| *** tonanhngo has quit IRC | 00:04 | |
| openstackgerrit | Wei Cao proposed openstack/kolla: Add header blocks in CONTRIBUTING.rst guide https://review.openstack.org/393959 | 00:25 |
|---|---|---|
| *** saneax-_-|AFK is now known as saneax | 00:29 | |
| Pavo | https://www.youtube.com/watch?v=dXMfdpjnAHU | 00:31 |
| Pavo | opps | 00:31 |
| openstackgerrit | Wei Cao proposed openstack/kolla: remove fedora support https://review.openstack.org/394074 | 00:37 |
| *** saneax is now known as saneax-_-|AFK | 00:40 | |
| *** caowei has joined #openstack-kolla | 00:41 | |
| *** dwalsh has joined #openstack-kolla | 00:55 | |
| *** dave-mccowan has joined #openstack-kolla | 01:10 | |
| *** srwilkers has joined #openstack-kolla | 01:18 | |
| *** dave-mccowan has quit IRC | 01:23 | |
| *** neilus has joined #openstack-kolla | 01:26 | |
| *** neilus has quit IRC | 01:30 | |
| *** srwilkers has quit IRC | 01:40 | |
| *** huikang has quit IRC | 01:44 | |
| *** huikang has joined #openstack-kolla | 01:44 | |
| *** srwilkers has joined #openstack-kolla | 01:48 | |
| *** srwilkers has left #openstack-kolla | 01:48 | |
| *** huikang has quit IRC | 01:49 | |
| *** dwalsh has quit IRC | 01:57 | |
| *** huikang has joined #openstack-kolla | 02:09 | |
| *** bmace_ has quit IRC | 02:12 | |
| *** bmace_ has joined #openstack-kolla | 02:13 | |
| *** huikang has quit IRC | 02:20 | |
| *** huikang has joined #openstack-kolla | 02:40 | |
| *** haplo37 has quit IRC | 02:48 | |
| *** g3ek has quit IRC | 02:48 | |
| *** g3ek has joined #openstack-kolla | 02:54 | |
| *** haplo37 has joined #openstack-kolla | 02:55 | |
| openstackgerrit | Wei Cao proposed openstack/kolla: Add freezer container https://review.openstack.org/374908 | 03:12 |
| openstackgerrit | Wei Cao proposed openstack/kolla: Add freezer ansible role https://review.openstack.org/375874 | 03:57 |
| openstackgerrit | Merged openstack/kolla: when condition corrected in ceilometer config.yml https://review.openstack.org/394067 | 04:04 |
| openstackgerrit | Merged openstack/kolla: baremetal group defined in all-in-one https://review.openstack.org/393832 | 04:19 |
| openstackgerrit | Merged openstack/kolla: Add header blocks to Dockerfiles https://review.openstack.org/393955 | 04:20 |
| *** sdake has joined #openstack-kolla | 04:35 | |
| openstackgerrit | Wei Cao proposed openstack/kolla: Add solum ansible role https://review.openstack.org/358938 | 04:38 |
| openstackgerrit | Merged openstack/kolla: Remove hostname from 127.0.0.1 https://review.openstack.org/388722 | 04:42 |
| openstackgerrit | prameswar proposed openstack/kolla: Custom policy files https://review.openstack.org/394147 | 04:44 |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 04:46 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Use openstack_service_workers variable in gate https://review.openstack.org/394148 | 04:49 |
| openstackgerrit | prameswar proposed openstack/kolla: when condition corrected in ceilometer config.yml https://review.openstack.org/394149 | 04:49 |
| openstackgerrit | Merged openstack/kolla: Remove inputs.filestat from telegraf configuration https://review.openstack.org/387935 | 04:58 |
| openstackgerrit | Merged openstack/kolla: Include oslo.vmware along with other oslo packages in openstack-base. https://review.openstack.org/393744 | 04:59 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Install all python packages in global-requirements.txt file https://review.openstack.org/394150 | 05:04 |
| sdake | lots o patches | 05:07 |
| sdake | evening folks | 05:07 |
| *** huikang has quit IRC | 05:07 | |
| sdake | does this channel ever sleep ;) | 05:07 |
| *** Jeffrey4l has joined #openstack-kolla | 05:13 | |
| openstackgerrit | Merged openstack/kolla: Add reconfigure.yml for bifrost https://review.openstack.org/390880 | 05:18 |
| sdake | Jeffrey4l did you see my emial re the repo split | 05:24 |
| * Jeffrey4l is reading sdake | 05:25 | |
| sdake | the tldr is I may need a little help ;) | 05:25 |
| sdake | (on project-config) | 05:25 |
| sdake | I'll start the work on the 10th when i return from travel | 05:26 |
| Jeffrey4l | gate fix should be the hardest, ino | 05:26 |
| Jeffrey4l | imo | 05:26 |
| sdake | agree | 05:26 |
| sdake | i was hoping we cant ackle it together | 05:27 |
| Jeffrey4l | np | 05:27 |
| sdake | after i get through the rigamarlole required by the repo split | 05:27 |
| sdake | not that i don't thikn others could do the job, but if we could just wait until the 10th when i return, I will get cracking | 05:28 |
| sdake | because the repo additions in particular are tricky | 05:28 |
| sdake | not hard, just tricky | 05:28 |
| Jeffrey4l | i do not think other guys is working this expect u. ;) | 05:28 |
| Jeffrey4l | yep. git history. | 05:28 |
| sdake | do you mean others expect me to tackle this? | 05:29 |
| sdake | (your sentence didn't parse for me) | 05:29 |
| Jeffrey4l | i mean other guys is not working on repo split. | 05:29 |
| Jeffrey4l | btw, i can take the work of fixing gate jobs. ;) | 05:30 |
| sdake | cool - i will help | 05:37 |
| sdake | i want to get back into dev more | 05:37 |
| sdake | specificially around kolla-kubernetes | 05:37 |
| sdake | but job needs to be finished on repo split first | 05:38 |
| sdake | ptl is sort of an engineering management job of a project as busy as kolla | 05:38 |
| sdake | feel a bit like i'm out of the tech | 05:38 |
| sdake | whereas I used to write half the commits to the repo | 05:38 |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 05:45 |
| openstackgerrit | Merged openstack/kolla: Fix broken test in vagrant boostrap script https://review.openstack.org/389266 | 06:00 |
| openstackgerrit | Merged openstack/kolla: nova-spicehtml5proxy dependency package added https://review.openstack.org/392896 | 06:00 |
| *** sdake_ has joined #openstack-kolla | 06:05 | |
| *** sdake has quit IRC | 06:05 | |
| *** sdake has joined #openstack-kolla | 06:07 | |
| *** sdake_ has quit IRC | 06:10 | |
| openstackgerrit | prameswar proposed openstack/kolla: nova-spicehtml5proxy dependency package added https://review.openstack.org/394151 | 06:14 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Make service_checks easier to run independently of kolla-ansible https://review.openstack.org/381161 | 06:16 |
| openstackgerrit | Merged openstack/kolla: Fix plugin parsing for base images https://review.openstack.org/393291 | 06:16 |
| openstackgerrit | Praveen N proposed openstack/kolla: fix magnum [keystone_authtoken] configuration https://review.openstack.org/388305 | 06:36 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Adding logrotate configuration for ceilometer https://review.openstack.org/389710 | 06:40 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Check api_interface instead of network_interface https://review.openstack.org/386654 | 06:43 |
| *** tonanhngo has joined #openstack-kolla | 06:44 | |
| openstackgerrit | Merged openstack/kolla: Add upgrade.yml for bifrost https://review.openstack.org/390863 | 06:45 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Deploy Telegraf on all nodes https://review.openstack.org/388885 | 06:55 |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 07:17 |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 07:23 |
| *** csj has joined #openstack-kolla | 07:29 | |
| openstackgerrit | prameswar proposed openstack/kolla: disable firewall in bootstrap-server https://review.openstack.org/393102 | 07:30 |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 07:32 |
| *** tonanhngo has quit IRC | 07:35 | |
| *** csj has quit IRC | 07:39 | |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 07:45 |
| *** csj has joined #openstack-kolla | 07:58 | |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 07:59 |
| *** sdake has quit IRC | 08:08 | |
| *** csj has quit IRC | 08:10 | |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 08:11 |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 08:16 |
| openstackgerrit | zhubingbing proposed openstack/kolla: add panko dockerfile https://review.openstack.org/387232 | 08:25 |
| *** saneax-_-|AFK is now known as saneax | 08:52 | |
| *** saneax is now known as saneax-_-|AFK | 09:03 | |
| *** dwalsh has joined #openstack-kolla | 09:52 | |
| *** dwalsh has quit IRC | 10:01 | |
| *** caowei has quit IRC | 11:41 | |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla: Add custom policies in service.json https://review.openstack.org/394177 | 12:46 |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla: Add custom policies in service.json https://review.openstack.org/394177 | 12:48 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Uses tunnel_interface as ovs tunnel instead of api_interface https://review.openstack.org/392094 | 12:51 |
| *** neilus has joined #openstack-kolla | 12:53 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: remove override_install_langs option in /etc/yum.conf https://review.openstack.org/391631 | 12:57 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Configure log path for kibana https://review.openstack.org/394178 | 13:01 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: updates sources.list and apt prefereses filename to match dockerfile. https://review.openstack.org/386592 | 13:06 |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla: Allow flexibility in mysql backend for ceilometer https://review.openstack.org/392111 | 13:08 |
| *** Jeffrey4l has quit IRC | 13:19 | |
| *** neilus has quit IRC | 13:21 | |
| *** neilus has joined #openstack-kolla | 13:50 | |
| *** neilus has quit IRC | 13:55 | |
| Pavo | morning | 14:00 |
| *** srwilkers has joined #openstack-kolla | 14:15 | |
| *** dwalsh has joined #openstack-kolla | 14:39 | |
| *** neilus has joined #openstack-kolla | 14:45 | |
| *** neilus has quit IRC | 14:46 | |
| *** neilus has joined #openstack-kolla | 14:47 | |
| *** dwalsh has quit IRC | 14:56 | |
| *** srwilkers has quit IRC | 15:06 | |
| *** srwilkers has joined #openstack-kolla | 15:07 | |
| *** csj has joined #openstack-kolla | 15:32 | |
| csj | hi, I use kolla+vagrant to install a AIO. as the document mentioned `source ~/openrc`, I don't find this file in my operator kvm :( | 15:33 |
| csj | doc is here: http://docs.openstack.org/developer/kolla/devenv-vagrant.html | 15:34 |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla: Horizon custom policies https://review.openstack.org/394187 | 15:47 |
| *** sdake has joined #openstack-kolla | 16:31 | |
| *** Pavo has quit IRC | 16:45 | |
| *** Pavo has joined #openstack-kolla | 16:49 | |
| *** v1k0d3n has joined #openstack-kolla | 16:50 | |
| *** g3ek has quit IRC | 16:52 | |
| *** haplo37 has quit IRC | 16:53 | |
| *** haplo37 has joined #openstack-kolla | 16:53 | |
| *** g3ek has joined #openstack-kolla | 16:54 | |
| *** matrohon has joined #openstack-kolla | 17:21 | |
| *** srwilkers has quit IRC | 17:35 | |
| *** matrohon has quit IRC | 18:02 | |
| *** neilus has quit IRC | 18:07 | |
| *** bmace_ has quit IRC | 18:13 | |
| *** HyperJohnGraham has joined #openstack-kolla | 18:22 | |
| *** sdake has quit IRC | 18:34 | |
| *** neilus has joined #openstack-kolla | 18:37 | |
| *** neilus has quit IRC | 18:39 | |
| *** dave-mccowan has joined #openstack-kolla | 18:42 | |
| *** sdake has joined #openstack-kolla | 18:44 | |
| sdake | hey peeps | 18:47 |
| *** dave-mccowan has quit IRC | 18:48 | |
| openstackgerrit | Merged openstack/kolla: nova-spicehtml5proxy dependency package added https://review.openstack.org/394151 | 19:30 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Fix typo in doc/bifrost.rst https://review.openstack.org/392625 | 19:48 |
| openstackgerrit | Merged openstack/kolla: Fix typo in doc/bifrost.rst https://review.openstack.org/392625 | 19:51 |
| openstackgerrit | Steven Dake proposed openstack/kolla: remove override_install_langs option in /etc/yum.conf https://review.openstack.org/394200 | 19:52 |
| openstackgerrit | Merged openstack/kolla: remove override_install_langs option in /etc/yum.conf https://review.openstack.org/391631 | 19:54 |
| *** jtriley has joined #openstack-kolla | 20:03 | |
| openstackgerrit | Merged openstack/kolla: Allow flexibility in mysql backend for ceilometer https://review.openstack.org/392111 | 20:04 |
| openstackgerrit | Merged openstack/kolla: Adding logrotate configuration for ceilometer https://review.openstack.org/389710 | 20:04 |
| openstackgerrit | Merged openstack/kolla: Remove haproxy_socket volume from telegraf container https://review.openstack.org/387639 | 20:11 |
| *** jtriley has quit IRC | 20:13 | |
| *** matrohon has joined #openstack-kolla | 20:19 | |
| Pavo | sup sdake | 20:23 |
| sdake | yo pavo | 20:23 |
| Pavo | hows it been | 20:23 |
| sdake | i've been looking into that problem related to fips-140 for you | 20:23 |
| Pavo | awesome | 20:23 |
| sdake | the short answer is we use a library called pycrypto | 20:24 |
| sdake | i don't yet know if this library is fips-140 compliant, but the general belief on the ml is it is not | 20:24 |
| Pavo | as long as there nothing that md5 hashes its fine | 20:24 |
| sdake | we dont use md5 in our code - but pycrypto implements md5 i suspect | 20:24 |
| Pavo | needs to be atleast sha-256 | 20:24 |
| sdake | md5 is insecure as you probably are aware | 20:25 |
| Pavo | yeap | 20:25 |
| Pavo | so I have been playing around with custom ini files | 20:25 |
| Pavo | can't seem to get it working | 20:25 |
| Pavo | guess I am just doing it wrong | 20:25 |
| sdake | possible | 20:25 |
| sdake | tell me what your doing | 20:25 |
| Pavo | I created a folder in /etc/kolla named config | 20:26 |
| Pavo | inside that folder I have a neutron-server folder and inside that folder I have a ml2.conf.ini | 20:26 |
| Pavo | I mean ml2_conf.ini | 20:26 |
| Pavo | I added tenant_network_types = vxlan, vlan under [ml2] section and did a kolla-ansible reconfigure | 20:27 |
| Pavo | checked the new deployment, nothing changed | 20:27 |
| Pavo | so either I am not understanding the custom ini files or configured the folders wrongly | 20:28 |
| Pavo | dunno | 20:28 |
| Pavo | there are a few changes I have to have customized for our environment and normally changed them in their respective ini files, like ml2 and dhcp_agent and horizon-dashboard.ini | 20:30 |
| Pavo | but for the life of me can not figure how to change them with kolla | 20:30 |
| openstackgerrit | Merged openstack/kolla: Add profile usage to kolla-build guide https://review.openstack.org/378230 | 20:39 |
| sdake | pavo sorry i got mixed up in some work :) | 20:40 |
| Pavo | no problem | 20:40 |
| openstackgerrit | Merged openstack/kolla: Adjust object-replicator heading within object.conf.j2 template. https://review.openstack.org/391521 | 20:40 |
| sdake | pavo rename neutron-server to ust neutron i think | 20:40 |
| sdake | pavo also i can't guarantee that ml2 reconfiguration works - it should but i have never personally tested going rom vlan to vxlan or visa-versas | 20:41 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Adjust object-replicator heading within object.conf.j2 template. https://review.openstack.org/394203 | 20:41 |
| Pavo | yeah I am not going from to, I am just adding to have more options | 20:42 |
| Pavo | but I will try just using neutron as the folder name | 20:42 |
| sdake | ya so for top level services our model is config/nova/nova.conf not /config/nova/nova-conductor/nova.conf | 20:45 |
| sdake | we used to have the second convention, but IIRC it was removed. | 20:45 |
| Pavo | in the source code, where is the code that looks for theses ini files? | 20:46 |
| Pavo | because there has to be a list of what kolla will except and what it will not except | 20:46 |
| Pavo | because its really hard to know what ini file to change with there being the same ini file for multiple containers | 20:48 |
| *** srwilkers has joined #openstack-kolla | 20:55 | |
| openstackgerrit | Steven Dake proposed openstack/kolla: Adding logrotate configuration for ceilometer https://review.openstack.org/394204 | 20:56 |
| sdake | pavo moment | 20:56 |
| sdake | pavo https://github.com/openstack/kolla/blob/master/ansible/roles/nova/tasks/config.yml#L51-L57 | 20:57 |
| Pavo | so its located in the roles folder for every service? | 20:58 |
| Pavo | ok well changing it from neutron-server to just neutron worked | 21:00 |
| Pavo | this is going to be VERY tedious task to narrow down which ini files I need to customize | 21:00 |
| Pavo | sdake take a look at this http://ddi.hopto.org:3000/henry.brock/OpenstackDDI/src/master/config/install.sh this is all the settings I use with packstack, which set the settings in the ini files for said service | 21:02 |
| sdake | pavo one option you ahve is just to copy your ini files over | 21:02 |
| Pavo | there are a ton that doesn't reflect what kolla names them as, so it going to take some time to figure which ones kolla names them as to customize | 21:02 |
| *** jtriley has joined #openstack-kolla | 21:02 | |
| sdake | pavo without doing a three way diff :) | 21:03 |
| Pavo | true | 21:03 |
| sdake | its not "optimal' but should get you going | 21:03 |
| Pavo | yeah | 21:04 |
| openstackgerrit | Merged openstack/kolla: Add freezer container https://review.openstack.org/374908 | 21:07 |
| *** jtriley has quit IRC | 21:07 | |
| *** jtriley has joined #openstack-kolla | 21:08 | |
| openstackgerrit | Merged openstack/kolla: Add header blocks in CONTRIBUTING.rst guide https://review.openstack.org/393959 | 21:08 |
| *** neilus has joined #openstack-kolla | 21:11 | |
| *** jtriley has quit IRC | 21:13 | |
| *** dave-mccowan has joined #openstack-kolla | 21:23 | |
| openstackgerrit | Merged openstack/kolla: Mariadb performance tuning https://review.openstack.org/386130 | 21:26 |
| *** haplo37 has quit IRC | 21:39 | |
| *** g3ek has quit IRC | 21:41 | |
| *** g3ek has joined #openstack-kolla | 21:42 | |
| *** haplo37 has joined #openstack-kolla | 21:46 | |
| openstackgerrit | Merged openstack/kolla: Add file validator for *.json.j2 file https://review.openstack.org/388851 | 21:47 |
| *** v1k0d3n has quit IRC | 21:47 | |
| *** v1k0d3n has joined #openstack-kolla | 21:48 | |
| openstackgerrit | Steven Dake proposed openstack/kolla: Add file validator for *.json.j2 file https://review.openstack.org/394206 | 21:50 |
| sdake | sup v1k0d3n | 21:50 |
| *** v1k0d3n has quit IRC | 21:51 | |
| *** v1k0d3n has joined #openstack-kolla | 21:51 | |
| v1k0d3n | hey sdake what's up man | 21:52 |
| v1k0d3n | when are you taking off for kubecon? | 21:52 |
| sdake | 3pm - arrive 5pm | 21:56 |
| *** dave-mcc_ has joined #openstack-kolla | 22:04 | |
| *** dave-mccowan has quit IRC | 22:06 | |
| *** jtriley has joined #openstack-kolla | 22:09 | |
| *** matrohon has quit IRC | 22:10 | |
| *** dave-mccowan has joined #openstack-kolla | 22:13 | |
| *** jtriley has quit IRC | 22:14 | |
| sdake | v1k0d3n if you are not aware, we are splitting the repositories of kolla into 3 | 22:14 |
| sdake | https://www.gliffy.com/go/publish/11338125 | 22:14 |
| sdake | there is a diagram, process that and tell me what you think | 22:15 |
| sdake | have you had an opportunity to check out ryan's helm spec? | 22:15 |
| sdake | in ryan's spec, there is some discussion of using entrypoint project | 22:16 |
| sdake | however, this is problematic I htink because entrypoint provides its own JSON API model | 22:16 |
| *** dave-mcc_ has quit IRC | 22:17 | |
| sdake | and we already have an API model in Kolla | 22:17 |
| *** dave-mccowan has quit IRC | 22:17 | |
| sdake | making these two inconsistent models work well together seems troubling to me | 22:17 |
| *** neilus has quit IRC | 22:21 | |
| *** v1k0d3n has quit IRC | 22:30 | |
| *** v1k0d3n has joined #openstack-kolla | 22:35 | |
| *** v1k0d3n has quit IRC | 22:38 | |
| *** v1k0d3n has joined #openstack-kolla | 22:38 | |
| *** neilus has joined #openstack-kolla | 22:39 | |
| v1k0d3n | not sure how to correct the API issue, just know that having a self-contained, controller/operator model makes the solution cloud native. | 22:40 |
| v1k0d3n | not sure if current isn't exactly cloud-native, but having a watchdog is what we're going to be after eventually. | 22:41 |
| v1k0d3n | even entrypoint will eventually turn into that operator model like: https://coreos.com/blog/introducing-the-etcd-operator.html | 22:41 |
| v1k0d3n | entrypoint is most likely a precursor to a much bigger movement, which seems to be supported by others from what i can tell. | 22:42 |
| v1k0d3n | sdake: ^^ | 22:43 |
| *** neilus has quit IRC | 22:44 | |
| sdake | other people that have tried that model didn't like it | 22:44 |
| sdake | it works 95% of the time | 22:44 |
| sdake | 5% of the time is fails. | 22:45 |
| sdake | we tried heavily during compose work | 22:45 |
| sdake | and asalkeld and co tried again with mesos | 22:45 |
| sdake | a completely decentralized model of self-orchestration | 22:45 |
| sdake | so looking at the etcd-oeprator github i see this: | 22:46 |
| sdake | The controller collects anonymous usage statistics to help us learning how the software is being used and how we can improve it. To disbale collection, run the controller with the flag -analytics=false | 22:46 |
| sdake | groan... | 22:46 |
| sdake | https://github.com/coreos/etcd-operator | 22:46 |
| sdake | the etcd operator is only for etcd | 22:47 |
| sdake | sounds like whta we really need is a kolla-operator | 22:47 |
| sdake | as in rework kolla_start.sh into kolla_start_operator.sh | 22:47 |
| sdake | but again it didn't work in the past ;) | 22:47 |
| sdake | two separate time | 22:48 |
| *** dave-mccowan has joined #openstack-kolla | 22:51 | |
| sdake | s | 22:55 |
| sdake | v1k0d3n are you napping at the keyboard again ;) | 22:56 |
| sdake | v1k0d3n that etcd_operator tool is written by two cats | 22:58 |
| sdake | https://github.com/coreos/etcd-operator/graphs/contributors | 22:58 |
| sdake | while that particular implementation looks appealing, i'd like to see something a little more resilient to busses | 22:58 |
| sdake | (noting it only works for etcd:) | 22:59 |
| v1k0d3n | yeah, from what i've been discussing with a lot of folks...the ultimate goal would be to have app-specific operators. | 23:00 |
| v1k0d3n | for each specific app logic. | 23:00 |
| sdake | cool | 23:00 |
| sdake | so have you seen this? https://github.com/stackanetes/kubernetes-entrypoint | 23:00 |
| v1k0d3n | and really only for things that aren't cloud-native to start. galera/maria, rabbitmq, memcache, etc. for our purposes if it would be considered. | 23:00 |
| v1k0d3n | yeah, that's what i've been talking about over the past couple of weeks | 23:01 |
| sdake | define cloud native for me please | 23:01 |
| v1k0d3n | (operator was just "released" this past week). | 23:01 |
| sdake | this is the api to that code base: https://github.com/stackanetes/stackanetes/blob/master/nova/manifest.jsonnet | 23:01 |
| sdake | ya probably a private repo then | 23:01 |
| sdake | i guess i saw "July" as the start point but if it was a private repo, then i'd see why it only has two authors | 23:02 |
| v1k0d3n | container, dynamically scheduled, microservice in nature. | 23:02 |
| v1k0d3n | the rub is in the dynamically scheduled...because if it's in "any old order" things like galera, rabbit...those aren't reliable anymore working with other applications. | 23:02 |
| *** jtriley has joined #openstack-kolla | 23:03 | |
| v1k0d3n | it was private. this was just release this past tuesday or wedsnesday maybe? | 23:03 |
| sdake | could you back that with an argument :) | 23:03 |
| v1k0d3n | in what way do you mean? | 23:03 |
| sdake | "aren't reliable anymore" | 23:03 |
| sdake | or facts | 23:04 |
| sdake | or something rather then speculation ;) | 23:04 |
| sdake | not that rabbitmq was all that reliable to begin with | 23:04 |
| v1k0d3n | ok, so i want to upgrade newton to ocata.... | 23:04 |
| sdake | but lets leave that part out | 23:04 |
| v1k0d3n | then i want to downgrade the entire openstack deployment. | 23:04 |
| sdake | downgrade is a terrible example | 23:04 |
| v1k0d3n | cloud-native would assume that this can be done | 23:04 |
| sdake | because downgrade doesn't work with databases | 23:04 |
| sdake | of any type | 23:04 |
| sdake | any time you have changing state, downgrade becomes a non-starter | 23:05 |
| v1k0d3n | i think the end goal is that operator makes this possible eventually. | 23:05 |
| sdake | can you pick a different example that i'll actually buy ;) | 23:05 |
| sdake | v1k0d3n - for downgrade to work, database migrations would have to be unwound | 23:05 |
| sdake | this is not possible in a SQL environment | 23:05 |
| sdake | and not really possible in a pure KV environment either | 23:05 |
| v1k0d3n | ok | 23:05 |
| v1k0d3n | rabbit | 23:06 |
| v1k0d3n | i'm just saying. | 23:06 |
| sdake | ya rabbit is full of trouble, but it has nothign to do with dynamic scheduling | 23:06 |
| sdake | and everything to do with erlang | 23:06 |
| *** jtriley has quit IRC | 23:08 | |
| v1k0d3n | i'm just worried that kolla-kubernetes could be held back because of the need for kolla-ansible. | 23:08 |
| sdake | could you expand on that | 23:08 |
| v1k0d3n | some of the ansible items are addressed with other, simpler options and because of the basic nature of a kubernetes deployment. | 23:09 |
| v1k0d3n | why ansible? | 23:09 |
| sdake | no the held back part | 23:09 |
| v1k0d3n | why not use entry point because it breaks an ansible deployment | 23:09 |
| v1k0d3n | ? | 23:09 |
| sdake | you mean a comma to be in that sentence? | 23:09 |
| sdake | why not use entry pooint; because it breaks an ansible deployment? | 23:10 |
| *** saneax-_-|AFK is now known as saneax | 23:10 | |
| v1k0d3n | i'd rather talk in person. just have too much going on (making dinner, errands, etc). | 23:11 |
| sdake | i understand but would like to hear your take before we get to kubecon | 23:11 |
| v1k0d3n | it's always going to come to this discussion, i guess. | 23:11 |
| v1k0d3n | we're looking for a controller. i think SAP is looking for that too. | 23:12 |
| v1k0d3n | others too. | 23:12 |
| sdake | right a controller | 23:12 |
| v1k0d3n | just want to remove parity. | 23:12 |
| sdake | would you consider entrypoint a controller? | 23:12 |
| v1k0d3n | it's a start; more than what kolla has today for kubernetes. | 23:12 |
| sdake | i understand it provides some functionality | 23:13 |
| sdake | (keep in mind i'm coing up to speed) | 23:13 |
| sdake | my take on controllers is they are sort of "top down" | 23:13 |
| v1k0d3n | i know...it's all good. | 23:13 |
| sdake | meaning they operate with dependencies in order | 23:13 |
| *** saneax is now known as saneax-_-|AFK | 23:13 | |
| v1k0d3n | the operator (etcd/prometheus) is a controller for sure. | 23:14 |
| v1k0d3n | that's the reason. | 23:14 |
| sdake | ok,, and entrypoint is what? | 23:14 |
| sdake | besides a start - does it have a clasification | 23:14 |
| v1k0d3n | i said earlier, precursor to this etcd operator concept. | 23:15 |
| v1k0d3n | lesson's learned. | 23:15 |
| sdake | so why not just write an openstack operator? | 23:15 |
| v1k0d3n | but i think that it will end up being a controller eventually for that project. i would love to see that work in kolla, rather than somewhere else. | 23:15 |
| sdake | that would be compatible with an ansible deployment. | 23:15 |
| v1k0d3n | that's my suggestion :) | 23:15 |
| v1k0d3n | that's been my whole point of this convo. hahaha | 23:15 |
| sdake | i thought the whole point you were making is we should use entrypoint ;) | 23:16 |
| v1k0d3n | why does it need to compatible with ansible...just a hypothetical question? | 23:16 |
| v1k0d3n | until earlier this week...yes. because it's a start. | 23:16 |
| sdake | sure - because we maintain one set of containers for all orchestration systems | 23:16 |
| sdake | not two sets | 23:16 |
| sdake | one for ansible one for kubernetes iss nutty | 23:16 |
| v1k0d3n | ^^ agreed. | 23:17 |
| v1k0d3n | but kolla cannot be deployed with salt | 23:17 |
| v1k0d3n | or puppet | 23:17 |
| v1k0d3n | today | 23:17 |
| v1k0d3n | but if entrypoint was used... | 23:17 |
| sdake | who cares | 23:17 |
| v1k0d3n | it could be. | 23:17 |
| v1k0d3n | EVERYONE CARES | 23:17 |
| sdake | about salt and puppet? | 23:17 |
| v1k0d3n | that's the single run i think | 23:17 |
| v1k0d3n | yes. | 23:17 |
| v1k0d3n | it's containers...why pick an orchestrator? that becomes opinionated then. | 23:18 |
| sdake | we haven't picked an orchestrator | 23:18 |
| sdake | we have chosen a json API to the containers however | 23:18 |
| sdake | so rather then entrypoint, why not make our own version | 23:18 |
| v1k0d3n | is there an architecture diagram for the json API? | 23:19 |
| sdake | if we depend on a third party dep for our core runtime of containers it will come to haunt us | 23:19 |
| v1k0d3n | maybe that would help? | 23:19 |
| sdake | v1k0d3n i dont have one | 23:19 |
| sdake | i can point you at the json api exmaple if you like | 23:19 |
| sdake | the reason the json api was invented was so that multiple orchestration engines could be used | 23:20 |
| sdake | entrypoint took angus's work there and expanded it | 23:20 |
| sdake | but it was never rolled back upstream | 23:21 |
| sdake | instead it was forked off into a new thing | 23:21 |
| *** neilus has joined #openstack-kolla | 23:21 | |
| sdake | without our original requirements being met | 23:21 |
| sdake | fwiw when angus and i had this conversation with a bunch of other people in the channel he said that his approach was wrong | 23:24 |
| sdake | that was the second failure I was speaking of earlier | 23:24 |
| sdake | not that I dont mind trying again | 23:24 |
| sdake | if we are going to make an openstack operator, I think that makes perfect sense - lets make it work with everything we care about today (ansible, kubernetes) using the json API we have today | 23:25 |
| *** HyperJohnGraham has quit IRC | 23:27 | |
| sdake | v1k0d3n have you seen this tweet: https://twitter.com/robynbergeron/status/792104354899787776 | 23:27 |
| sdake | v1k0d3n i could see a future outcome where each type of application (that is complex) is controlled by an operator | 23:34 |
| sdake | i dont think that is the same thing as entrypoint | 23:34 |
| sdake | entrypoint half of the equation there | 23:34 |
| v1k0d3n | it's not | 23:37 |
| v1k0d3n | at summit...entrypoint was handing dependancies via container, not an orchestrator. | 23:37 |
| v1k0d3n | kubernetes does not need an orchestrator. | 23:37 |
| v1k0d3n | kolla-ansible does | 23:38 |
| v1k0d3n | so ansible holds it back | 23:38 |
| v1k0d3n | operator is new, better. | 23:38 |
| v1k0d3n | but does ansible need it? | 23:38 |
| sdake | you mean OPENSTACK needs an orchestrator | 23:38 |
| v1k0d3n | yes | 23:38 |
| v1k0d3n | but we are talking about kolla | 23:38 |
| sdake | kolla-ansible just delivers on the requirements | 23:38 |
| v1k0d3n | because kolla takes on building openstack | 23:39 |
| sdake | in the future it doess | 23:39 |
| v1k0d3n | if kolla was just containers, it would solve this via container | 23:39 |
| *** neilus has quit IRC | 23:39 | |
| sdake | if you are talking of the repo | 23:39 |
| sdake | ok, well thats not going to ever happen | 23:39 |
| sdake | do you know kolla's mission? | 23:39 |
| v1k0d3n | i belive containerized openstack | 23:40 |
| v1k0d3n | ? | 23:40 |
| sdake | https://twitter.com/robynbergeron/status/792104354899787776 | 23:41 |
| sdake | wrong pste | 23:41 |
| sdake | moment | 23:41 |
| sdake | https://github.com/openstack/governance/blob/master/reference/projects.yaml#L2458-L2459 | 23:41 |
| v1k0d3n | lol first one is funny considering though :) | 23:41 |
| sdake | ya it was already in my paste buffer when i mentioned it above a screen ago ;) | 23:42 |
| sdake | so our mission isn't really "containerized openstack" | 23:42 |
| v1k0d3n | yeah, so i understand | 23:42 |
| sdake | people could interpret that to mean "to put openstack in containers." | 23:42 |
| sdake | our mission is to do that AND operate openstack with them | 23:42 |
| v1k0d3n | i consider it to be an openstack deployment that is prod-ready, just using containers. | 23:42 |
| v1k0d3n | i get it. | 23:43 |
| v1k0d3n | but what kubernetes needs is different | 23:43 |
| sdake | ok - the reason i mention all this | 23:43 |
| sdake | is because there was alot of contention early on | 23:43 |
| *** neilus has joined #openstack-kolla | 23:43 | |
| sdake | and it persists to this day | 23:43 |
| v1k0d3n | a container get's whacked and comes up somewhere else...is something going to orchestrate or operate that? | 23:43 |
| sdake | which is why we are making the kolla repo | 23:43 |
| sdake | within ansible? | 23:43 |
| v1k0d3n | so container native means that the container would be able to address that type of issue. | 23:43 |
| v1k0d3n | no... | 23:43 |
| v1k0d3n | nevermind. | 23:44 |
| sdake | its all good | 23:44 |
| sdake | dont get frustrated ;) | 23:44 |
| v1k0d3n | probably better convo for later. i really think we're on the same page. | 23:44 |
| v1k0d3n | but the detail is ansible, and keeping that. | 23:44 |
| v1k0d3n | i get it. | 23:44 |
| sdake | i just want to understand what you want to propose before it gets done | 23:44 |
| v1k0d3n | it NEEDS to be there. | 23:44 |
| v1k0d3n | for kolla-ansible | 23:44 |
| v1k0d3n | but if the containers are used differently (kubernetes), than a different approach should at least be explored. | 23:45 |
| sdake | totally agree | 23:45 |
| v1k0d3n | if it's not...people will HAVE to look at only parts of kolla. | 23:45 |
| sdake | we spent all week on irc brainstorming approaches | 23:45 |
| sdake | that is what kolla-kubernetes repository is about | 23:45 |
| v1k0d3n | so operator is a good path. | 23:45 |
| v1k0d3n | my favorite. | 23:45 |
| v1k0d3n | it's way better than entrypoint. | 23:45 |
| sdake | i think your missing all this brainstorming context | 23:45 |
| sdake | what is better then entrypoint? | 23:46 |
| v1k0d3n | but at the time a couple of weeks ago...entrypoint allows to not be dependent on ansible....again. | 23:46 |
| v1k0d3n | operator. i mentioned above. | 23:46 |
| v1k0d3n | was operator mentioned last week? | 23:46 |
| sdake | not at all | 23:46 |
| v1k0d3n | i wish i was in it...meant to be, this week has been rough | 23:46 |
| sdake | here are the two independent areas of research we have underway | 23:47 |
| v1k0d3n | so optional operator containers for galera/mariadb, rabbit, etc...that would be great. | 23:47 |
| sdake | ust to bring you up to speed | 23:47 |
| sdake | we have a couple dudes looking at adding dependencies to helm so helm can handle all of openstack operational experience | 23:47 |
| v1k0d3n | what is missing? | 23:48 |
| sdake | we also have couple dudes looking at integrating entrypoint into our kolla_start | 23:48 |
| *** neilus has quit IRC | 23:48 | |
| v1k0d3n | i sent some things to ryan...helm can handle dependancies. | 23:48 |
| sdake | the reason we have these dudes looking at this stuff is to determine which path to take | 23:48 |
| sdake | it sounds like your proposing a third path, an operator | 23:48 |
| v1k0d3n | but for blips, where container is rescheduled or something unexpected...operator handles this. | 23:48 |
| sdake | yes it handles them incorrectly for the kolla case however | 23:48 |
| v1k0d3n | we had a meeting with deis this week. | 23:48 |
| v1k0d3n | and also been attending the community meetings. i invited ryan to come because SAP was demoing. | 23:49 |
| sdake | ok, well it would be better if we could keep all the communicatio nin one place ;) | 23:49 |
| sdake | so are you proposing then a third line of investigation? | 23:50 |
| v1k0d3n | yes. operators. | 23:50 |
| v1k0d3n | because through discussions...it keeps coming back to this. | 23:50 |
| sdake | ok - I think your caught up on context | 23:50 |
| sdake | or possibly I'm caught up | 23:50 |
| v1k0d3n | hard to tell anymore :) things are moving so fast. | 23:51 |
| sdake | see if we have the convo in one place, then everyone is caught up all the time ;) | 23:51 |
| v1k0d3n | what time do you get in tomorrow? | 23:51 |
| sdake | 5pm | 23:51 |
| v1k0d3n | yeah, just hard to do that...because there are so many other projects involved now. | 23:51 |
| sdake | going to try to moe my flight up tho because i can | 23:51 |
| v1k0d3n | helm has it;s own meetings. kubernetes has an operators meeting, or SIG-apps. there is openstack-SIG...there's just so much. | 23:51 |
| sdake | these cats don't use irc? | 23:51 |
| v1k0d3n | nope....sorry. they use slack | 23:52 |
| v1k0d3n | it's where all the hipsters are now man :) | 23:52 |
| sdake | ya | 23:52 |
| sdake | so here is the basic deal | 23:53 |
| sdake | an operator sounds fine | 23:53 |
| sdake | i'd like to investigate all 3 approaches as a community | 23:54 |
| sdake | so we can find the best approach to proceed with | 23:54 |
| sdake | however, for kolla-kubernetes to work properly, it has to integrate with kolla containers | 23:54 |
| *** srwilkers has quit IRC | 23:54 | |
| sdake | for kolla-ansible to work properly, it has to integrate with kolla-containers | 23:54 |
| sdake | just because kubernetes is hip doesn't obliviate the fact that 16% of deployments are interested in kolla-ansible (today) | 23:55 |
| *** srwilkers has joined #openstack-kolla | 23:55 | |
| sdake | that will continue to grow | 23:55 |
| sdake | kolla-ansible is proven and based upon well tested tech | 23:55 |
| sdake | so if we make an operator, it needs to integrate with what we got | 23:56 |
| sdake | or we failed our mission | 23:56 |
| sdake | and let down the 500 people who have contributed to kolla over the last 26 months | 23:57 |
| v1k0d3n | so today, if ansible deploys out of order (serialized), will the containers be initialize correctly? | 23:57 |
| v1k0d3n | what if ansible just throws them in any order? | 23:57 |
| sdake | i totally get the problem you after | 23:57 |
| sdake | we tried it with compose | 23:57 |
| v1k0d3n | so they're not the same. as it stands today, kolla containers cannot really work for kubernetes. | 23:57 |
| sdake | we had something similar to entrypoint, but much more crude | 23:57 |
| v1k0d3n | that's the problem. | 23:57 |
| v1k0d3n | kolla-ansible needs ansible. | 23:58 |
| v1k0d3n | containers need structure. | 23:58 |
| v1k0d3n | when deployed. | 23:58 |
| v1k0d3n | i know you understand what i'm saying. | 23:58 |
| v1k0d3n | i'm just trying to drive to something better. | 23:58 |
| v1k0d3n | and i think it can be done. | 23:58 |
| sdake | it would be rockin if we could just launch all the containers and they owuld do their thing | 23:59 |
| sdake | we have tried it twice | 23:59 |
| sdake | both times failed | 23:59 |
| v1k0d3n | yeah, that's what is needed. | 23:59 |
| sdake | for different reasons | 23:59 |
| v1k0d3n | operator would fix that | 23:59 |
| v1k0d3n | even entrypoint would fix that, if the entrypoint container is called. | 23:59 |
| v1k0d3n | and that's why i say them together sometimes. | 23:59 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!