |
v1k0d3n | the reality is the one that ends up in our shop; the one that our execs would believe at the end of the day. | 12:50 |
sdake | these siloed groups are trying to compete back on new ground with kubernetes | 12:50 |
v1k0d3n | from what i have been told everywhere, it's not in there interest to maintain container builds. | 12:51 |
sdake | well lets agree to disagree then and move on | 12:51 |
v1k0d3n | this is a really big part though. | 12:51 |
v1k0d3n | i'm telling each "meet us in kolla". | 12:51 |
sdake | what does container builds have to do with it? | 12:52 |
sdake | v1k0d3n kolla is 3 things -> containers, kolla-ansible, kolla-kubernetes | 12:52 |
sdake | it isn't containers only | 12:53 |
sdake | I thought we covered that last night ;) | 12:53 |
sdake | the fact that 3 siloed implementations of kubernetes exists has zero to do with kolla-ansible | 12:53 |
sdake | if anything, they followed our lead and are attempting a leapfrog | 12:53 |
v1k0d3n | yeah, but if kolla's container strategy isn't willing to budge because of ansible, i think this holds back opportunities in kolla-kubernetes. | 12:54 |
sdake | v1k0d3n we never said that | 12:54 |
v1k0d3n | it's always about leapfrogging if you want to sell something | 12:54 |
sdake | v1k0d3n anyway with operators the container implementation doesn't need to change | 12:54 |
sdake | v1k0d3n good, I'm glad we are in agreement that kolla-ansible is not the "cause" of the silos | 12:56 |
sdake | in fact, the cause of the silos is those companies wanting to compete with kolla | 12:56 |
v1k0d3n | you guys are good with operators now? because regardless, i think that's a huge part. | 12:56 |
sdake | as you so eloquently put it :) | 12:56 |
sdake | v1k0d3n i think thats a done deal in my mind | 12:56 |
v1k0d3n | no, i don't think coreos really wants to maintain their own. i really don't. | 12:57 |
v1k0d3n | the other...well...that's probably a lot more complicated as we know :) | 12:57 |
*** dwalsh has joined #openstack-kolla | 12:57 | |
sdake | coreos doesn't want to maintain their own kubernetes implementation you mean? | 12:58 |
sdake | rather openstack on kubernetes? | 12:58 |
v1k0d3n | huh? | 13:00 |
v1k0d3n | they do not want to maintain their own container images of openstack <<< | 13:01 |
rhallisey | eh | 13:01 |
rhallisey | want to have a ghangout talk? | 13:01 |
sdake | not really | 13:01 |
v1k0d3n | it's MUCH easier to let "openstack" do the container image creation. | 13:01 |
sdake | irc is better | 13:01 |
rhallisey | well there seems to be a lot of mis communication here | 13:02 |
v1k0d3n | since we are talking about kubernetes guys, i really think we should bring to the kubernetes SIGS | 13:02 |
sdake | you guys jump around so much, I have a hard time keeping track of the comms :) | 13:02 |
sdake | irc slows that down so i can keep track of the various threads floating aroudn inside your minds | 13:02 |
v1k0d3n | that's probably the way to go. otherwise we're not getting that advice and we are certain to deviate down our own path. | 13:02 |
sdake | v1k0d3n kolla community produces its own distinct work | 13:03 |
sdake | v1k0d3n while the kubernetes sig that sgordon runs is interesting, i'd rather us see what we come up with on our own | 13:03 |
sdake | so, back on helm... | 13:03 |
*** tonanhngo has joined #openstack-kolla | 13:04 | |
v1k0d3n | why on our own? | 13:05 |
v1k0d3n | i don't understand that. | 13:05 |
*** tonanhngo has quit IRC | 13:05 | |
v1k0d3n | if we want people to use openstack, they go to openstack | 13:05 |
v1k0d3n | if kubernetes they go to kubernetes | 13:05 |
sdake | openstack on kubernetes, do they go to a sig? | 13:05 |
*** ccesario has joined #openstack-kolla | 13:05 | |
sdake | ok, so back on helm.. | 13:06 |
sdake | can we argue about one thing at a time :) | 13:06 |
v1k0d3n | huh? | 13:06 |
sdake | helm looks like it provides a pypi type repo thing | 13:06 |
sdake | A Repository is the place where charts can be collected and shared. It's like Perl's CPAN archive or the Fedora Package Database, but for Kubernetes packages. | 13:06 |
sdake | v1k0d3n is that accurate? | 13:06 |
v1k0d3n | it is a repo as well. yes. | 13:06 |
sdake | v1k0d3n if so, who hosts this repository? | 13:06 |
pprokop | you can host it locally | 13:07 |
v1k0d3n | think apt, for kubernetes and that is similar to helm. but helm is also part of the development pipeline. | 13:07 |
v1k0d3n | end user. | 13:07 |
pprokop | just like docker registry | 13:07 |
v1k0d3n | i think it would be good to reach out to helm to see the benefits guys. | 13:07 |
sdake | but there is no global hosting? | 13:07 |
v1k0d3n | there is a huge miss here. | 13:07 |
v1k0d3n | there can be global hosting as well, yes | 13:07 |
pprokop | kolla doesn't host images globally too, i guess | 13:07 |
sdake | is anyon doing that? | 13:07 |
*** lamt has joined #openstack-kolla | 13:07 | |
v1k0d3n | that's the kubernetes repo/incubator. | 13:08 |
sdake | pprokop in dockerhub we do | 13:08 |
v1k0d3n | jenkins can do this whole chart creation for the developers. | 13:08 |
v1k0d3n | so there's not a massive amount of debt after creation. | 13:08 |
sdake | v1k0d3n jenkins is dead dude | 13:08 |
v1k0d3n | oh my.... | 13:08 |
v1k0d3n | ok | 13:08 |
sdake | or on its way to death | 13:08 |
pprokop | I don't think right now there is any commercial hosting of helm | 13:09 |
v1k0d3n | ok....use what you want. i don't care. example steve | 13:09 |
sdake | v1k0d3n huh? | 13:09 |
v1k0d3n | pprokop: the commercial hosting, i think you're talking about what is in the kubernetes repo. | 13:09 |
pprokop | but anyway most of environments wants to have a local registry | 13:09 |
v1k0d3n | there is stable and incubator. | 13:09 |
pprokop | I mean docker-registry for helm charts | 13:10 |
v1k0d3n | pprokop: i am so confused now..... | 13:10 |
v1k0d3n | docker registry for helm charts? | 13:10 |
pprokop | https://helm.sh/ | 13:10 |
sdake | v1k0d3n thinking about 20 different things at onc ewill do that :) | 13:11 |
openstackgerrit | zhubingbing proposed openstack/kolla: add panko role https://review.openstack.org/394298 | 13:11 |
*** lamt has quit IRC | 13:11 | |
pprokop | okay my bad | 13:12 |
pprokop | https://console.cloud.google.com/storage/browser/kubernetes-charts/ | 13:12 |
pprokop | https://github.com/kubernetes/charts | 13:13 |
*** b_bezak has joined #openstack-kolla | 13:13 | |
v1k0d3n | pprokop: yes, exactly. | 13:14 |
v1k0d3n | think we're on the same page now | 13:14 |
pprokop | You guys are too multi-task for me :D | 13:15 |
v1k0d3n | this is so true :) | 13:16 |
openstackgerrit | Andrew Widdersheim proposed openstack/kolla: Allow cinder-volume to be configured to use NFS https://review.openstack.org/279913 | 13:17 |
sdake | pprokop ya, thats why i prefer irc | 13:17 |
sdake | pprokop one thing at a time + incremental = winning | 13:17 |
pprokop | Right now helm simplifies deployment of kolla-kubernetes | 13:21 |
pprokop | cause it is a standard | 13:21 |
pprokop | not another tool | 13:21 |
pprokop | so one can reuse many thing | 13:22 |
*** jheroux has joined #openstack-kolla | 13:22 | |
openstackgerrit | Vladislav Belogrudov proposed openstack/kolla: Exclude port from ceilometer_database_mysql_address https://review.openstack.org/394424 | 13:22 |
*** dave-mccowan has joined #openstack-kolla | 13:23 | |
*** b_bezak has quit IRC | 13:23 | |
v1k0d3n | pprokop: it's more than that. there is a development framework that could be leveraged as well with charts. | 13:24 |
pprokop | can you gie a link ? | 13:25 |
v1k0d3n | we're missing the huge value with that too, if we don't use helm. | 13:25 |
v1k0d3n | i have to leave for kubecon. | 13:25 |
v1k0d3n | already running late. | 13:25 |
v1k0d3n | look for charts in dies repos. | 13:25 |
openstackgerrit | Mauricio Lima proposed openstack/kolla: Move manila-share service to network node https://review.openstack.org/393469 | 13:25 |
pprokop | Thanks. Have a nice trip. | 13:25 |
rhallisey | v1k0d3n, have a safe journey | 13:25 |
v1k0d3n | also reach out to "levo" in deis slack group. | 13:25 |
v1k0d3n | lachie is a huge openstacker | 13:26 |
*** b_bezak has joined #openstack-kolla | 13:30 | |
*** fguillot has joined #openstack-kolla | 13:32 | |
sdake | see ya v1k0d3n | 13:33 |
*** dwalsh_ has joined #openstack-kolla | 13:34 | |
*** tonanhngo has joined #openstack-kolla | 13:34 | |
*** tonanhngo has quit IRC | 13:35 | |
*** dwalsh has quit IRC | 13:36 | |
*** lamt has joined #openstack-kolla | 13:38 | |
*** dwalsh_ has quit IRC | 13:42 | |
*** saneax-_-|AFK is now known as saneax | 13:43 | |
*** v1k0d3n has quit IRC | 13:43 | |
*** neilus has joined #openstack-kolla | 13:45 | |
*** inc0 has joined #openstack-kolla | 13:46 | |
inc0 | good morning | 13:47 |
egonzalez90 | Morning inc0 | 13:47 |
inc0 | I'm back to action :) | 13:47 |
sdake | inc0 welcome to the party | 13:47 |
sdake | inc0 i'm out | 13:47 |
sdake | BYE | 13:47 |
*** sdake has quit IRC | 13:48 | |
inc0 | lol | 13:48 |
*** neilus has quit IRC | 13:50 | |
inc0 | so, anything particularly interesting happened last week? | 13:50 |
*** fguillot has quit IRC | 13:50 | |
openstackgerrit | Merged openstack/kolla: Load all sections defined in code and conf file https://review.openstack.org/393896 | 13:50 |
openstackgerrit | Merged openstack/kolla: Run collect static during container startup when necessarily https://review.openstack.org/394020 | 13:51 |
*** saneax has left #openstack-kolla | 13:53 | |
*** fguillot has joined #openstack-kolla | 13:54 | |
*** dwalsh_ has joined #openstack-kolla | 13:55 | |
*** srwilkers has quit IRC | 13:56 | |
*** neilus has joined #openstack-kolla | 13:57 | |
*** skramaja_ has joined #openstack-kolla | 14:00 | |
*** skramaja has quit IRC | 14:00 | |
*** caowei has quit IRC | 14:01 | |
*** caowei has joined #openstack-kolla | 14:03 | |
openstackgerrit | zhubingbing proposed openstack/kolla: add panko role https://review.openstack.org/394298 | 14:05 |
*** skramaja has joined #openstack-kolla | 14:06 | |
*** skramaja_ has quit IRC | 14:06 | |
*** zhurong has quit IRC | 14:07 | |
zhubingbing | sdake aroud | 14:14 |
inc0 | zhubingbing, he won't be around for few days | 14:19 |
zhubingbing | ok | 14:19 |
zhubingbing | inco | 14:19 |
zhubingbing | can u help me review it ? | 14:20 |
zhubingbing | https://review.openstack.org/#/c/383997/ | 14:20 |
zhubingbing | thanks | 14:20 |
*** mgoddard has quit IRC | 14:22 | |
*** duonghq has joined #openstack-kolla | 14:22 | |
duonghq | egonzalez90, thank you very much | 14:22 |
inc0 | zhubingbing, what is stored in /var/lib/kolla/collectd | 14:22 |
inc0 | ? | 14:22 |
zhubingbing | Collected information? | 14:24 |
inc0 | what information are these? that's my question | 14:25 |
inc0 | we don't add named volume there | 14:25 |
*** eaguilar has joined #openstack-kolla | 14:26 | |
*** shardy has joined #openstack-kolla | 14:27 | |
zhubingbing | memory disk and other information | 14:27 |
*** mgoddard has joined #openstack-kolla | 14:27 | |
zhubingbing | so, we don't add /var/lib/kolla/colletcd ? | 14:28 |
*** yingjun has joined #openstack-kolla | 14:28 | |
duonghq | pbourke, egonzalez90, thanks I'll recheck tommorow | 14:32 |
*** lrensing has joined #openstack-kolla | 14:33 | |
duonghq | wish that somebody can review my bp implementation: https://review.openstack.org/#/c/358539/21 | 14:34 |
*** tonanhngo has joined #openstack-kolla | 14:34 | |
*** pprokop has quit IRC | 14:36 | |
*** tonanhngo has quit IRC | 14:36 | |
inc0 | zhubingbing, this dir will be deleted after container redeploy | 14:39 |
inc0 | so we need to make sure that it won't hold anything persistant | 14:39 |
inc0 | if it does, we need to make it named volume | 14:39 |
zhubingbing | roger | 14:40 |
inc0 | and then we need to keep chown code | 14:40 |
inc0 | for upgrades | 14:40 |
duonghq | inc0, I didn't notice bp/ks-rolling-upgrade-role before, so I partial implenent something like it in ps: https://review.openstack.org/#/c/388544/ wish that you can take a look at this | 14:41 |
lrensing | good morning | 14:42 |
duonghq | morning lrensing | 14:42 |
inc0 | duonghq, will do! | 14:44 |
inc0 | just need to wake up before I start reviewing code;) | 14:44 |
duonghq | thank inc0 | 14:44 |
duonghq | just drop a note | 14:44 |
*** eaguilar has quit IRC | 14:45 | |
*** srwilkers has joined #openstack-kolla | 14:45 | |
inc0 | srwilkers, lrensing ping | 14:46 |
lrensing | hello | 14:46 |
*** eaguilar has joined #openstack-kolla | 14:46 | |
inc0 | hey guys:) so qwang and jascott1 (probably still asleep) will work with you on helm:) | 14:46 |
inc0 | between 4 of you we should be able to make PoC soon | 14:47 |
lrensing | https://blueprints.launchpad.net/kolla-kubernetes/+spec/helm-orchestration this is the blueprint, right? | 14:47 |
inc0 | thats the one | 14:48 |
*** egonzalez has joined #openstack-kolla | 14:49 | |
*** egonzalez90 has quit IRC | 14:49 | |
*** MarMat has joined #openstack-kolla | 14:49 | |
*** v1k0d3n has joined #openstack-kolla | 14:52 | |
srwilkers | inc0: sounds good | 14:52 |
v1k0d3n | inc0: hey man | 14:52 |
v1k0d3n | did you catch up with all the reading from this moring? :) | 14:52 |
inc0 | nope | 14:53 |
inc0 | still kicking off my brain | 14:53 |
inc0 | caffeine is pouring | 14:53 |
inc0 | so I talked with snap guys, they'll help us | 14:53 |
inc0 | we might even get sth as soon as Dec | 14:53 |
v1k0d3n | if you're like me and srwilkers...we need caffeine iv's | 14:53 |
*** pprokop has joined #openstack-kolla | 14:54 | |
srwilkers | +1 | 14:54 |
v1k0d3n | oh that's awesome! | 14:54 |
*** stvnoyes has joined #openstack-kolla | 14:54 | |
inc0 | I have mountain dew kick start (0.5l one...) | 14:54 |
duonghq | I'm always wish that I can try coffee from as many as country posible | 14:54 |
inc0 | almost the same | 14:54 |
duonghq | how do you brew coffee? | 14:55 |
inc0 | I'm not coffee drinker..I'm a tea person tbh | 14:56 |
inc0 | (and I'm miserable in US...) | 14:56 |
*** f13o has quit IRC | 14:56 | |
v1k0d3n | lol | 14:56 |
rhallisey | inc0, nice :) | 14:56 |
v1k0d3n | you can get decent tea at world market. | 14:56 |
v1k0d3n | and the areo brew is awesome btw :) | 14:57 |
rhallisey | about the tea :) | 14:57 |
inc0 | I buy my tea in online store | 14:57 |
inc0 | there is one good one | 14:57 |
v1k0d3n | yeah, makes sense. | 14:57 |
inc0 | but I'm yet to find coffee shop with good tea | 14:58 |
duonghq | which type of tea do you like inc0? | 14:58 |
inc0 | or anywhere to drink something decent outside home;) | 14:58 |
inc0 | duonghq, I drink all;) but earl grey would be my choice of every day tea | 14:58 |
*** caowei has quit IRC | 14:59 | |
v1k0d3n | this is how i brew tea (i'm a lazy 'merican): https://www.bedbathandbeyond.com/store/product/adagio-teas-ingenuitea-16-ounce-teapot/1041716847?skuId=41716847&mcid=PS_googlepla_nonbrand_coffeetea_&product_id=41716847&adpos=1o5&creative=43742634109&device=c&matchtype=&network=g&gclid=CODdho70ltACFZVahgodlx0A3Q | 14:59 |
duonghq | in my city, most people go to so-called coffee shop for drink non-coffee, lol | 14:59 |
inc0 | yeah, in Poland we have tea-shops | 14:59 |
inc0 | with decent stuff | 15:00 |
inc0 | I have couple infusers;) | 15:00 |
inc0 | and best thing ever - a kettle that heats up water to temperature I want | 15:00 |
srwilkers | i just chew coffee grounds | 15:00 |
srwilkers | straight to the head | 15:00 |
duonghq | it's hard to find good infuser in my city, I'm looking for that for at least 6 months | 15:00 |
inc0 | so I can have proper 85C for green tea;) | 15:00 |
inc0 | lol | 15:01 |
inc0 | this country needs legalized amphetamine;) | 15:01 |
inc0 | would boost up economy:D | 15:01 |
srwilkers | just pump it into the water | 15:01 |
srwilkers | can we draft a spec for that? | 15:01 |
duonghq | in my country, people use tea pot, | 15:02 |
duonghq | with green tea, | 15:02 |
duonghq | something like this: http://chebuptancuong.com/wp-content/uploads/2015/10/am-tra-0.jpg | 15:03 |
duonghq | but tea infusers is much more convenient at office | 15:03 |
inc0 | duonghq, you're from vietnam afair right? | 15:03 |
duonghq | right | 15:03 |
inc0 | yeah, but infusers in offices in US are lousy | 15:04 |
duonghq | any reason? | 15:04 |
inc0 | I bring my own stuff to office, and black tea not green tea because water is too hot | 15:04 |
inc0 | US is coffee culture | 15:04 |
inc0 | lots of brewed quality coffee | 15:05 |
inc0 | and some cheap tea because nobody (besides me) drinks it anyway | 15:05 |
duonghq | which tea brand do you use? | 15:05 |
v1k0d3n | all this talk of tea...and coffee. shows what our priorities are gentlemen :) | 15:05 |
lrensing | that depends if you go to teavana or not inc0 | 15:05 |
inc0 | which I think should be revisited given how many asian people now works in IT | 15:05 |
inc0 | lrensing, I bought some tea from there once | 15:05 |
inc0 | it was overpriced and not that good | 15:06 |
lrensing | agreed | 15:06 |
inc0 | https://www.tealyra.com <- this is where I get my stuff now | 15:06 |
v1k0d3n | inc0: has a sensitive palette. | 15:06 |
srwilkers | v1k0d3n: just caught up on all the reading from earlier | 15:06 |
inc0 | lol | 15:06 |
v1k0d3n | :) | 15:06 |
inc0 | especially in my wallet | 15:06 |
v1k0d3n | don't we all! | 15:06 |
v1k0d3n | isn't tevanna starbucks? or something else? | 15:07 |
lrensing | i buy my coffee from https://www.happymugcoffee.com/ | 15:07 |
v1k0d3n | * teavana | 15:07 |
lrensing | their site looks pretty terrible | 15:07 |
lrensing | but they have good coffee, and a good tea selection | 15:07 |
srwilkers | starbucks bought teavana | 15:07 |
v1k0d3n | i get my coffee like i get my tech....amazon | 15:07 |
v1k0d3n | to me, i've found that the way i make coffee is everything. | 15:08 |
lrensing | pour over master race | 15:08 |
v1k0d3n | just get the roast right and i will do the rest. | 15:08 |
v1k0d3n | yeah, pour over you need really good coffee. that lets coffee do the work for sure. | 15:08 |
v1k0d3n | again, i digress to the iv statement earlier. | 15:09 |
v1k0d3n | so inc0 have you caught up on the whole operator and helm topics? | 15:09 |
inc0 | reading now | 15:09 |
v1k0d3n | ok. i think we own kolla a demo of helm, and possibly the development pipeline. i am trying to find a video on this, but lrensing and srwilkers were on a call earlier this week that talked about the charts development and dev pipeline. | 15:10 |
v1k0d3n | there's a LOT of value in this approach and it's a lot more than surface deep. | 15:11 |
inc0 | yeah, as I said above, let's get PoC up asap | 15:11 |
v1k0d3n | well, what about SAP? | 15:12 |
v1k0d3n | they already have a demo they are sharing with the helm community. | 15:12 |
v1k0d3n | would that work? | 15:12 |
inc0 | I need to dig through their code, but afair they didn't exactly solve dependency issue | 15:12 |
v1k0d3n | they did something huge this morning; removed all the make/prebuilds and now it's more "helm" native. | 15:13 |
v1k0d3n | so dependency issue will ALWAYS exists with kubernetes. i am more and more convinced of this. | 15:13 |
v1k0d3n | that's where the operator concept comes into play, but they are two separate things to solve two separate issues. | 15:13 |
v1k0d3n | they updated this morning fyi see latest notes: https://github.com/sapcc/openstack-helm | 15:14 |
inc0 | so pprokop had this entrypoint that solves dependenc | 15:14 |
inc0 | ies | 15:14 |
inc0 | add that to liveness probes and we can deal with deps | 15:15 |
v1k0d3n | hve you seen the recent news with coreos? | 15:15 |
v1k0d3n | i know...this stuff gets really hard to keep up with. | 15:15 |
v1k0d3n | changes frequently. | 15:15 |
pprokop | it's just an application which can handle etcd lifecycle | 15:15 |
pprokop | and interact with k8s | 15:15 |
v1k0d3n | https://coreos.com/blog/introducing-operators.html | 15:16 |
v1k0d3n | yes, but i'm referring to the model | 15:16 |
pprokop | yes but for openstack it means write an operator for each service | 15:16 |
*** sean-k-mooney has joined #openstack-kolla | 15:16 | |
v1k0d3n | approach. i know it's not a drop in replacement. it's just showing the basic framework for what would be an operator (controller) per service that isn't cloud-native....the usual suspects. | 15:16 |
v1k0d3n | i don't think so. | 15:16 |
v1k0d3n | some services are just an API, and don't care. | 15:17 |
pprokop | Yep. | 15:17 |
v1k0d3n | those are cloud native. spin up as many as you want, and its handled via the RC | 15:17 |
v1k0d3n | but rabbit, memcache, galera/maria...these are the problems. | 15:17 |
v1k0d3n | those need "operators" in my example. | 15:17 |
pprokop | +1 | 15:18 |
v1k0d3n | since they need hard ip knowledge, etc. | 15:18 |
v1k0d3n | entrypoint is good! don't get my wrong. poor mans "operator" container per service. :) | 15:18 |
pprokop | it wasn't meant to be an operator | 15:18 |
pprokop | :D | 15:18 |
inc0 | v1k0d3n, so issue we need to consider is our containers *has to remain* compatible with ansible | 15:19 |
pprokop | It was developed to do one thing (maybe two) | 15:19 |
inc0 | and non-k8s model | 15:19 |
inc0 | but I think we can pull that off | 15:19 |
rhallisey | inc0, yes, but we can template this stuff | 15:20 |
pprokop | I am more the confident that for most services dependencies can be handled via init container | 15:20 |
inc0 | not really, template will work on container build | 15:20 |
pprokop | so just template manifest | 15:20 |
rhallisey | inc0, yes exactly | 15:20 |
inc0 | and I don't want to have different containers per orch engine | 15:20 |
rhallisey | you include the scripts for kolla-k9s | 15:20 |
rhallisey | you are going to have to in this case | 15:20 |
v1k0d3n | here's better info (obviously etcd specific, but that is the point): https://github.com/coreos/etcd-operator/tree/master/doc/design | 15:20 |
rhallisey | because kolla-k8s orchestartion is done in the container | 15:21 |
inc0 | not necessarly if we make kolla-start intelligent | 15:21 |
rhallisey | this is a mix internal/external mix model | 15:21 |
v1k0d3n | ^^ yes | 15:21 |
rhallisey | inc0, that will cover the entrypoint case | 15:21 |
v1k0d3n | i think that's where some of the issues creep in. | 15:21 |
rhallisey | but it would stil vary | 15:22 |
v1k0d3n | because ansible needs an orchestrator (opinionated to ansible...which is fine), but kubernetes needs container self-awareness logic. | 15:22 |
inc0 | so etcd under k8s is meant to have full information about cluster status | 15:22 |
*** TxGirlGeek has joined #openstack-kolla | 15:22 | |
inc0 | ? | 15:22 |
pprokop | yes | 15:22 |
pprokop | operator is deployed as pod on k8s | 15:22 |
v1k0d3n | remember, only in it's namespace | 15:22 |
pprokop | and queries api | 15:22 |
v1k0d3n | so kolla namespace, and that's ok | 15:22 |
pprokop | to see state | 15:22 |
v1k0d3n | ^^ yes. | 15:23 |
inc0 | yeah, so we have at least 3 sources of truth here - k8s itself, operator and actual state of nodes | 15:23 |
v1k0d3n | and steve's argument (valid, i get it) is that kolla already has it's own API. | 15:23 |
v1k0d3n | so perhaps that can be leveraged? | 15:23 |
v1k0d3n | or modified? | 15:23 |
rhallisey | https://github.com/openstack/kolla/blob/master/docker/keystone/keystone/extend_start.sh#L3 | 15:23 |
*** v1k0d3n has quit IRC | 15:23 | |
rhallisey | kolla-k8s started with a similar model | 15:23 |
inc0 | we can run into all sorts of funny stuff - especially before (if ever) operator becomes de facto standard | 15:24 |
*** b_bezak has quit IRC | 15:24 | |
rhallisey | I think the operator container looks pretty good imo | 15:24 |
inc0 | 1 tool will work on k8s, another on operator | 15:24 |
*** v1k0d3n has joined #openstack-kolla | 15:24 | |
inc0 | we kinda have similar container already - fencing pod | 15:24 |
inc0 | PoC quality, but still | 15:24 |
*** b_bezak has joined #openstack-kolla | 15:24 | |
rhallisey | ya that's a similar idea | 15:25 |
inc0 | however tbh I'd wait till idea becomes more...stable | 15:25 |
v1k0d3n | yeah, so like i said...perhaps this could be leveraged but not sure if this would remain exactly the same or not to fit an operator model | 15:25 |
inc0 | right now it seems to me that every company in the bunch have their own idea about managing stateful apps in k8s | 15:25 |
inc0 | and none of it is perfect | 15:25 |
inc0 | I'm ok with changing it if a) will remain backward compatible and b) will be upgradable | 15:26 |
v1k0d3n | i asked steve for an architecture of this api. | 15:26 |
inc0 | I'm pretty sure we can achieve both | 15:26 |
v1k0d3n | tbh, that is really important. | 15:26 |
inc0 | it's really simple | 15:26 |
v1k0d3n | would be nice to include with the docs | 15:27 |
inc0 | https://github.com/openstack/kolla/blob/master/ansible/roles/nova/templates/nova-compute.json.j2 | 15:27 |
inc0 | you just specify files you want to copy and command you want to run | 15:27 |
inc0 | so as Steve said, we had similar experiment already | 15:28 |
inc0 | with mesos and zookeeper | 15:28 |
v1k0d3n | well, i think the save snapshot is pretty key | 15:28 |
zhubingbing | inco https://review.openstack.org/#/c/353261/ | 15:28 |
zhubingbing | please remove -2 | 15:28 |
zhubingbing | inc0 | 15:29 |
zhubingbing | thanks | 15:29 |
*** b_bezak has quit IRC | 15:29 | |
inc0 | done | 15:29 |
zhubingbing | ok | 15:29 |
*** eaguilar has quit IRC | 15:29 | |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Wait the wsrep_ready to be ON in mariadb https://review.openstack.org/394368 | 15:30 |
inc0 | v1k0d3n, so yeah, this looks like fencing pod we have | 15:31 |
inc0 | I fundamentally mistrust any external tool that queries api and do stuff | 15:31 |
v1k0d3n | that's ok. i get that. | 15:31 |
inc0 | as it might run into all sorts of funny race conditions | 15:31 |
v1k0d3n | i'm just asking to use a similar framework for kubernetes | 15:31 |
inc0 | alternative idea would be k8s-native hooks that are called whenever something happends | 15:32 |
v1k0d3n | obviously just for kolla, and specifically for the individual services. | 15:32 |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Load all sections defined in code and conf file https://review.openstack.org/394494 | 15:32 |
*** eaguilar has joined #openstack-kolla | 15:32 | |
v1k0d3n | it doesn't need to be an openstack-wide fencing solution. i feel like a microservice-based operator model is easier to maintain long term, and more flexible. | 15:33 |
inc0 | well, there are few things that needs fencing-like behaviour tho | 15:33 |
v1k0d3n | what coreos is talking about is just "showing others" how to build these operators. they're not building them on behalf of others....so i don't think there's a concern there. | 15:33 |
inc0 | and anything with volume (so anything with data-persistence) would also be vulnerable | 15:33 |
*** tonanhngo has joined #openstack-kolla | 15:34 | |
v1k0d3n | i agree. | 15:35 |
v1k0d3n | i think that each component should have it's own volume claim though. | 15:35 |
v1k0d3n | that's better even from a security model. | 15:35 |
*** absubram has joined #openstack-kolla | 15:36 | |
openstackgerrit | Jeffrey Zhang proposed openstack/kolla: NO_MERGE: test mitaka branch https://review.openstack.org/393877 | 15:38 |
v1k0d3n | inc0: you may or may not find this interesting too: https://github.com/coreos/prometheus-operator/blob/master/Documentation/service-monitor.md | 15:40 |
v1k0d3n | for instance the parts at the bottom: Currently, targets are always discovered from all namespaces. In the future, the ServiceMonitor should allow to restrict this to one or more namespaces. How such a configuration would look like, i.e. explicit namespaces, selection by labels, or both, and what the default behavior should be is still up for discussion. | 15:40 |
*** harlowja_at_home has joined #openstack-kolla | 15:41 | |
*** serverascode has quit IRC | 15:41 | |
*** coolsvap has quit IRC | 15:42 | |
v1k0d3n | an interesting scenario is using kubernetes federated API for kolla operators: https://github.com/kubernetes/kubernetes/blob/master/docs/proposals/federated-api-servers.md | 15:42 |
*** cloudnull has quit IRC | 15:42 | |
*** alanmeadows has quit IRC | 15:42 | |
*** smekel_ has quit IRC | 15:42 | |
*** amaged__ has quit IRC | 15:42 | |
*** mgkwill has quit IRC | 15:42 | |
*** qwang has quit IRC | 15:42 | |
*** zhenguo has quit IRC | 15:42 | |
*** fguillot has quit IRC | 15:43 | |
*** mkoderer has quit IRC | 15:43 | |
*** Kdecherf has quit IRC | 15:43 | |
*** classicsnail has quit IRC | 15:43 | |
*** mrhillsman has quit IRC | 15:43 | |
Pavo | morning everyone | 15:45 |
*** coolsvap has joined #openstack-kolla | 15:45 | |
*** zhenguo has joined #openstack-kolla | 15:46 | |
*** mgkwill has joined #openstack-kolla | 15:46 | |
*** qwang has joined #openstack-kolla | 15:46 | |
*** smekel_ has joined #openstack-kolla | 15:47 | |
*** serverascode has joined #openstack-kolla | 15:47 | |
*** alanmeadows has joined #openstack-kolla | 15:48 | |
*** amaged__ has joined #openstack-kolla | 15:48 | |
v1k0d3n | morning Pavo | 15:49 |
v1k0d3n | where is kfox1111 lately? | 15:49 |
*** fguillot has joined #openstack-kolla | 15:50 | |
v1k0d3n | what are your thoughts on all of these discussions? curious. | 15:50 |
*** cloudnull has joined #openstack-kolla | 15:51 | |
openstackgerrit | zhubingbing proposed openstack/kolla: Ansible-ize OpenStack Designate https://review.openstack.org/353261 | 15:52 |
*** f13o has joined #openstack-kolla | 15:52 | |
lrensing | v1k0d3n: what sort of prechecks do you think would be valuable to have in kolla-kubernetes? | 15:52 |
*** dave-mccowan has quit IRC | 15:53 | |
*** classicsnail has joined #openstack-kolla | 15:53 | |
*** v1k0d3n has quit IRC | 15:54 | |
*** mkoderer has joined #openstack-kolla | 15:54 | |
*** mrhillsman has joined #openstack-kolla | 15:55 | |
rhallisey | lrensing, go ahead and post what you have | 15:55 |
rhallisey | the community can iterate on it | 15:55 |
*** Kdecherf has joined #openstack-kolla | 15:56 | |
*** radeksmg has joined #openstack-kolla | 15:56 | |
*** HyperJohnGraham has quit IRC | 15:57 | |
*** g3ek has quit IRC | 15:58 | |
*** Guest92 has joined #openstack-kolla | 15:58 | |
*** neilus has quit IRC | 15:58 | |
*** Kimmo__ has quit IRC | 16:00 | |
*** mgoddard_ has joined #openstack-kolla | 16:01 | |
*** Guest92_ has joined #openstack-kolla | 16:01 | |
*** neilus has joined #openstack-kolla | 16:01 | |
*** v1k0d3n has joined #openstack-kolla | 16:02 | |
*** msimonin has left #openstack-kolla | 16:02 | |
*** g3ek has joined #openstack-kolla | 16:03 | |
v1k0d3n | lrensing: i think most prechecks can be done via kubernetes api once that is up. | 16:03 |
v1k0d3n | even checking to see if kubernetes itself exists can be done through api queries. | 16:04 |
v1k0d3n | are prechecks ansible? | 16:05 |
*** Guest92 has quit IRC | 16:05 | |
*** mgoddard has quit IRC | 16:05 | |
*** dave-mccowan has joined #openstack-kolla | 16:06 | |
*** neilus has quit IRC | 16:06 | |
inc0 | v1k0d3n, prechecks are ansible | 16:08 |
inc0 | but thats just playbook | 16:08 |
inc0 | and totally optional | 16:08 |
inc0 | we can do it with k8s, or not do it at all and let k8s handle it | 16:09 |
*** v1k0d3n has quit IRC | 16:09 | |
*** v1k0d3n has joined #openstack-kolla | 16:10 | |
v1k0d3n | inc0: yeah, that's where i was going. are the pre-checks really needed for k8s? | 16:11 |
v1k0d3n | i mean...they ARE, but the approach is different. | 16:11 |
inc0 | not really imho | 16:11 |
v1k0d3n | ^^ agreed. | 16:11 |
v1k0d3n | simple is k8s available, are persistant volumes available, etc | 16:12 |
openstackgerrit | zhubingbing proposed openstack/kolla: Ansible-ize OpenStack Designate https://review.openstack.org/353261 | 16:12 |
v1k0d3n | looks different. so what lrensing is working on was before we started talking about entrypoints and operator logic, right? | 16:12 |
v1k0d3n | (assumption here) | 16:12 |
rhallisey | anything that is environment specific | 16:13 |
rhallisey | so any thing that kolla-k8s documents as part of setup | 16:13 |
rhallisey | those would be prechecked | 16:13 |
rhallisey | I think that will be more defined as we define our orchestartion | 16:13 |
rhallisey | there are also sanity checks that could go on top of that | 16:14 |
inc0 | yeah, my approach would be - write deployment while keeping upgradability in mind | 16:14 |
inc0 | all the rest - we'll deal with it as we go | 16:14 |
sean-k-mooney | o/ how are everyone today | 16:15 |
inc0 | stuff that makes sense in ansible might not make sense in k8s and vice versa | 16:15 |
rhallisey | for sure | 16:15 |
inc0 | good morning Sean:) | 16:15 |
v1k0d3n | inc0: ...this is going to be my new twitter quote of the day "stuff that makes sense in ansible might not make sense in k8s and vice versa" OMG YES | 16:16 |
inc0 | I'm all for what makes sense | 16:16 |
inc0 | I don't care about "omg omg this has to look like ansible" | 16:16 |
inc0 | no it doesn't | 16:16 |
v1k0d3n | burn it all to the ground like our elections? yes. | 16:16 |
v1k0d3n | (clearly...a joke here guys) | 16:16 |
sean-k-mooney | inc0: i see ye are haveing some fun with kolla-k8s | 16:16 |
v1k0d3n | :) | 16:16 |
inc0 | sean-k-mooney, it's our local hype in Kolla;) | 16:17 |
v1k0d3n | seriously though, just trying to determine parity between what is needed for each. | 16:17 |
inc0 | and hype is important...also it is damn cool;) | 16:17 |
inc0 | v1k0d3n, deploy and upgrade | 16:17 |
v1k0d3n | sean-k-mooney: lol yes...fun. | 16:17 |
inc0 | that's it | 16:17 |
v1k0d3n | inc0: agreed | 16:17 |
v1k0d3n | well...downgrade in cases where downgrade could be possible. | 16:18 |
v1k0d3n | just throwing that out there. | 16:18 |
*** berendt has joined #openstack-kolla | 16:18 | |
inc0 | downgrade is not supported by openstack | 16:18 |
sean-k-mooney | i have an open question for ye guys. do we have a solution today for runing different contienres in the same group? | 16:18 |
v1k0d3n | not possible with ansible today, but definitely possible with k9s | 16:18 |
v1k0d3n | or k8s minus the special letter. | 16:18 |
sean-k-mooney | or rather running the porcess in in two different contiers with the same group. | 16:18 |
*** eaguilar has quit IRC | 16:19 | |
v1k0d3n | inc0: to that point (downgrade) even more reason to use an operator, honestly. :) | 16:19 |
inc0 | v1k0d3n, operator or not, won't happen nicely | 16:20 |
inc0 | there is no downgrade of database, period | 16:20 |
v1k0d3n | operator won't allow upgrade if there is a failure << what i'm saying. | 16:20 |
inc0 | this is only real issue we have, but hard enough to block whole thing | 16:20 |
v1k0d3n | there is a logic diagram in that repo i sent earlier. | 16:20 |
inc0 | sure, but still, no downgrade of db | 16:20 |
v1k0d3n | does db need downgraded? | 16:21 |
inc0 | as in - no reverse migration | 16:21 |
inc0 | ofc | 16:21 |
v1k0d3n | for other services? | 16:21 |
inc0 | if you do dbsync | 16:21 |
inc0 | you have new schema | 16:21 |
inc0 | if you save sth to new schema | 16:21 |
v1k0d3n | TPC entrypoint is a dbsync | 16:21 |
v1k0d3n | which is interesting approach. | 16:21 |
inc0 | you will either lose data or do manual reverse migration | 16:21 |
v1k0d3n | not saying i like or dislike. just noting it. | 16:21 |
inc0 | what I'm saying is downgrade won't happen not because of orchiestration mechanism | 16:22 |
inc0 | it would be super easy for us to downgrade even with ansible | 16:22 |
v1k0d3n | ok | 16:22 |
inc0 | reverse db migration -> that's blocker | 16:22 |
v1k0d3n | i'm just going to concede. your missing my point. | 16:22 |
v1k0d3n | kubernetes allows downgrade. openstack does not. openstack will need to change at some point. | 16:23 |
v1k0d3n | ^ point | 16:23 |
v1k0d3n | not a kolla issue to solve. | 16:23 |
inc0 | yeah, and if that ever happends we'll be ready | 16:23 |
sean-k-mooney | v1k0d3n: that will be a long way off | 16:23 |
inc0 | imho won't ever happen | 16:23 |
v1k0d3n | back to agreement on all points again. | 16:23 |
v1k0d3n | not sure about that. | 16:23 |
sean-k-mooney | v1k0d3n: i have more faith in the nova scheduler refactor and a swap to pure python3 before db downgrades are added | 16:23 |
inc0 | as they had reverse migrations and they are removing them | 16:24 |
v1k0d3n | if we said that applications would get to this point 3 years ago we said never going to happen :) | 16:24 |
v1k0d3n | sean-k-mooney: hahaha! fair enough | 16:24 |
inc0 | which means, OpenStack will actively block downgrade | 16:24 |
dmsimard | inc0: fyi I've revived an internal thread about licensing stuff (re: ansible vs apache v2), I'll let you know if anything comes out of it | 16:24 |
inc0 | I don't like it, bu thats what it is | 16:24 |
inc0 | dmsimard, I would appreciate it! | 16:24 |
*** n3v3rm0r3r has quit IRC | 16:25 | |
inc0 | I would be willing to code in really cool feature to ansible itself if I could | 16:25 |
*** sdake has joined #openstack-kolla | 16:25 | |
inc0 | would incubate in Kolla and then we could move it to ansible core | 16:25 |
inc0 | but GPL really hurt our effort there | 16:25 |
v1k0d3n | btw, did you guys hear about the ceph GPL stuff? | 16:25 |
sean-k-mooney | inc0: you could add some nice features as plugins | 16:25 |
sean-k-mooney | but ya i avoid the gpl | 16:26 |
dmsimard | inc0: FWIW Zuul is (still) apache v2 with a bunch of ansible stuff in there https://github.com/openstack-infra/zuul | 16:26 |
inc0 | sean-k-mooney, one particular plugin I am thinking about adding will be GPL-tainted | 16:26 |
v1k0d3n | also kind of wondering what happens to GPL in container world, since it's contained. | 16:26 |
inc0 | dmsimard, there are things that allows certain things to be coded to ansible with apache 2 | 16:26 |
inc0 | like our kolla_docker module | 16:26 |
rhallisey | v1k0d3n, I think the ceph issue got resolved | 16:26 |
rhallisey | license is chaning | 16:27 |
inc0 | but strategy plugin will be tainted. | 16:27 |
rhallisey | I don't have a link though | 16:27 |
dmsimard | rhallisey: license for zuul ? | 16:27 |
v1k0d3n | rhallisey: i saw that ceph docker was resolved. ceph itself is still GPL though. | 16:27 |
v1k0d3n | does that matter? | 16:27 |
dmsimard | rhallisey: oh nm | 16:27 |
rhallisey | dmsimard, no for cpeh | 16:27 |
rhallisey | ceph | 16:27 |
inc0 | I haven't seen ceph gpl story...link anyone? | 16:27 |
v1k0d3n | https://github.com/ceph/ceph-docker/blob/master/LICENSE | 16:27 |
kfox1111 | morning. | 16:27 |
sean-k-mooney | inc0: i was think that it would be cool to add a stratagey plugin that understood groups actully | 16:27 |
srwilkers | kfox1111: morning | 16:27 |
sdake | yo cats | 16:27 |
inc0 | sean-k-mooney, strategy plugin I was thking about would effectively add mechanism similar to --template-override but to ansible | 16:28 |
v1k0d3n | https://github.com/ceph/ceph/blob/0a5d8946004060ca8d848d5c9107e1596ebc7f33/README.md | 16:28 |
sdake | kfox1111 how do you feel about making a "operator" for kolla-kubernetes | 16:28 |
inc0 | things like dpdk could be just plugin to kolla instead of full patchset to core | 16:28 |
sean-k-mooney | inc0: i was think you could do somthing similar with the kolla-ansible tool for that issue. | 16:28 |
kfox1111 | sdake: I don't follow. | 16:28 |
sean-k-mooney | thought not generically i guess | 16:28 |
v1k0d3n | LGPL version 2.1 | 16:28 |
sdake | kfox1111 your probably missing ontext then - here is some docs: | 16:29 |
kfox1111 | yeah. been buys readying a system, so had to keep my head down. | 16:29 |
sdake | https://coreos.com/blog/introducing-operators.html | 16:29 |
inc0 | sean-k-mooney, so strategically my plan for this would be to have something like contrib dir for both dockerfiles and ansible | 16:29 |
sean-k-mooney | inc0: i would love to be able to do that or to have a mechanium similar to devstacks plugins | 16:30 |
*** rhallisey has quit IRC | 16:30 | |
inc0 | something like that | 16:30 |
sdake | kfox i havve to hit the airport in 5 mins, so make a decision quckly please :) | 16:30 |
v1k0d3n | no pressure, right? lol | 16:30 |
kfox1111 | sdake: looking... | 16:30 |
*** rhallisey has joined #openstack-kolla | 16:30 | |
inc0 | like place for code where you can put your own stuff that will be optional during deployment | 16:30 |
sdake | if kfox and sbezverk arn't in, we might as well shelve that idea | 16:30 |
inc0 | easier to manage and will allow us to have code with lower quality requirements in there | 16:31 |
sean-k-mooney | inc0: well my taught were i would lovs to add a mechanium where i could add a kolla sub directory to an arbty git directory and load it as a plugin with docker file and ansible playbooks | 16:31 |
v1k0d3n | inc0: helm has that btw. | 16:31 |
v1k0d3n | deis steward too | 16:31 |
inc0 | sean-k-mooney, yup, that's the plan | 16:31 |
inc0 | and I think I can do this with ansible native way | 16:31 |
v1k0d3n | oh, just talking about ansible. | 16:31 |
inc0 | with proper strategy plugin | 16:31 |
v1k0d3n | not for the kubernetes part. | 16:31 |
v1k0d3n | right? | 16:31 |
inc0 | ...if not for GPL | 16:32 |
inc0 | yeah v1k0d3n | 16:32 |
sean-k-mooney | inc0: that is definetly something i would love to test out. | 16:32 |
inc0 | this is not a feature I care about in k8s initial release;) | 16:32 |
v1k0d3n | ok. just checking. man, gets SO confusing having kolla blob like this. | 16:32 |
zhubingbing | good bye guys | 16:32 |
inc0 | trust me, I know | 16:32 |
zhubingbing | -) | 16:32 |
srwilkers | bye zhubingbing | 16:32 |
inc0 | have a good night zhubingbing | 16:32 |
zhubingbing | thanks | 16:32 |
v1k0d3n | later zhubingbing | 16:32 |
zhubingbing | bye | 16:32 |
sdake | v1k0d3n reading scrollback now | 16:32 |
sean-k-mooney | inc0: speaking of dpdk https://review.openstack.org/#/c/342354/ i have a lot of work still to do but it is at least fucntional now | 16:33 |
sdake | v1k0d3n noticed you mentioned what might make sense in ansible might not in kubernetes | 16:33 |
sdake | i totallly agree | 16:33 |
sdake | that is why we make our containers non-specific to any deployment engine | 16:33 |
kfox1111 | sdake: half way through reading... so far seems like a good idea. | 16:33 |
v1k0d3n | sdake: inc0 get's credit...i made it a daily quote. | 16:34 |
kfox1111 | though I'm undersleeped at the moment too... | 16:34 |
rhallisey | sdake, +1 :) | 16:34 |
sdake | v1k0d3n this is another reason why i am not keen on having containers magically orchestrate themselves | 16:34 |
inc0 | lol | 16:34 |
sdake | v1k0d3n because it makes the containers non-generic and specific to a platform | 16:35 |
sdake | v1k0d3n i think your spot in with your operator approach | 16:35 |
sdake | lets figure out how to make that a reality | 16:35 |
sdake | assumign kfox1111 is in ;) | 16:35 |
kfox1111 | so are operators implemnted now? | 16:35 |
v1k0d3n | lol | 16:35 |
v1k0d3n | this discussion is hilarious | 16:35 |
sdake | kfox1111 you implement one on your own | 16:35 |
rhallisey | kfox1111, we're containerizing you | 16:35 |
v1k0d3n | "very who's on first" | 16:35 |
*** yingjun has quit IRC | 16:35 | |
sdake | lets push off the helm discussion to later, i think theere isn't any major pushback on that idea | 16:36 |
*** yingjun has joined #openstack-kolla | 16:36 | |
kfox1111 | hehe. | 16:36 |
inc0 | we kinda did already | 16:36 |
inc0 | as fencing pod | 16:36 |
v1k0d3n | kfox1111: talking about operator concept for kolla/openstack | 16:36 |
sdake | but its hard to focus on more then one thing at a time | 16:36 |
inc0 | ugly-as-fuck but well | 16:36 |
kfox1111 | I've asked for clone(kfox1111) for years... that would be cool. :) | 16:36 |
kfox1111 | are operator support in k8s implmented today? | 16:36 |
rhallisey | docker run kfox1111 | 16:36 |
kfox1111 | I don't remember seeing it in the specs currently. | 16:36 |
sdake | kfox1111 no, k8s nor helm implement operators | 16:37 |
v1k0d3n | ok, so helm (for the most part) we agree is good, am i getting this correctly? | 16:37 |
sdake | there isn't a standardized operator implementation that works across service | 16:37 |
v1k0d3n | just because i'm slow, and need clarity.... | 16:37 |
rhallisey | kfox1111, no spec yet | 16:37 |
sdake | each operator is customized to the service | 16:37 |
kfox1111 | sdake: k. | 16:37 |
v1k0d3n | ^^ yes | 16:37 |
sdake | v1k0d3n right i thiink so | 16:37 |
kfox1111 | at the faw now... | 16:37 |
inc0 | uaf in short - uniform alpha foxtrot | 16:37 |
v1k0d3n | sdake: totally agree. | 16:37 |
v1k0d3n | ok...thanks sdake clears up helm for me. | 16:37 |
kfox1111 | faq | 16:38 |
v1k0d3n | kfox1111: you've been up on the operator things coming from coreos? | 16:38 |
sdake | v1k0d3n where the disagreement around helm is at present is "when" - and I think that is second (operator first) | 16:38 |
kfox1111 | v1k0d3n: reading now. | 16:38 |
v1k0d3n | for prometheus and etcd. ok | 16:38 |
sdake | kfox1111 going to miss my fliight | 16:38 |
sdake | pull the trigger on a decision bro ;) | 16:38 |
v1k0d3n | oh ok sdake...i see now! | 16:38 |
v1k0d3n | yes, ok...totally back that. | 16:38 |
kfox1111 | sdake: I'm for it, so long as its a good implementation. | 16:38 |
rhallisey | sdake, ml vote | 16:38 |
rhallisey | wel | 16:38 |
kfox1111 | have a good flight. | 16:39 |
rhallisey | spec :) | 16:39 |
v1k0d3n | later sdake | 16:39 |
sdake | roger spec | 16:39 |
sdake | ml vote = overkill imo | 16:39 |
rhallisey | we'll comment on a spec | 16:39 |
sdake | v1k0d3n one lsat thing we will need to discuss is implementation language | 16:39 |
inc0 | python | 16:39 |
v1k0d3n | group has python | 16:39 |
sdake | v1k0d3n i know nobody is going to really like it but we may hae to go with python in the short term | 16:39 |
kfox1111 | I think if we adopt helm, it will be important soon. | 16:39 |
v1k0d3n | so python | 16:39 |
sdake | and port to go later | 16:39 |
v1k0d3n | agreed kfox1111 100% | 16:39 |
kfox1111 | as it will allow us to do multiple mariadb/rabbit more easily. | 16:39 |
inc0 | python is good endgoal too | 16:39 |
sdake | inc0 agree | 16:40 |
inc0 | really, golang is great language to certain stuff | 16:40 |
kfox1111 | which we need for medium/large clouds. | 16:40 |
v1k0d3n | yes, but there's also the whole build process around helm that is EXTREMELY powerful. | 16:40 |
sdake | so can I have a summary from v1k0d3n of what we have all agreed to before I write a spec ;) | 16:40 |
v1k0d3n | operators in python. first to go would be rabbitmq, mariadb, and memcached. | 16:40 |
*** yingjun has quit IRC | 16:40 | |
kfox1111 | sdake: we can work out the details after spec's started. | 16:40 |
inc0 | v1k0d3n, how soon can we get PoC of this up? | 16:40 |
sdake | kfox1111 going to cloudnativecon | 16:40 |
inc0 | can we please start with PoC and decide afterwards? | 16:41 |
sdake | people will want to know our plans | 16:41 |
kfox1111 | sdake: ah. sweet. :) | 16:41 |
sdake | i dont want to shoow up and say "taada - we got nothign": ) | 16:41 |
*** berendt has quit IRC | 16:41 | |
v1k0d3n | inc0: is this a cross community PoC? | 16:41 |
sdake | i'm an openstack ambassador in this case | 16:41 |
rhallisey | inc0, I think poc and spec will be in parallel | 16:41 |
v1k0d3n | because we have hardware man. | 16:41 |
kfox1111 | sdake: +1 to operators and helm currently. devils in the details though. | 16:41 |
inc0 | I've seen too many things break after long and detailed discussion and first hit on writing the stuff | 16:41 |
kfox1111 | if we find showstoppers we need to either fix or abandon. | 16:41 |
sean-k-mooney | inc0: isnt the main politcal issue with golang that its not approved byoned pocs for any openstack project to use | 16:41 |
sdake | inc0 right | 16:41 |
v1k0d3n | inc0: the spec is the plan ;) | 16:41 |
v1k0d3n | am i right? | 16:41 |
sdake | sean-k-mooney go is possible with work from us | 16:41 |
inc0 | v1k0d3n, I'll go with poc in pararell | 16:42 |
v1k0d3n | PoC works against the spec | 16:42 |
*** n3v3rm0r3r has joined #openstack-kolla | 16:42 | |
v1k0d3n | yeah, i dont mind that. but would like to work all together on this. | 16:42 |
sdake | so i've got3% battery power | 16:42 |
sdake | here is my understanding | 16:42 |
inc0 | one poc speaks more than 1000 specs, or words in specs...you know | 16:42 |
sdake | 1. operator first ASAP in python | 16:42 |
kfox1111 | +1 to working together. | 16:42 |
*** jrich523_ is now known as jrich523 | 16:42 | |
v1k0d3n | ^^ oia | 16:42 |
sdake | 2. helm second | 16:42 |
inc0 | helm poc is already happening | 16:42 |
kfox1111 | poc's are ok, so long as we're all willing to throw them out. | 16:42 |
sdake | 3. possibly port operator to python | 16:42 |
kfox1111 | sdake: operators aren't implmented. | 16:42 |
kfox1111 | helm is. | 16:42 |
v1k0d3n | yes. sdake srwilkers already put out a helm blueprint i think, but spec is better | 16:42 |
kfox1111 | so implementing 2 first would be quicker. | 16:43 |
kfox1111 | or at least poc'ing. | 16:43 |
rhallisey | kfox1111, we would write the operator | 16:43 |
inc0 | qwang, jascott1 up already? | 16:43 |
v1k0d3n | kfox1111: i think we can PoC both at the same time. | 16:43 |
v1k0d3n | it's all good. | 16:43 |
v1k0d3n | the helm stuff is already started... | 16:43 |
sdake | kfox1111 how about this | 16:43 |
kfox1111 | rhallisey: but you need operator support in k8s first. | 16:43 |
v1k0d3n | very close | 16:43 |
sdake | 1. operator and helm in parallel but not integrated | 16:43 |
kfox1111 | sdake: +1. | 16:43 |
rhallisey | kfox1111, I think we can use kubectl | 16:43 |
v1k0d3n | kfox1111: https://github.com/sapcc/openstack-helm/commits/master | 16:43 |
sean-k-mooney | kfox1111: i think the dsl and jinja templates pocs show that we can trow them away | 16:43 |
sdake | 2. integrate em | 16:43 |
kfox1111 | sdake: sounds godo. | 16:44 |
kfox1111 | godo | 16:44 |
sdake | sean-k-mooney right - we can move past ideas to implementation | 16:44 |
kfox1111 | good | 16:44 |
kfox1111 | stupid daylight savings time. :/ | 16:44 |
*** berendt has joined #openstack-kolla | 16:44 | |
sdake | v1k0d3n in agreement then? | 16:44 |
v1k0d3n | ok...so here guys...we will start prepping lab for this PoC work. want everyone to help if possible. | 16:44 |
v1k0d3n | cores have access, and code in clear. | 16:44 |
v1k0d3n | sdake: +1 | 16:45 |
v1k0d3n | i agree | 16:45 |
v1k0d3n | in fact, +2 if i could | 16:46 |
v1k0d3n | :) | 16:46 |
rhallisey | :) | 16:46 |
v1k0d3n | rhallisey: you in? | 16:46 |
v1k0d3n | kfox1111: ? | 16:46 |
rhallisey | yes of course | 16:46 |
sdake | pbourke is bringing people to the party too | 16:46 |
v1k0d3n | inc0: ? | 16:46 |
kfox1111 | I'm in. | 16:46 |
inc0 | v1k0d3n, ofc | 16:46 |
sdake | about 4 or 5 engineers from my understanding | 16:46 |
v1k0d3n | pbourke: yeeaah! | 16:46 |
sdake | we need to catch them up | 16:46 |
inc0 | make it open in gerrit | 16:46 |
inc0 | that's it | 16:46 |
sdake | via the spec process unfortunately | 16:46 |
inc0 | whoever wants to join in, will join in | 16:46 |
*** zhubingbing has quit IRC | 16:46 | |
sdake | I'd ask that we not have a rollcall ote on teh spec tho | 16:46 |
sdake | because thats a apin in the ass and this isn't contentious | 16:46 |
srwilkers | agreed | 16:47 |
inc0 | so ok, guys, can we please not add more beaurocracy than we need to? | 16:47 |
inc0 | k8s needs to move fast | 16:47 |
sdake | inc0 precisely | 16:47 |
v1k0d3n | inc0: omg YES! | 16:47 |
inc0 | let's not overengineer process | 16:47 |
sdake | inc0 i dont even want ot right a spec | 16:47 |
v1k0d3n | second quote of the day. | 16:47 |
sdake | but you demandedo ne ;) | 16:47 |
v1k0d3n | now you've finally had your freaking tea! | 16:47 |
kfox1111 | sdake: dong forget to get on the plane. :) | 16:47 |
kfox1111 | dont | 16:47 |
sdake | right- gott ajet | 16:47 |
inc0 | let's just do this poc and go from there | 16:48 |
v1k0d3n | alright...this is all great guys. see you at kubecon sdake | 16:48 |
kfox1111 | I think we're all on the same page. | 16:48 |
inc0 | that will give us idea if it makes sense or not | 16:48 |
kfox1111 | just gota work out the details. | 16:48 |
rhallisey | :) | 16:48 |
sean-k-mooney | sdake: such a bad pun i hope that was intentional :) | 16:48 |
rhallisey | lul | 16:48 |
v1k0d3n | yes. lock it in. and when you start to doubt yourself...remember to read back through this monster thread. | 16:48 |
v1k0d3n | :D | 16:48 |
rhallisey | sean-k-mooney, sdake isn't punny | 16:48 |
v1k0d3n | not punny at all | 16:49 |
sean-k-mooney | oh what have i done | 16:49 |
*** egonzalez has quit IRC | 16:49 | |
sean-k-mooney | :0 | 16:49 |
*** egonzalez90 has joined #openstack-kolla | 16:49 | |
inc0 | wirehead_, is our local pun source | 16:49 |
inc0 | you might say that he can pack a PUNch | 16:49 |
sean-k-mooney | ok i need to grab more caffine be right back | 16:50 |
v1k0d3n | inc0: perhaps too much tea | 16:50 |
inc0 | :D | 16:50 |
*** csj_ has quit IRC | 16:50 | |
inc0 | I'm about half way through my mountain dew | 16:51 |
inc0 | wait till I finish it.. | 16:51 |
v1k0d3n | so glad we worked this out before my flight. | 16:52 |
v1k0d3n | and before kubecon. | 16:52 |
inc0 | well we just agreed to make a poc, which in my book shouldn't even be matter of discussion;) | 16:52 |
inc0 | if somebody feels it's worth doing, then it's worth making a poc | 16:52 |
sdake | inc0 we agreed to a course of action as well | 16:52 |
inc0 | and poc helps with discussion temendously | 16:52 |
v1k0d3n | yeah, but hopefully we all agreed to work on the PoC :) | 16:52 |
v1k0d3n | SAP has been doing the helm PoC for a little while now. | 16:53 |
inc0 | yeah helm seems to be almost decided | 16:53 |
v1k0d3n | sdake: yes...biggest part, course of action. | 16:53 |
inc0 | how we deal with helm shortcommings, we'll see | 16:53 |
v1k0d3n | that's why i feel good now. | 16:53 |
inc0 | I have few ideas on my own:) | 16:53 |
v1k0d3n | they are working through these too. made some big improvements today actually. | 16:53 |
inc0 | yeah it's on my todo list - fully understand their code | 16:54 |
sdake | helm has no shortcoming for us in this model | 16:54 |
inc0 | oh it does | 16:54 |
sdake | I think its not endgame (if I was maintaining helm) | 16:54 |
sdake | but SEP :) | 16:54 |
v1k0d3n | lets just leave on a good note today? | 16:54 |
v1k0d3n | lol | 16:54 |
inc0 | Helm doesn't help us with certain stuff | 16:55 |
v1k0d3n | no, but operator makes up for that actually. | 16:55 |
inc0 | we need to work through these before we jump head first | 16:55 |
inc0 | yeah, it might | 16:55 |
*** v1k0d3n has quit IRC | 16:55 | |
sdake | ok gotta jet | 16:55 |
sdake | later - will send trip report to the cores when I return | 16:55 |
inc0 | have fun | 16:56 |
*** v1k0d3n has joined #openstack-kolla | 16:56 | |
sdake | fun = work? | 16:56 |
sdake | are you kidding :) | 16:56 |
*** sdake has quit IRC | 16:56 | |
inc0 | I kinda feel bad about kubecon | 16:56 |
inc0 | I was in Seattle yesterday.. | 16:56 |
v1k0d3n | seriously? you shouldve stayed man. | 16:56 |
kfox1111 | operators and helm are complimentery. | 16:56 |
v1k0d3n | would really like to have you there.... | 16:56 |
v1k0d3n | kfox1111: yes | 16:56 |
v1k0d3n | work hand in hand. | 16:57 |
kfox1111 | operators look to be implemented in containers. helm is packages for container sets. | 16:57 |
inc0 | v1k0d3n, tickets were already gone | 16:57 |
kfox1111 | a good match. :) | 16:57 |
inc0 | next time | 16:57 |
*** berendt has quit IRC | 16:57 | |
*** bmace has joined #openstack-kolla | 16:57 | |
inc0 | also I have work to do here:P can't afford to run through all the cool conferences | 16:57 |
sean-k-mooney | can someone send a link to what operators are in a k8s context | 16:57 |
sean-k-mooney | is it related to this https://coreos.com/blog/introducing-operators.html | 16:58 |
inc0 | sean-k-mooney, preciselly | 16:58 |
v1k0d3n | inc0: next time reach out :) | 16:58 |
sean-k-mooney | a ok ill add it to my reading list | 16:59 |
*** fragatina has joined #openstack-kolla | 16:59 | |
v1k0d3n | may be able to dig one up as a member. you guys too actually. | 16:59 |
v1k0d3n | but you're right...this late tough. | 16:59 |
inc0 | don't worry, I'm sure you guys can handle it | 16:59 |
*** harlowja_at_home has quit IRC | 17:00 | |
qwang | morning | 17:00 |
v1k0d3n | sean-k-mooney: you may find these useful too | 17:00 |
v1k0d3n | https://github.com/coreos/etcd-operator/tree/master/doc/design | 17:00 |
*** mgoddard_ has quit IRC | 17:00 | |
*** mgoddard has joined #openstack-kolla | 17:00 | |
v1k0d3n | and even this as prometheus https://github.com/coreos/prometheus-operator/tree/master/Documentation | 17:01 |
*** fragatina has quit IRC | 17:01 | |
v1k0d3n | like i said before, there may be opportunity for kolla to use kubernetes federated api as well at some point. | 17:01 |
*** fragatina has joined #openstack-kolla | 17:01 | |
v1k0d3n | deis steward builds upon this concept i believe | 17:02 |
*** diogogmt has joined #openstack-kolla | 17:02 | |
kfox1111 | I'm talking with the trove ptl too. | 17:02 |
kfox1111 | it makes a lot of sence for trove to provide an operator for all their db's. | 17:02 |
kfox1111 | we could even reuse it if it existed then. | 17:02 |
*** strigazi is now known as strigazi_AFK | 17:02 | |
sean-k-mooney | v1k0d3n: thanks ill add that too | 17:03 |
v1k0d3n | yes, kfox1111 i like that | 17:03 |
kfox1111 | no reason to implement everything ourselves. :) | 17:03 |
*** radeksmg has quit IRC | 17:04 | |
*** matrohon has quit IRC | 17:09 | |
v1k0d3n | ^^ totally agree wtih that too | 17:11 |
v1k0d3n | i would actually LIKE to see the projects themselves build the operator logic. | 17:11 |
*** egonzalez90 has quit IRC | 17:11 | |
v1k0d3n | ok guys...im off in the air. later! | 17:14 |
kfox1111 | I think a operator based trove would solve a lot of other problem for trove too. :) | 17:14 |
kfox1111 | l8r | 17:14 |
kfox1111 | I gota head out for a bit too. | 17:14 |
*** v1k0d3n has quit IRC | 17:15 | |
*** duonghq has quit IRC | 17:15 | |
*** v1k0d3n has joined #openstack-kolla | 17:15 | |
*** v1k0d3n has quit IRC | 17:20 | |
*** berendt has joined #openstack-kolla | 17:22 | |
*** unicell1 has quit IRC | 17:23 | |
*** eaguilar has joined #openstack-kolla | 17:24 | |
*** fragatina has quit IRC | 17:28 | |
*** Serlex has quit IRC | 17:31 | |
*** diogogmt has quit IRC | 17:33 | |
*** lamt has quit IRC | 17:38 | |
*** lamt has joined #openstack-kolla | 17:38 | |
*** diogogmt has joined #openstack-kolla | 17:43 | |
*** lamt has quit IRC | 17:43 | |
*** neilus has joined #openstack-kolla | 17:43 | |
*** athomas has quit IRC | 17:47 | |
*** neilus has quit IRC | 17:48 | |
openstackgerrit | Andrew Widdersheim proposed openstack/kolla: Allow cinder-volume to be configured to use NFS https://review.openstack.org/279913 | 17:58 |
*** Kimmo__ has joined #openstack-kolla | 17:58 | |
*** neilus has joined #openstack-kolla | 18:00 | |
*** neilus has quit IRC | 18:03 | |
*** neilus has joined #openstack-kolla | 18:03 | |
*** fragatina has joined #openstack-kolla | 18:04 | |
*** unicell has joined #openstack-kolla | 18:05 | |
*** dave-mccowan has quit IRC | 18:06 | |
*** shardy has quit IRC | 18:07 | |
*** rhallisey has quit IRC | 18:09 | |
*** rhallisey has joined #openstack-kolla | 18:10 | |
*** dave-mccowan has joined #openstack-kolla | 18:11 | |
*** neilus has quit IRC | 18:11 | |
inc0 | srwilkers, lrensing up? | 18:13 |
srwilkers | zzzzz | 18:13 |
srwilkers | jk | 18:13 |
srwilkers | im here | 18:13 |
inc0 | qwang, Qin, meet Steve | 18:14 |
inc0 | Steve, meet Qin:) | 18:14 |
inc0 | let's work together on helm | 18:14 |
lrensing | hello | 18:14 |
srwilkers | hello:) | 18:14 |
jascott1 | did someone say helm? | 18:14 |
srwilkers | \o/ | 18:14 |
inc0 | oh, and there is Justin | 18:15 |
*** rhallisey has quit IRC | 18:15 | |
*** rhallisey has joined #openstack-kolla | 18:15 | |
inc0 | I see Brandon already left, but I met with Justin and Qin last week | 18:15 |
inc0 | and our quick plan for poc will be - define deps in helm | 18:16 |
inc0 | spawn a container from helm with these deps passed with ENV variable | 18:16 |
inc0 | and entrypoint inside container will query kubectl to spawn after all the deps are met | 18:16 |
inc0 | so this will be kinda similar to coreos operator, but not quite | 18:16 |
srwilkers | okay | 18:17 |
qwang | Hi Steve | 18:17 |
inc0 | srwilkers, let's make a race, which 2 person team will make patchet with helm first;) | 18:18 |
inc0 | team Intel or team ATT;) | 18:18 |
srwilkers | well, if i was a gambling man | 18:18 |
inc0 | it's ok to have 2 for now, we'll look at them and go from there | 18:18 |
inc0 | mariadb + keystone is the goal | 18:18 |
inc0 | for poc:) | 18:18 |
srwilkers | okay, sounds good | 18:19 |
*** eaguilar has quit IRC | 18:19 | |
lrensing | ha alright | 18:19 |
inc0 | adding a bit of unhealthy competition to community is good for popcorn consumption | 18:19 |
inc0 | ;) | 18:19 |
srwilkers | back to the amphetamines in the water discussion | 18:20 |
lrensing | i was getting ready to say…. circling back to that | 18:20 |
inc0 | but really, I think it'll be cool to get 2 approaches independently if we can do it within few days | 18:20 |
inc0 | after that we start cooperating fully, but let's get as much ideas on table before we commit | 18:21 |
inc0 | as possible* | 18:21 |
srwilkers | i think thats a good idea | 18:21 |
qwang | inc0: what do you mean by "2 approaches"? what is the other approach? | 18:23 |
inc0 | qwang, one is what we thought of in Friday | 18:25 |
inc0 | I'm pretty sure srwilkers and lrensing already paid some thought themselves on the matter | 18:25 |
inc0 | let make quick and dirty poc of both | 18:25 |
inc0 | and see which one sticks | 18:25 |
*** tyrola has joined #openstack-kolla | 18:26 | |
*** eaguilar has joined #openstack-kolla | 18:27 | |
openstackgerrit | Larry Rensing proposed openstack/kolla-kubernetes: Adding prechecks script https://review.openstack.org/394569 | 18:28 |
*** v1k0d3n has joined #openstack-kolla | 18:28 | |
v1k0d3n | ...and now...the room is silent. | 18:29 |
v1k0d3n | :) | 18:29 |
lrensing | v1k0d3n: we were just given a challenge | 18:30 |
v1k0d3n | hey, kfox1111...question for lrensing actually...how do you think prechecks should be done for kolla-k8s? | 18:30 |
*** fragatina has quit IRC | 18:30 | |
v1k0d3n | oh? what now lrensing? | 18:30 |
lrensing | intel vs at&t helm POC shootout | 18:30 |
*** fragatina has joined #openstack-kolla | 18:31 | |
inc0 | v1k0d3n, read the log from last 15min | 18:31 |
v1k0d3n | i need to find a link. | 18:32 |
rhallisey | dims, ping | 18:32 |
v1k0d3n | i see some of it.though | 18:32 |
v1k0d3n | srwilkers: sent to me. | 18:32 |
dims | rhallisey : pong | 18:32 |
*** mkoderer has quit IRC | 18:32 | |
v1k0d3n | so, inc0 basically both working on PoC, right? | 18:33 |
inc0 | http://eavesdrop.openstack.org/irclogs/%23openstack-kolla/%23openstack-kolla.2016-11-07.log.html | 18:33 |
inc0 | yeah, for next few days | 18:33 |
v1k0d3n | so i want to include the SAP folks. they have most of what you're talking about with helm already done...in fact making big strides. | 18:33 |
*** neilus has joined #openstack-kolla | 18:33 | |
inc0 | get ideas up there without tainting each others brains | 18:33 |
v1k0d3n | thanks for link, very helpful! | 18:33 |
v1k0d3n | this is one of our larger chat logs in a while :) | 18:34 |
inc0 | v1k0d3n, sure, just they expressed that they don't really have time to work on that stuff with us, and thats fair | 18:34 |
v1k0d3n | lol | 18:34 |
v1k0d3n | and it's already only few hours into the day. | 18:34 |
v1k0d3n | lol | 18:34 |
inc0 | yeah, caffeine is kicking in | 18:34 |
v1k0d3n | inc0: well...they did...but they are also working on it still | 18:34 |
inc0 | yeah, I'd love to have them here:) | 18:34 |
v1k0d3n | the makefiles and prebuilds are gone today. that's a big step. | 18:35 |
v1k0d3n | we're working with them, and specifically including deis in the mix too...since they can help us the most. | 18:35 |
inc0 | totally for it | 18:35 |
v1k0d3n | in fact...deis expressed a lot of interest working on this, because then they understand what it takes and can help us solve. | 18:35 |
v1k0d3n | total win for everyone. | 18:36 |
inc0 | I agree, I just don't want to pressure anyone | 18:36 |
v1k0d3n | i will be talking with gabe and his team this week at kubecon. | 18:36 |
v1k0d3n | no pressure on anyone. | 18:36 |
v1k0d3n | well, except for us to get the lab going :) | 18:36 |
inc0 | yeah, maybe if they deliver what they have to deliver that will give them some breathing room | 18:36 |
v1k0d3n | so i have to work that out and get that going ASAP. we're almost there. hardware is actually ready now. | 18:36 |
v1k0d3n | let me try to sync up with them today. | 18:37 |
inc0 | cool, do that, let me know if you need any ad-hoc call or whatnot | 18:37 |
v1k0d3n | planned a meeting with them too to discuss a path forward and using lab. | 18:37 |
v1k0d3n | sounds like a plan | 18:37 |
inc0 | btw Deis is a company..? | 18:37 |
inc0 | or a name?;) | 18:38 |
v1k0d3n | yes, focused only on open source. | 18:38 |
inc0 | I'm not familiar with it | 18:38 |
inc0 | I see | 18:38 |
v1k0d3n | deis makes it's money by helping companies as hired guns. | 18:38 |
inc0 | cool, would love to talk to them | 18:38 |
v1k0d3n | in fact, i would really like to talk to my management about bringing them in to assist. i think that would help us all here honestly. | 18:39 |
v1k0d3n | they are very un-opinionated, since they aren't revenue-generated via a product. | 18:39 |
v1k0d3n | want me to set up a meeting for cores? | 18:39 |
v1k0d3n | we met with them last week. impressive stuff with a helm deployment/development strategy. | 18:39 |
inc0 | yeah, we can make hangout with them | 18:41 |
inc0 | maybe after kubecon? | 18:41 |
v1k0d3n | yeah, that would be perfect. | 18:41 |
inc0 | in the meantime, if they would like to join us here on irc | 18:41 |
v1k0d3n | i will suggest a larger discussion with kolla team when i talk with them | 18:41 |
inc0 | we can include them in our little group with strange sense of humor | 18:42 |
inc0 | they either blend in or run out scared | 18:42 |
v1k0d3n | most (as in probably 90%+) are going to be at kubecon | 18:42 |
v1k0d3n | lol | 18:42 |
inc0 | cool, extend our invitation then:) | 18:43 |
v1k0d3n | they also hang out in the kubernetes sleack group as well. | 18:43 |
v1k0d3n | that's their main hangout | 18:43 |
v1k0d3n | oh and deis sorry. | 18:43 |
v1k0d3n | they have their own | 18:43 |
v1k0d3n | so they are in sig apps all day, and deis slack | 18:43 |
v1k0d3n | will invite them to irc also | 18:43 |
inc0 | kk | 18:46 |
*** Marcellin__ has joined #openstack-kolla | 18:50 | |
*** lamt has joined #openstack-kolla | 18:53 | |
v1k0d3n | inc0 oh btw...prometheus went to an operator model. can we just deploy this as part of the kolla-k8s tasks? | 18:55 |
v1k0d3n | thoughts? | 18:55 |
inc0 | well I talked with snap guys about prometheus | 18:55 |
inc0 | apperantly prometheus has issues at scale | 18:55 |
inc0 | I'll pick their brains more about monitoring | 18:56 |
v1k0d3n | interesting. perhaps we can help address that on behalf of the k8s community then? | 18:57 |
v1k0d3n | since prometheus is part of cncf | 18:57 |
v1k0d3n | also, can we submit an issue if there are scale issues? | 18:57 |
sean-k-mooney | so just reading a little about operators, could they be used to manage mysql in a galarea cluster? that is a gap in kolla-k8s today is that correct? | 18:58 |
v1k0d3n | sean-k-mooney: that is the intent yes | 18:59 |
v1k0d3n | why i brought it up | 18:59 |
sean-k-mooney | ok cool. the demo ye did was nice at the summit but it would be even more awsome if there was no downtime because of active active deployments of our stateful services | 19:00 |
v1k0d3n | i agree definitely | 19:01 |
sean-k-mooney | even with operators would it still be better to keep a single rabbit mq or perhaps isolated clusters of rabbitmq instances in the k8s deployment | 19:02 |
v1k0d3n | for our purposes preferably isolated clusters of rabbitmq. we're huge. our current implementation needs to support more than just one rabbitmq. | 19:05 |
sean-k-mooney | inc0: do you know if there is any intergration between snap/prometheus with any openstack services. heat/gnocci/aodh/ceilomenter. | 19:05 |
sean-k-mooney | v1k0d3n: i really like the idea of haveing 1 rabbitmq per service by default and allow each to be clustered if needed for scaling. | 19:06 |
*** ayoung has joined #openstack-kolla | 19:06 | |
inc0 | clusters will hurt scalability rather than help | 19:07 |
sean-k-mooney | v1k0d3n: for example one rabbitmq insnace might be fine for glance but i might need 3 for nova or neutron so it would be nice to be able to scale out only the service that need it . | 19:07 |
sean-k-mooney | inc0: really that is interesting | 19:07 |
inc0 | yeah, local disk will always be faster than mirroring over network | 19:07 |
sean-k-mooney | yes i guess that is true | 19:08 |
inc0 | eventually consistent systems helps with scalling, but rabbitmq is not that | 19:08 |
sean-k-mooney | so rabbit only scales vertically well? | 19:08 |
inc0 | well one thing I need to check is what is impact of losing an rabbimq data really | 19:09 |
v1k0d3n | this will be an interesting area to explore later, but i'm good for keeping it simple for now. | 19:09 |
sean-k-mooney | e.g. if you need to scale rabbit but it on a faster cpu with faster disk. | 19:09 |
inc0 | because if impact is low, we could do sth like multiple single-node rabbitmqs | 19:09 |
v1k0d3n | another problem for another day as i see it. | 19:09 |
inc0 | and round robin through them | 19:09 |
inc0 | add a bit of heuristics for services/nodes to prefer clusters they're more likely to find messages they care about | 19:10 |
inc0 | well, interesting experiment...one day | 19:10 |
sean-k-mooney | speaking of another day its time grab dinner. talk to everyone tomorow | 19:10 |
inc0 | have a good night Sean | 19:11 |
*** radeksmg has joined #openstack-kolla | 19:12 | |
inc0 | brb | 19:12 |
*** inc0 has quit IRC | 19:12 | |
*** inc0 has joined #openstack-kolla | 19:12 | |
v1k0d3n | later sean-k-mooney | 19:13 |
*** v1k0d3n has quit IRC | 19:16 | |
*** v1k0d3n has joined #openstack-kolla | 19:17 | |
*** dwalsh_ has quit IRC | 19:21 | |
kfox1111 | v1k0d3n: maybe a kollakube subcommand? | 19:24 |
*** sdake has joined #openstack-kolla | 19:24 | |
*** sdake has quit IRC | 19:25 | |
*** sdake has joined #openstack-kolla | 19:25 | |
*** sdake has quit IRC | 19:26 | |
*** sdake has joined #openstack-kolla | 19:27 | |
sdake | sup peeps | 19:28 |
kfox1111 | wb | 19:28 |
sdake | ya snakes on a palne | 19:28 |
kfox1111 | :) | 19:29 |
inc0 | sdakes on the plane | 19:29 |
sdake | whats cool bout this plane is | 19:29 |
sdake | a) in an exit row | 19:29 |
v1k0d3n | v1k0d3n on the plane too | 19:29 |
sdake | b) have power avilable for my laptop | 19:29 |
sdake | sup v1k0d3n | 19:29 |
inc0 | which airlines? | 19:29 |
v1k0d3n | ha! | 19:29 |
v1k0d3n | exit row here too | 19:29 |
v1k0d3n | no power though :/ | 19:29 |
v1k0d3n | i wish | 19:30 |
v1k0d3n | just really conserving ... so no docker or other services running :) | 19:30 |
kfox1111 | anyone seen sbezverk? | 19:30 |
v1k0d3n | kfox1111 good with a kolla kube command, but something smaller maybe? | 19:31 |
v1k0d3n | this is picky...i know | 19:31 |
rhallisey | kfox1111, he's on vacation | 19:31 |
kfox1111 | v1k0d3n: the commands just a plugable set of subcommands. so pluging it in should be pretty light weight. | 19:31 |
v1k0d3n | wish it kollactl ? | 19:31 |
kfox1111 | rhallisey: ah. ok. thanks. :) | 19:31 |
v1k0d3n | kollaadm ? | 19:31 |
v1k0d3n | :) | 19:31 |
kfox1111 | rhallisey: can you workflow the gate jobs then? | 19:32 |
rhallisey | yar | 19:32 |
v1k0d3n | makes sense | 19:32 |
kfox1111 | rhallisey: https://review.openstack.org/#/c/386966/ and https://review.openstack.org/#/c/389299/ please | 19:32 |
kfox1111 | v1k0d3n: yeah, that would be nice. but kolla itself probably should be kollactl. but then we'd be kollakubectl, which is too long. :) | 19:33 |
kfox1111 | v1k0d3n: I also kind of think we'll be providing kollakube ultipmately as a container. so deps and such isn't a big issue for the cli. | 19:34 |
v1k0d3n | this is the problem with having three projects in 1 | 19:34 |
kfox1111 | I should stop trying to talk today... I just can't type it seems. | 19:34 |
kfox1111 | yeah. | 19:34 |
v1k0d3n | well, i think that container is good, but compiled tools/scripts can be copied to the main control host or group of hosts. | 19:35 |
v1k0d3n | if the container gets messed up, i guess you can always rollback, and i like things in containers, but if the container gets messed up, user has no way to control anymore. | 19:36 |
v1k0d3n | just thinking through | 19:36 |
kfox1111 | its python. I don't think we will compile them ever. | 19:36 |
v1k0d3n | maybe it's not even an issue. | 19:36 |
*** neilus has quit IRC | 19:36 | |
v1k0d3n | well, i know...sorry...bad example. | 19:36 |
kfox1111 | we just have to seperate config from the container. | 19:36 |
kfox1111 | so the config outlasts the container. | 19:37 |
kfox1111 | you then think of the container as the compiled program. | 19:37 |
rhallisey | looks good kfox1111 | 19:37 |
kfox1111 | I do stuff like 'docker -it --rm -v/a:/b mycontainer myprogram "$*" ' in a bash script. | 19:37 |
rhallisey | we'll move to two +2's when the core team expands | 19:37 |
kfox1111 | then you can call "myprogram --foo" and it does stuff. but all containerized. :) | 19:38 |
kfox1111 | rhallisey: thanks. | 19:38 |
kfox1111 | yeah. more eyes woudl be good. but its just too hard with so few people. | 19:38 |
v1k0d3n | yeah, i get it. | 19:38 |
openstackgerrit | Merged openstack/kolla-kubernetes: Modular gate https://review.openstack.org/386966 | 19:38 |
*** dave-mccowan has quit IRC | 19:39 | |
openstackgerrit | Merged openstack/kolla-kubernetes: Multinode Gate & Misc fixes https://review.openstack.org/389299 | 19:39 |
sdake | kfox1111 sbezverk is on pto until the 15th | 19:40 |
kfox1111 | sdake: ok. thanks. :) | 19:41 |
kfox1111 | amrith | 19:42 |
*** awidders_ has joined #openstack-kolla | 19:43 | |
*** msimonin has joined #openstack-kolla | 19:44 | |
*** awiddersheim has quit IRC | 19:45 | |
*** dwalsh has joined #openstack-kolla | 19:46 | |
*** msimonin1 has joined #openstack-kolla | 19:50 | |
*** dwalsh has quit IRC | 19:50 | |
*** msimonin has quit IRC | 19:51 | |
*** dave-mccowan has joined #openstack-kolla | 19:53 | |
*** dwalsh has joined #openstack-kolla | 19:54 | |
*** dwalsh has quit IRC | 19:56 | |
*** dwalsh has joined #openstack-kolla | 19:56 | |
v1k0d3n | v1k off one plane and then onto the next soon. | 19:58 |
v1k0d3n | later peeps | 19:58 |
*** dwalsh_ has joined #openstack-kolla | 19:59 | |
*** dwalsh has quit IRC | 19:59 | |
*** v1k0d3n has quit IRC | 20:02 | |
*** papacz has quit IRC | 20:02 | |
*** v1k0d3n has joined #openstack-kolla | 20:03 | |
*** dwalsh_ has quit IRC | 20:05 | |
*** neilus has joined #openstack-kolla | 20:06 | |
*** dwalsh has joined #openstack-kolla | 20:06 | |
*** TxGirlGeek has quit IRC | 20:09 | |
*** v1k0d3n has quit IRC | 20:10 | |
*** TxGirlGeek has joined #openstack-kolla | 20:13 | |
*** TxGirlGeek has quit IRC | 20:18 | |
*** TxGirlGeek has joined #openstack-kolla | 20:19 | |
sdake | dims about? | 20:20 |
dims | sdake : hey | 20:20 |
sdake | sup dims, you going to kubecon - iirc you weren't | 20:21 |
dims | could not make it | 20:21 |
*** dwalsh has quit IRC | 20:23 | |
*** ppalacios has quit IRC | 20:24 | |
*** dwalsh has joined #openstack-kolla | 20:24 | |
*** dwalsh has quit IRC | 20:26 | |
*** dwalsh has joined #openstack-kolla | 20:26 | |
sdake | dims right - thought so thanks | 20:26 |
*** ppalacios has joined #openstack-kolla | 20:27 | |
*** ppalacios has quit IRC | 20:28 | |
*** lrensing has quit IRC | 20:33 | |
*** matrohon has joined #openstack-kolla | 20:41 | |
*** tyrola has quit IRC | 20:50 | |
*** TxGirlGeek has quit IRC | 20:55 | |
*** TxGirlGeek has joined #openstack-kolla | 20:56 | |
*** lrensing has joined #openstack-kolla | 20:57 | |
*** ayoung has quit IRC | 21:04 | |
*** rhallisey has quit IRC | 21:12 | |
*** neilus has quit IRC | 21:16 | |
*** TxGirlGeek has quit IRC | 21:17 | |
*** fguillot has quit IRC | 21:18 | |
*** TxGirlGeek has joined #openstack-kolla | 21:18 | |
*** neilus has joined #openstack-kolla | 21:19 | |
*** schwicht has joined #openstack-kolla | 21:21 | |
*** sdake has quit IRC | 21:22 | |
*** neilus has quit IRC | 21:24 | |
*** TxGirlGeek has quit IRC | 21:27 | |
*** TxGirlGeek has joined #openstack-kolla | 21:27 | |
*** ayoung has joined #openstack-kolla | 21:27 | |
*** jemcevoy has joined #openstack-kolla | 21:28 | |
*** eaguilar has quit IRC | 21:28 | |
*** markmcclain has joined #openstack-kolla | 21:29 | |
*** neilus has joined #openstack-kolla | 21:29 | |
jemcevoy | Ho do I set the dns search path for the VM... With Nova Network I set dhcp_domain in the nova.conf. It looks like Neutron uses dns_domain but I cannot figure out how to get the value set... I created /etc/kolla/config/neutron/neutron.conf containing: [DEFAULT] \n dns_domain = scm.penguincomputing.com penguincomputing.com | 21:33 |
jemcevoy | then ran /home/jmcevoy/Kolla/kolla/tools/kolla-ansible reconfigure -i /home/jmcevoy/Kolla/kolla/ansible/inventory/multinode | 21:34 |
*** TxGirlGeek has quit IRC | 21:34 | |
*** Jeffrey4l has quit IRC | 21:35 | |
*** TxGirlGeek has joined #openstack-kolla | 21:35 | |
*** dwalsh has quit IRC | 21:35 | |
*** lrensing has quit IRC | 21:36 | |
*** srwilkers has quit IRC | 21:37 | |
*** neilus has quit IRC | 21:48 | |
*** neilus has joined #openstack-kolla | 21:50 | |
*** dwalsh has joined #openstack-kolla | 21:50 | |
*** jheroux has quit IRC | 21:55 | |
*** schwicht has quit IRC | 21:58 | |
*** v1k0d3n has joined #openstack-kolla | 22:06 | |
*** neilus has quit IRC | 22:07 | |
*** schwicht has joined #openstack-kolla | 22:15 | |
imcsk8 | is there a way to stop a service using a kolla command? | 22:21 |
*** v1k0d3n has quit IRC | 22:22 | |
*** v1k0d3n has joined #openstack-kolla | 22:22 | |
*** schwicht has quit IRC | 22:23 | |
Pavo | evening | 22:24 |
v1k0d3n | afternoon | 22:24 |
v1k0d3n | :) | 22:24 |
*** schwicht has joined #openstack-kolla | 22:25 | |
*** radeksmg has quit IRC | 22:26 | |
*** absubram has quit IRC | 22:26 | |
*** matrohon has quit IRC | 22:31 | |
*** Guest92_ has quit IRC | 22:34 | |
*** fragatin_ has joined #openstack-kolla | 22:37 | |
*** v1k0d3n has quit IRC | 22:38 | |
*** v1k0d3n has joined #openstack-kolla | 22:39 | |
*** fragatina has quit IRC | 22:40 | |
*** v1k0d3n has quit IRC | 22:40 | |
*** v1k0d3n has joined #openstack-kolla | 22:41 | |
*** v1k0d3n has quit IRC | 22:42 | |
*** srwilkers has joined #openstack-kolla | 22:42 | |
*** v1k0d3n has joined #openstack-kolla | 22:43 | |
*** v1k0d3n has quit IRC | 22:44 | |
*** v1k0d3n has joined #openstack-kolla | 22:46 | |
*** neilus has joined #openstack-kolla | 22:46 | |
*** neilus has quit IRC | 22:51 | |
*** srwilkers has quit IRC | 23:00 | |
*** msimonin1 has quit IRC | 23:05 | |
*** schwicht has quit IRC | 23:05 | |
*** srwilkers has joined #openstack-kolla | 23:05 | |
*** v1k0d3n has quit IRC | 23:09 | |
openstackgerrit | BorneMace proposed openstack/kolla: Added support for a stop playbook in kolla https://review.openstack.org/388254 | 23:09 |
*** lamt has quit IRC | 23:10 | |
*** Guest80760 has joined #openstack-kolla | 23:11 | |
*** pbourke has quit IRC | 23:13 | |
*** imcsk8 has quit IRC | 23:16 | |
*** Guest80760 has quit IRC | 23:18 | |
*** dwalsh has quit IRC | 23:19 | |
*** gfidente has quit IRC | 23:21 | |
*** Guest92 has joined #openstack-kolla | 23:21 | |
*** srwilkers has left #openstack-kolla | 23:23 | |
*** msimonin has joined #openstack-kolla | 23:24 | |
*** b_bezak has joined #openstack-kolla | 23:24 | |
*** msimonin has quit IRC | 23:25 | |
*** srwilkers has joined #openstack-kolla | 23:28 | |
*** schwicht has joined #openstack-kolla | 23:30 | |
*** ayoung has quit IRC | 23:31 | |
*** Jeffrey4l has joined #openstack-kolla | 23:31 | |
*** b_bezak has quit IRC | 23:32 | |
*** b_bezak has joined #openstack-kolla | 23:32 | |
*** TxGirlGeek has quit IRC | 23:33 | |
*** Guest92_ has joined #openstack-kolla | 23:34 | |
*** ayoung has joined #openstack-kolla | 23:34 | |
*** b_bezak has quit IRC | 23:36 | |
*** Guest92 has quit IRC | 23:37 | |
*** sdake has joined #openstack-kolla | 23:44 | |
*** sdake has quit IRC | 23:45 | |
*** sdake has joined #openstack-kolla | 23:45 | |
*** neilus has joined #openstack-kolla | 23:45 | |
*** ayoung has quit IRC | 23:49 | |
*** inc0 has quit IRC | 23:52 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!