| *** eaguilar has quit IRC | 00:06 | |
| *** ssurana has joined #openstack-kolla | 00:12 | |
| *** eaguilar has joined #openstack-kolla | 00:14 | |
| *** yangyapeng has quit IRC | 00:18 | |
| *** huzhengchuan has quit IRC | 00:23 | |
| *** huzhengchuan_ has quit IRC | 00:24 | |
| *** jamesbenson has joined #openstack-kolla | 00:25 | |
| *** jamesbenson has quit IRC | 00:29 | |
| *** Pavo has joined #openstack-kolla | 00:32 | |
| *** yingjun has joined #openstack-kolla | 00:32 | |
| Pavo | quick question guys | 00:33 |
|---|---|---|
| Pavo | https://docs.openstack.org/project-deploy-guide/kolla-ansible/ocata/multinode.html says to use tools/start-registry script, just downloaded kolla using pip and there is no start-registry script | 00:34 |
| *** yingjun has quit IRC | 00:35 | |
| *** jamesbenson has joined #openstack-kolla | 00:40 | |
| *** unicell has quit IRC | 00:44 | |
| *** jamesbenson has quit IRC | 00:44 | |
| *** unicell has joined #openstack-kolla | 00:45 | |
| *** tovin07_ has joined #openstack-kolla | 00:45 | |
| *** schwicht has joined #openstack-kolla | 00:45 | |
| *** emccormick has joined #openstack-kolla | 00:46 | |
| jascott1 | pavo did you search around in repos? | 00:51 |
| jascott1 | (sorry I dont know the answer) | 00:51 |
| Pavo | no, just saying that the docs say there is a script that can start a Docker registry for you and should be located in tools directory and its not there | 00:52 |
| Pavo | only thing in tools dir is cleanup-containers cleanup-host cleanup-images stop-containers validate-docker-execute.sh | 00:53 |
| *** emccormick has quit IRC | 00:53 | |
| Pavo | now that script is in master if someone was working off that | 00:54 |
| *** Pavo has quit IRC | 00:58 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla-ansible master: [TEST][MASTER][KOLLA-ANSIBLE]TEST CLEAN CEILOMETER https://review.openstack.org/475670 | 01:01 |
| *** ssurana has quit IRC | 01:12 | |
| *** Pavo has joined #openstack-kolla | 01:17 | |
| Pavo | where is a list of all the stable/ocata images that kolla builds and has on docker hub? | 01:17 |
| Pavo | wanna just docker pull yours and not have to build | 01:18 |
| *** caowei has joined #openstack-kolla | 01:18 | |
| Pavo | because kolla-ansible pull isn't working | 01:19 |
| Pavo | also there is no kolla-build command anymore | 01:20 |
| Pavo | you guys made alot of changes I see | 01:20 |
| *** yangyapeng has joined #openstack-kolla | 01:22 | |
| *** huzhengchuan has joined #openstack-kolla | 01:26 | |
| *** yangyapeng has quit IRC | 01:26 | |
| *** huzhengchuan_ has joined #openstack-kolla | 01:26 | |
| *** yangyapeng has joined #openstack-kolla | 01:27 | |
| *** cuongnv has joined #openstack-kolla | 01:29 | |
| *** krtaylor has joined #openstack-kolla | 01:30 | |
| *** Pavo has quit IRC | 01:30 | |
| *** rwallner has quit IRC | 01:31 | |
| *** sbezverk has quit IRC | 01:34 | |
| *** caowei_ has joined #openstack-kolla | 01:36 | |
| *** caowei_1 has joined #openstack-kolla | 01:37 | |
| *** caowei has quit IRC | 01:39 | |
| *** caowei_1 is now known as caowei | 01:39 | |
| *** caowei_ has quit IRC | 01:41 | |
| *** lucasxu has joined #openstack-kolla | 01:41 | |
| *** jamesbenson has joined #openstack-kolla | 01:43 | |
| *** StephenWang1991 has joined #openstack-kolla | 01:44 | |
| *** schwicht has quit IRC | 01:46 | |
| *** jamesbenson has quit IRC | 01:47 | |
| sdake | jascott1 i was speaking with sbezverk a bit - you will be able our meeting tomorrow re deployment? | 01:47 |
| *** StephenW_ has joined #openstack-kolla | 01:47 | |
| *** yangyapeng has quit IRC | 01:48 | |
| sdake | jascott1 I think I have finally put things together in a pretty coherent way - kolla-kubernetes is more or less a bunch of helm microcharts which can be deployed in many ways | 01:48 |
| *** yangyapeng has joined #openstack-kolla | 01:48 | |
| sdake | jascott1 what we need is a second "component" that actually manages openstack (possibly modeled after kolla-ansible but specific to kubernetes) | 01:49 |
| *** StephenWang1991 has quit IRC | 01:50 | |
| *** dave-mccowan has joined #openstack-kolla | 01:58 | |
| *** StephenWang1991 has joined #openstack-kolla | 01:59 | |
| *** StephenW_ has quit IRC | 02:00 | |
| *** unicell has quit IRC | 02:00 | |
| *** rhallisey has quit IRC | 02:01 | |
| jascott1 | sdake sounds good. when is the meeting? | 02:04 |
| sdake | jascott1 not sure - inc0 mntioned it today during the team meeting | 02:04 |
| *** dave-mccowan has quit IRC | 02:05 | |
| jascott1 | cool I'll be there | 02:05 |
| sdake | jascott1 i think the deploy guide is pretty good - i haeven't tried yoru code but if its modeled after that, that might be viable | 02:05 |
| sdake | although I'd sort of like fine grained steps for deployment as well | 02:05 |
| sdake | reference destroy.ym | 02:06 |
| sdake | l | 02:06 |
| *** dave-mccowan has joined #openstack-kolla | 02:06 | |
| sdake | i think one place where we have strugged is actually sorting out that a second component is needed | 02:06 |
| jascott1 | agreed | 02:07 |
| jascott1 | or what is needed | 02:07 |
| sdake | service charts might be a thing some day | 02:07 |
| sdake | i'd rather just get on with it and have a functional implementation rather then jerk around until the end of time making thata happen ;) | 02:07 |
| *** ssurana has joined #openstack-kolla | 02:08 | |
| *** MarginHu has joined #openstack-kolla | 02:08 | |
| jascott1 | machette our way to success :) | 02:08 |
| sdake | with gates - machette wins :) | 02:08 |
| jemcevoy | inc0: I finally got to filling the bug... https://bugs.launchpad.net/kolla-ansible/+bug/1701148 it has how to repro but I have not tested the repro method... | 02:09 |
| openstack | Launchpad bug 1701148 in kolla-ansible "Ceph OSD Containers Fail when Device Names Change" [Undecided,New] | 02:09 |
| sdake | i jsut think openstack is too complicated to make helm install openstack work without any type of orchestration | 02:09 |
| jascott1 | imean im glad the dream is there | 02:09 |
| sdake | 180 microcharts | 02:09 |
| sdake | we can make helm install openstack work with a behind-the-scenes orchestration | 02:09 |
| sdake | and even helm upgrade ;) | 02:10 |
| sdake | i know some teams working on this problem are looking at helm plugins for that | 02:10 |
| sdake | to me that feels like a whole lotta pain | 02:10 |
| jascott1 | unfortunately for me I dont run a real cluster so I dont know how ppl actually subdivide and manage things | 02:10 |
| sdake | after 3 years working on kolla i have a pretty clear idea | 02:11 |
| jascott1 | seems like each larger chunk could be its own release (from helm pov) | 02:11 |
| sdake | although i also lack operational experience | 02:11 |
| *** ssurana has quit IRC | 02:13 | |
| *** jamesbenson has joined #openstack-kolla | 02:14 | |
| *** jamesbenson has quit IRC | 02:19 | |
| *** yangyapeng has quit IRC | 02:35 | |
| *** yangyapeng has joined #openstack-kolla | 02:41 | |
| *** jamesbenson has joined #openstack-kolla | 02:45 | |
| *** jamesbenson has quit IRC | 02:49 | |
| *** gfidente has quit IRC | 02:51 | |
| *** eaguilar has quit IRC | 02:55 | |
| *** dave-mccowan has quit IRC | 02:55 | |
| *** xinliang_ has quit IRC | 02:59 | |
| *** xinliang has joined #openstack-kolla | 03:00 | |
| *** salv-orlando has joined #openstack-kolla | 03:04 | |
| *** unicell has joined #openstack-kolla | 03:08 | |
| *** salv-orlando has quit IRC | 03:09 | |
| *** zhurong has quit IRC | 03:11 | |
| *** unicell has quit IRC | 03:12 | |
| spsurya | morning all | 03:14 |
| *** goldyfruit has joined #openstack-kolla | 03:22 | |
| *** unicell has joined #openstack-kolla | 03:29 | |
| *** ducttape_ has joined #openstack-kolla | 03:44 | |
| *** zhurong has joined #openstack-kolla | 03:49 | |
| *** ducttap__ has joined #openstack-kolla | 03:58 | |
| *** lucasxu has quit IRC | 04:01 | |
| *** ducttap__ has quit IRC | 04:02 | |
| *** ducttape_ has quit IRC | 04:02 | |
| *** salv-orlando has joined #openstack-kolla | 04:05 | |
| *** salv-orlando has quit IRC | 04:09 | |
| *** yuanying has joined #openstack-kolla | 04:18 | |
| *** yuanying_ has quit IRC | 04:19 | |
| *** goldyfruit has quit IRC | 04:31 | |
| *** dixiaoli has joined #openstack-kolla | 04:35 | |
| *** dixiaoli has quit IRC | 04:39 | |
| *** jamesbenson has joined #openstack-kolla | 04:45 | |
| *** jamesbenson has quit IRC | 04:49 | |
| *** yuanying has quit IRC | 05:04 | |
| *** lpetrut has joined #openstack-kolla | 05:04 | |
| *** yuanying has joined #openstack-kolla | 05:06 | |
| *** pcaruana has joined #openstack-kolla | 05:14 | |
| *** dixiaoli has joined #openstack-kolla | 05:19 | |
| *** prateek has joined #openstack-kolla | 05:25 | |
| *** ssurana has joined #openstack-kolla | 05:26 | |
| *** dixiaoli has quit IRC | 05:28 | |
| *** pcaruana has quit IRC | 05:30 | |
| *** jamesbenson has joined #openstack-kolla | 05:31 | |
| *** pcaruana has joined #openstack-kolla | 05:33 | |
| *** salv-orlando has joined #openstack-kolla | 05:34 | |
| *** jamesbenson has quit IRC | 05:35 | |
| *** zhubingbing_ has quit IRC | 05:37 | |
| *** ssurana has left #openstack-kolla | 05:39 | |
| *** pcaruana has quit IRC | 05:39 | |
| *** lpetrut has quit IRC | 05:51 | |
| *** lyang__ has joined #openstack-kolla | 05:52 | |
| *** zhurong has quit IRC | 05:56 | |
| *** skramaja has joined #openstack-kolla | 06:05 | |
| *** zhurong has joined #openstack-kolla | 06:11 | |
| *** zhurong has quit IRC | 06:19 | |
| *** zhurong has joined #openstack-kolla | 06:22 | |
| *** leseb has quit IRC | 06:25 | |
| *** salv-orlando has quit IRC | 06:25 | |
| *** salv-orlando has joined #openstack-kolla | 06:25 | |
| *** salv-orlando has quit IRC | 06:30 | |
| *** manheim has joined #openstack-kolla | 06:36 | |
| *** manheim has quit IRC | 06:41 | |
| *** serlex has joined #openstack-kolla | 06:46 | |
| *** StephenWang1991 has quit IRC | 06:47 | |
| *** manheim has joined #openstack-kolla | 06:53 | |
| *** StephenWang1991 has joined #openstack-kolla | 07:03 | |
| *** shardy has joined #openstack-kolla | 07:06 | |
| *** StephenW_ has joined #openstack-kolla | 07:07 | |
| *** StephenWang1991 has quit IRC | 07:08 | |
| *** StephenW_ has quit IRC | 07:11 | |
| *** jamesbenson has joined #openstack-kolla | 07:20 | |
| *** salv-orlando has joined #openstack-kolla | 07:21 | |
| *** jamesbenson has quit IRC | 07:24 | |
| *** reidrac has joined #openstack-kolla | 07:34 | |
| reidrac | morning | 07:34 |
| *** pcaruana has joined #openstack-kolla | 07:35 | |
| *** jamesbenson has joined #openstack-kolla | 07:36 | |
| *** StephenWang1991 has joined #openstack-kolla | 07:38 | |
| openstackgerrit | caoyuan proposed openstack/kolla-ansible master: Enable zun ui when zun enabled https://review.openstack.org/452020 | 07:39 |
| *** jamesbenson has quit IRC | 07:40 | |
| *** lpetrut has joined #openstack-kolla | 07:42 | |
| *** StephenWang1991 has quit IRC | 07:42 | |
| openstackgerrit | caoyuan proposed openstack/kolla master: Install zun ui into horizon image https://review.openstack.org/452019 | 07:44 |
| *** openstackgerrit has quit IRC | 07:47 | |
| *** StephenWang1991 has joined #openstack-kolla | 07:50 | |
| *** jamesbenson has joined #openstack-kolla | 07:51 | |
| *** egonzalez has joined #openstack-kolla | 07:52 | |
| *** jamesbenson has quit IRC | 07:56 | |
| serlex | Morning reidrac :) | 07:58 |
| *** openstackgerrit has joined #openstack-kolla | 07:58 | |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla stable/ocata: Fix rally create verifier error https://review.openstack.org/477553 | 07:58 |
| reidrac | hey serlex :) | 07:58 |
| *** gfidente has joined #openstack-kolla | 08:02 | |
| *** gfidente has quit IRC | 08:02 | |
| *** gfidente has joined #openstack-kolla | 08:02 | |
| *** junbo has quit IRC | 08:03 | |
| *** junbo has joined #openstack-kolla | 08:06 | |
| *** StephenWang1991 has quit IRC | 08:09 | |
| *** salv-orl_ has joined #openstack-kolla | 08:13 | |
| *** salv-orlando has quit IRC | 08:17 | |
| openstackgerrit | Bertrand Lallau proposed openstack/kolla-ansible master: Neutron: nova_metadata_ip property is deprecated https://review.openstack.org/478446 | 08:18 |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla-ansible master: Fix prechecks fail with docker not in deployment host https://review.openstack.org/478796 | 08:20 |
| *** ddyer has quit IRC | 08:23 | |
| *** StephenWang1991 has joined #openstack-kolla | 08:24 | |
| *** salv-orl_ has quit IRC | 08:25 | |
| *** ddyer has joined #openstack-kolla | 08:26 | |
| *** salv-orlando has joined #openstack-kolla | 08:28 | |
| *** jascott1 has quit IRC | 08:31 | |
| *** jascott1 has joined #openstack-kolla | 08:32 | |
| *** StephenW_ has joined #openstack-kolla | 08:35 | |
| *** jascott1 has quit IRC | 08:36 | |
| *** Mannheim has joined #openstack-kolla | 08:37 | |
| *** StephenWang1991 has quit IRC | 08:39 | |
| *** zhubingbing has joined #openstack-kolla | 08:48 | |
| openstackgerrit | Merged openstack/kolla master: Add gnocchi-statsd support for Debian systems https://review.openstack.org/477510 | 08:53 |
| *** athomas has joined #openstack-kolla | 08:55 | |
| *** tvignaud has quit IRC | 08:56 | |
| *** karimb has joined #openstack-kolla | 08:57 | |
| *** mgoddard has joined #openstack-kolla | 09:03 | |
| *** rmart04 has joined #openstack-kolla | 09:09 | |
| *** tvignaud has joined #openstack-kolla | 09:09 | |
| masber | hi, anyone knows why magnum decides to put the master on a private network which is different from the one setup on openstack? | 09:12 |
| masber | my magnum cluster fails to deploy and I just realised that the private network is different | 09:13 |
| *** caowei has quit IRC | 09:16 | |
| *** sambetts|afk is now known as sambetts | 09:17 | |
| *** serlex has quit IRC | 09:26 | |
| *** dixiaoli has joined #openstack-kolla | 09:34 | |
| reidrac | so my dev machine has a Kaby Lake CPU; disabled HT to see if my crashes go away; cross fingers | 09:39 |
| *** MarginHu has quit IRC | 09:55 | |
| *** MarginHu has joined #openstack-kolla | 09:55 | |
| *** MarginHu has quit IRC | 09:56 | |
| *** salv-orlando has quit IRC | 10:00 | |
| *** dixiaoli has quit IRC | 10:01 | |
| *** tovin07_ has quit IRC | 10:05 | |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla-ansible master: Split keepalived own role https://review.openstack.org/452177 | 10:21 |
| *** serlex has joined #openstack-kolla | 10:21 | |
| *** yangyapeng has quit IRC | 10:24 | |
| *** manheim has quit IRC | 10:28 | |
| *** karimb has quit IRC | 10:29 | |
| *** StephenW_ has quit IRC | 10:32 | |
| *** jascott1 has joined #openstack-kolla | 10:33 | |
| *** pbourke has quit IRC | 10:37 | |
| *** jascott1 has quit IRC | 10:38 | |
| *** pbourke has joined #openstack-kolla | 10:39 | |
| *** shardy has quit IRC | 10:49 | |
| *** karimb has joined #openstack-kolla | 10:51 | |
| reidrac | nah, it didn't help; still getting random kernel crashes | 10:58 |
| *** ducttape_ has joined #openstack-kolla | 11:03 | |
| *** manheim has joined #openstack-kolla | 11:07 | |
| *** ducttape_ has quit IRC | 11:08 | |
| *** skramaja has quit IRC | 11:27 | |
| *** k_mouza has joined #openstack-kolla | 11:27 | |
| k_mouza | hello guys | 11:30 |
| *** yangyapeng has joined #openstack-kolla | 11:32 | |
| k_mouza | I'm trying to installed Magnum via kolla, in a pre-installed environment. I've enabled it in the globals.conf file, but when I do another deploy, it won't provision the magnum containers in the node that I've specified in the inventory file | 11:32 |
| k_mouza | and then a reconfigure gives me and error when it tries to check if the magnum containers are up and running | 11:33 |
| *** karimb has quit IRC | 11:33 | |
| k_mouza | is there anything else that I need to do to make magnum work ? | 11:33 |
| k_mouza | using kolla stable/newton release | 11:33 |
| *** egonzalez has quit IRC | 11:33 | |
| *** huzhengchuan_ has quit IRC | 11:34 | |
| *** huzhengchuan has quit IRC | 11:34 | |
| gfidente | hi guys, I don't think any of the tests is testing https://review.openstack.org/#/c/478649/ but it's a change to match for rhel what happens already for centos | 11:43 |
| *** rwallner has joined #openstack-kolla | 11:43 | |
| *** kornicameister has quit IRC | 11:43 | |
| *** rwallner has quit IRC | 11:45 | |
| *** rwallner has joined #openstack-kolla | 11:45 | |
| *** huzhengchuan has joined #openstack-kolla | 11:47 | |
| *** Bico_Fino has joined #openstack-kolla | 11:47 | |
| *** MarginHu has joined #openstack-kolla | 11:48 | |
| *** egonzalez has joined #openstack-kolla | 11:50 | |
| *** huzhengchuan_ has joined #openstack-kolla | 11:50 | |
| *** manheim has quit IRC | 11:51 | |
| *** Mannheim has quit IRC | 11:51 | |
| *** manheim has joined #openstack-kolla | 11:51 | |
| *** Bico_Fino has quit IRC | 11:52 | |
| *** kornicameister has joined #openstack-kolla | 11:53 | |
| *** kornicameister has quit IRC | 11:54 | |
| *** ansmith has quit IRC | 11:56 | |
| *** manheim has quit IRC | 11:56 | |
| *** rhallisey has joined #openstack-kolla | 11:57 | |
| *** prateek has quit IRC | 12:08 | |
| *** dixiaoli has joined #openstack-kolla | 12:09 | |
| hrw | gfidente: nice find | 12:11 |
| gfidente | hi hrw the other day I was looking into the images creation part | 12:11 |
| gfidente | because in tripleo we use the ceph/daemon images for the daemons | 12:11 |
| gfidente | not the official kolla images | 12:12 |
| gfidente | but client part still needs to get into the openstack images | 12:12 |
| gfidente | and found it | 12:12 |
| *** StephenWang1991 has joined #openstack-kolla | 12:16 | |
| *** schwicht has joined #openstack-kolla | 12:23 | |
| *** hrw has quit IRC | 12:29 | |
| *** hrw has joined #openstack-kolla | 12:31 | |
| *** schwicht has quit IRC | 12:31 | |
| *** StephenWang1991 has quit IRC | 12:34 | |
| *** janki has joined #openstack-kolla | 12:34 | |
| *** schwicht has joined #openstack-kolla | 12:36 | |
| *** egonzalez has quit IRC | 12:38 | |
| *** dmellado has quit IRC | 12:39 | |
| *** karimb has joined #openstack-kolla | 12:40 | |
| *** cuongnv has quit IRC | 12:41 | |
| *** dmellado has joined #openstack-kolla | 12:42 | |
| *** masber has quit IRC | 12:48 | |
| *** egonzalez has joined #openstack-kolla | 12:51 | |
| *** trozet has joined #openstack-kolla | 12:53 | |
| *** shardy has joined #openstack-kolla | 12:54 | |
| *** prateek has joined #openstack-kolla | 12:57 | |
| *** schwicht has quit IRC | 12:57 | |
| *** lucasxu has joined #openstack-kolla | 13:01 | |
| *** ducttape_ has joined #openstack-kolla | 13:07 | |
| rwellum | ping kfox1111 or sdake - have a question about k8's - using my own source image. | 13:08 |
| openstackgerrit | Merged openstack/kolla-ansible master: Add possibility to configure tenant network types and type drivers https://review.openstack.org/469511 | 13:11 |
| *** ansmith has joined #openstack-kolla | 13:12 | |
| *** ducttape_ has quit IRC | 13:12 | |
| openstackgerrit | Andy Smith proposed openstack/kolla-ansible master: Add support for hybrid messaging backends https://review.openstack.org/468966 | 13:21 |
| *** srwilkers has joined #openstack-kolla | 13:22 | |
| *** salv-orlando has joined #openstack-kolla | 13:24 | |
| *** salv-orlando has quit IRC | 13:25 | |
| *** rmart04_ has joined #openstack-kolla | 13:25 | |
| *** salv-orlando has joined #openstack-kolla | 13:25 | |
| *** rmart04 has quit IRC | 13:26 | |
| *** rmart04_ is now known as rmart04 | 13:26 | |
| *** goldyfruit has joined #openstack-kolla | 13:28 | |
| *** schwicht has joined #openstack-kolla | 13:30 | |
| *** ducttape_ has joined #openstack-kolla | 13:31 | |
| *** ducttap__ has joined #openstack-kolla | 13:32 | |
| *** ducttap__ has quit IRC | 13:33 | |
| *** karimb has quit IRC | 13:33 | |
| *** ducttap__ has joined #openstack-kolla | 13:33 | |
| *** eaguilar has joined #openstack-kolla | 13:34 | |
| *** jtriley has joined #openstack-kolla | 13:34 | |
| *** ducttape_ has quit IRC | 13:36 | |
| *** karimb has joined #openstack-kolla | 13:37 | |
| bmaltais | I discovered something interesting regarding token in kolla-ansible | 13:38 |
| openstackgerrit | Paul Bourke (pbourke) proposed openstack/kolla master: rabbitmq-plugins enable no longer needs /bin/true https://review.openstack.org/478935 | 13:38 |
| bmaltais | Might be the same for kolla-kubernetes. Essentially token are piling up in the mariadb keystone token table | 13:38 |
| *** rmart04 has quit IRC | 13:39 | |
| bmaltais | Expired tokens are not deleted and endup taking all the DB space after days of running. I use ceilometer and ManageIQ and I think the frequent polling is filling up the token table quickly. | 13:39 |
| bmaltais | After 7 days there was more than 1 million expired token in the DB taking more that 20GB of storage | 13:40 |
| *** StephenWang1991 has joined #openstack-kolla | 13:40 | |
| bmaltais | I had to use a mysql query to delete all expired token to get back to normal | 13:40 |
| bmaltais | Anyone else experiencing this? | 13:41 |
| *** rmart04 has joined #openstack-kolla | 13:41 | |
| bmaltais | Should there be a token cleanup cron job added to kolla-ansible to clean this up automatically? | 13:41 |
| k_mouza | jemcevoy: Hello! I have an issue and I was told that you've faced a similar problem and that you've solved it! I basically want to have multiple pci_alias inputs in my controller node's nova.conf file. I've added those multiple inputs in my custom config files, but kolla-ansible reconfigure, overwrites the first entries and only the last one makes it to the final nova.conf file. | 13:47 |
| k_mouza | jemcevoy: I understand that this is the default behaviour of a python directory structure, but have you found a way around it? Thanks a lot! | 13:48 |
| *** zhurong has quit IRC | 13:49 | |
| *** huzhengchuan_ has quit IRC | 13:50 | |
| *** huzhengchuan has quit IRC | 13:50 | |
| *** dmsimard is now known as dmsimard|afk | 13:50 | |
| *** rmart04 has quit IRC | 13:51 | |
| *** emccormick has joined #openstack-kolla | 13:52 | |
| *** eaguilar has quit IRC | 13:53 | |
| *** eaguilar has joined #openstack-kolla | 13:53 | |
| openstackgerrit | Sam Betts proposed openstack/kolla-kubernetes master: Make wait_for_pods.py work inside kubernetes cluster https://review.openstack.org/478941 | 13:56 |
| *** ducttap__ has quit IRC | 13:56 | |
| *** ducttape_ has joined #openstack-kolla | 13:57 | |
| *** rmart04 has joined #openstack-kolla | 13:59 | |
| *** mandre is now known as mandre|away | 14:05 | |
| *** ducttape_ has quit IRC | 14:05 | |
| *** ducttape_ has joined #openstack-kolla | 14:05 | |
| *** eaguilar has quit IRC | 14:17 | |
| *** eaguilar has joined #openstack-kolla | 14:22 | |
| *** ducttape_ has quit IRC | 14:23 | |
| *** huzhengchuan has joined #openstack-kolla | 14:26 | |
| *** huzhengchuan_ has joined #openstack-kolla | 14:26 | |
| *** yangyape_ has joined #openstack-kolla | 14:28 | |
| *** rmart04 has quit IRC | 14:29 | |
| *** yangyapeng has quit IRC | 14:29 | |
| *** zhurong has joined #openstack-kolla | 14:30 | |
| *** omenv has joined #openstack-kolla | 14:30 | |
| *** ducttape_ has joined #openstack-kolla | 14:30 | |
| *** ducttape_ has quit IRC | 14:39 | |
| *** zhurong has quit IRC | 14:40 | |
| *** ducttape_ has joined #openstack-kolla | 14:41 | |
| *** eaguilar has quit IRC | 14:45 | |
| *** ducttape_ has quit IRC | 14:47 | |
| *** StephenWang1991 has quit IRC | 14:55 | |
| *** lpetrut has quit IRC | 14:56 | |
| *** renmak_ has joined #openstack-kolla | 15:02 | |
| *** bmaltais has quit IRC | 15:05 | |
| *** StephenWang1991 has joined #openstack-kolla | 15:05 | |
| *** mmehan has joined #openstack-kolla | 15:06 | |
| *** kbyrne has quit IRC | 15:06 | |
| *** ducttape_ has joined #openstack-kolla | 15:07 | |
| *** kbyrne has joined #openstack-kolla | 15:07 | |
| *** jamesbenson has joined #openstack-kolla | 15:08 | |
| openstackgerrit | Bertrand Lallau proposed openstack/kolla-ansible master: Cinder: modernize nova config https://review.openstack.org/478964 | 15:08 |
| *** jamesbenson_ has joined #openstack-kolla | 15:11 | |
| *** jamesbenson has quit IRC | 15:12 | |
| sdake | rwellum shoot | 15:13 |
| *** blallau has joined #openstack-kolla | 15:16 | |
| kfox1111 | morning. | 15:19 |
| *** jistr is now known as jistr|afk | 15:20 | |
| sdake | sup kfox1111 | 15:22 |
| openstackgerrit | Bertrand Lallau proposed openstack/kolla-ansible master: Cinder: modernize nova config https://review.openstack.org/478964 | 15:23 |
| jemcevoy | k_mouza: I did not need more than one pci_alias per host. The configuration at that customer uses two kinds of nVidia GRID cards the older servers have K2 cards and the newer have M60 cards. What I ended up doing is putting the whitelist and pci_alias for the M60's in the default nova.conf and made host specific nova.conf for the servers with K2's with different pci_alias and whitelist. | 15:26 |
| jemcevoy | I have no idea why this works because the nova_scheduler on the controller nodes only have the alias for the the M60 Cards. | 15:28 |
| egonzalez | jemcevoy, maybe is easier to maintain using a var in /etc/kolla/config/nova.conf <option> ={{ pci_alias }} and include that var in the inventory per host | 15:28 |
| rwellum | sdake: If I want to build my own service from my own source code - say keystone, and run with standard everything else, then I need to make sure the kolla image tag is the same release as my source code correct? So if my source code is Mitaka based then I am pretty much screwed kolla image wise? | 15:30 |
| jemcevoy | egonzalez: hmmm... Could that var include a \n or CRTL-J to create a two line entry because pci_alias is a (Mulit-Opt) that can exist multiple times in the nova.conf file. | 15:30 |
| sdake | if the source code for your service is mitaka- why not run the 3.0 images? | 15:31 |
| sdake | then you will be running mitaka of everything else | 15:31 |
| rwellum | sdake: having some issues relating the image tag to the openstack release. | 15:32 |
| egonzalez | jemcevoy, not sure how multiopts are evaluated, but a jinja `for` may work | 15:32 |
| jemcevoy | Then a jinja2/yaml question... How would a multi-line var be defined? Just open quote to a close quote? And how about indenting? Or is there another trick to create a newline in the var? | 15:36 |
| *** jascott1 has joined #openstack-kolla | 15:39 | |
| *** blallau has quit IRC | 15:40 | |
| *** jascott1 has quit IRC | 15:43 | |
| *** eaguilar has joined #openstack-kolla | 15:44 | |
| *** StephenWang1991 has quit IRC | 15:46 | |
| *** trozet has quit IRC | 15:48 | |
| k_mouza | jemcevoy: I have the aliases and the whitelists on each different node as well (K80, M40 and K40) and I've created the flavors, but when I try to use one of the ones that's not in the controller's nova.conf file I get a "pci alias not defined" error | 15:54 |
| openstackgerrit | Helena proposed openstack/kolla-ansible master: Enabled additional .conf files to be read by collectd https://review.openstack.org/477580 | 15:54 |
| openstackgerrit | Helena proposed openstack/kolla master: Enabled the collectd-ceilometer-plugin https://review.openstack.org/477581 | 15:55 |
| k_mouza | I also tried a dictionary syntax for the pci_alias, but that didn't work either (systax error) | 15:55 |
| openstackgerrit | Sam Betts proposed openstack/kolla-kubernetes master: Make wait_for_pods.py work inside kubernetes cluster https://review.openstack.org/478941 | 15:55 |
| *** sambetts is now known as sambetts|afk | 16:00 | |
| *** egonzalez has quit IRC | 16:04 | |
| *** jistr|afk is now known as jistr | 16:07 | |
| jemcevoy | k_mouza: The customer with the mixed GRID cards is in pre-production testing and I cannot test egonzalez jinja2 idea using for to iterate over a list of pci_aliases. | 16:07 |
| *** prateek_ has joined #openstack-kolla | 16:08 | |
| *** prateek has quit IRC | 16:09 | |
| jemcevoy | k_mouza: Your other option would be to manually edit the configs on all the controllers and add the additional pci_aliases. Then restart the nova_schedulers. Also do you see all the devices in the nova.pci_devices table of mariadb? | 16:11 |
| *** manheim has joined #openstack-kolla | 16:13 | |
| *** salv-orl_ has joined #openstack-kolla | 16:14 | |
| *** eaguilar has quit IRC | 16:16 | |
| *** salv-orlando has quit IRC | 16:17 | |
| jemcevoy | k_mouza: Thanks for bringing this up. Actually I think that I may have done the manual edit to /etc/kolla/nova-scheduler/nova.conf and a subsequent deploy wiped the change... I better fix that before a reboot breaks the K2's... | 16:17 |
| k_mouza | jemcevoy: yup, all the GPUs are shown in the pci_device table | 16:17 |
| *** manheim has quit IRC | 16:17 | |
| k_mouza | jemcevoy: I'll have a look at egonzalez jinja2 idea and I'll report back. Let me know if you find a syntax that works as well. | 16:18 |
| *** dixiaoli has quit IRC | 16:22 | |
| *** MarginHu has quit IRC | 16:24 | |
| openstackgerrit | Mark Goddard proposed openstack/kolla-ansible master: Allow empty compute group when ironic is in use https://review.openstack.org/478993 | 16:28 |
| *** karimb has quit IRC | 16:29 | |
| *** reidrac has quit IRC | 16:34 | |
| *** unicell has quit IRC | 16:36 | |
| *** prateek has joined #openstack-kolla | 16:38 | |
| *** goldyfruit has quit IRC | 16:39 | |
| *** prateek_ has quit IRC | 16:40 | |
| openstackgerrit | Mark Goddard proposed openstack/kolla master: Add e2fsprogs and xfsprogs to ironic-conductor https://review.openstack.org/478998 | 16:44 |
| inc0 | good morning | 16:45 |
| *** k_mouza has quit IRC | 16:45 | |
| *** pcaruana has quit IRC | 16:48 | |
| inc0 | jemcevoy: I have idea for pci for you | 16:49 |
| inc0 | I'll need to make change to kolla-ansible tho | 16:49 |
| jemcevoy | Sounds Good and we should include k_mouza... He has the same issue | 16:53 |
| jemcevoy | inc0 did you see that I finally got around to filing the osd bug late yesterday... https://bugs.launchpad.net/kolla-ansible/+bug/1701148 it has how to repro but I have not tested the repro method... | 16:55 |
| openstack | Launchpad bug 1701148 in kolla-ansible "Ceph OSD Containers Fail when Device Names Change" [Undecided,New] | 16:55 |
| inc0 | yeah I saw | 16:55 |
| inc0 | didn't get to it yet | 16:55 |
| *** leseb has joined #openstack-kolla | 16:57 | |
| *** jascott1 has joined #openstack-kolla | 17:00 | |
| *** mgoddard has quit IRC | 17:02 | |
| inc0 | jemcevoy: can you show me conf you want to land in end nova.conf? | 17:02 |
| inc0 | as in, how do you want it to be rendered? | 17:02 |
| jemcevoy | Yes, But in a call at he moment | 17:05 |
| *** eaguilar has joined #openstack-kolla | 17:06 | |
| *** unicell has joined #openstack-kolla | 17:12 | |
| jamesbenson_ | Some commands from JJB are failing, I'm not sure how to follow it. Any suggestions? | 17:12 |
| jamesbenson_ | ^^ These are in the gates. | 17:15 |
| *** jascott1 has quit IRC | 17:16 | |
| *** jascott1 has joined #openstack-kolla | 17:17 | |
| *** jascott1 has quit IRC | 17:17 | |
| *** jascott1 has joined #openstack-kolla | 17:17 | |
| *** rhallisey is now known as rhallisey-lunch | 17:26 | |
| *** rhallisey-lunch has quit IRC | 17:28 | |
| *** mgoddard has joined #openstack-kolla | 17:41 | |
| jamesbenson_ | horizon/sqlite3 is yelling at me that there is no django_session table? bad build? http://paste.openstack.org/show/614107/ | 17:42 |
| *** prateek has quit IRC | 17:50 | |
| *** tonanhngo has joined #openstack-kolla | 17:55 | |
| *** tonanhngo has quit IRC | 17:56 | |
| *** tonanhngo has joined #openstack-kolla | 17:56 | |
| jamesbenson_ | but funny thing, my dev that deployed correctly, horizon is giving the classics "Something went wrong". In the logs it states, OperationalError: no such table: django_session | 17:57 |
| *** salv-orl_ has quit IRC | 18:00 | |
| *** salv-orlando has joined #openstack-kolla | 18:00 | |
| *** ansmith has quit IRC | 18:01 | |
| *** mgoddard has quit IRC | 18:02 | |
| *** salv-orlando has quit IRC | 18:05 | |
| *** eaguilar has quit IRC | 18:06 | |
| *** serlex has quit IRC | 18:07 | |
| *** dmsimard|afk is now known as dmsimard | 18:13 | |
| *** salv-orlando has joined #openstack-kolla | 18:14 | |
| *** ansmith has joined #openstack-kolla | 18:16 | |
| *** lpetrut has joined #openstack-kolla | 18:19 | |
| *** gfidente is now known as gfidente|afk | 18:21 | |
| *** goldyfruit has joined #openstack-kolla | 18:29 | |
| *** rhallisey has joined #openstack-kolla | 18:30 | |
| *** janki has quit IRC | 18:30 | |
| *** bmace has quit IRC | 18:36 | |
| *** bmace has joined #openstack-kolla | 18:36 | |
| *** salv-orlando has quit IRC | 18:47 | |
| *** salv-orlando has joined #openstack-kolla | 18:48 | |
| *** sbezverk has joined #openstack-kolla | 18:52 | |
| *** salv-orlando has quit IRC | 18:53 | |
| SamYaple | jemcevoy: yea thats a "known" thing. youre supposed to rerun the playbooks on reboot to bring the osds back up proper | 18:53 |
| SamYaple | jemcevoy: but i thought it was switched to partuuid at some point | 18:53 |
| *** ansmith has quit IRC | 19:05 | |
| *** sbezverk has quit IRC | 19:16 | |
| *** ansmith has joined #openstack-kolla | 19:19 | |
| *** rhallisey has quit IRC | 19:20 | |
| *** eaguilar has joined #openstack-kolla | 19:20 | |
| *** ducttap__ has joined #openstack-kolla | 19:22 | |
| *** ducttape_ has quit IRC | 19:22 | |
| *** rhallisey has joined #openstack-kolla | 19:23 | |
| *** sbezverk has joined #openstack-kolla | 19:23 | |
| jemcevoy | SamYaple: The data partition uses the uuid but the co-resident journal used the /dev/sdh2 like names... | 19:31 |
| *** salv-orlando has joined #openstack-kolla | 19:34 | |
| jemcevoy | inc0: Is there a bug on the multiOpt nova.conf issue for pci_alias? | 19:39 |
| jemcevoy | SamYaple: Should I try rerunning a deploy or reconfigure? | 19:42 |
| *** salv-orlando has quit IRC | 19:44 | |
| *** salv-orlando has joined #openstack-kolla | 19:44 | |
| jemcevoy | inc0: I posted the nova.conf here: http://paste.openstack.org/show/614116/ There two entries for the pci which can have multiple entries in the rendered nova.conf | 19:52 |
| *** sbezverk has quit IRC | 19:57 | |
| SamYaple | jemcevoy: remove the ceph-osd containers and rerun the playbooks, that will fix it | 20:05 |
| SamYaple | jemcevoy: you may not even need to remove the ceph-osd containres first | 20:05 |
| jemcevoy | SamYaple: So run kolla-ansible deploy? | 20:06 |
| SamYaple | jemcevoy: yep | 20:06 |
| jemcevoy | And should the osd containers be looping in restart or stopped? | 20:07 |
| SamYaple | jemcevoy: the osd containers should be running | 20:08 |
| jemcevoy | Here goes deploy... Let you know when the deploy completes | 20:09 |
| *** sbezverk has joined #openstack-kolla | 20:15 | |
| *** Pavo has joined #openstack-kolla | 20:22 | |
| *** Pavo has quit IRC | 20:24 | |
| *** lpetrut has quit IRC | 20:25 | |
| emccormick | Does anyone have any advice / links to share on doing a brownfield deploy using Kolla? I've got a deployment previously done without containers that I'd like to start using Kolla on. Most things I can make sense of rotating hosts in and out as they are deployed to containers, but I'm kind of stuck on how to do the computes | 20:31 |
| *** Pavo has joined #openstack-kolla | 20:31 | |
| emccormick | Would I need to make new computes in containers and then migrate? | 20:32 |
| *** ansmith has quit IRC | 20:32 | |
| *** karimb has joined #openstack-kolla | 20:34 | |
| *** zhubingbing has quit IRC | 20:35 | |
| *** zhubingbing has joined #openstack-kolla | 20:36 | |
| inc0 | emccormick: we had discussion on ptg | 20:38 |
| inc0 | live migration would be safest bet, you can experiment with killing libvirt and starting it in container, you might need to manually copy all required xmls and stuff | 20:39 |
| inc0 | neutron should just work if you kill agents on compute and start it in containers | 20:40 |
| inc0 | if you would keep libvirt running on host and just deploy nova-compute in container that should work too, however might be tricky configuration | 20:40 |
| inc0 | jemcevoy: yeah...openstack and it's broken ini | 20:42 |
| inc0 | it's hard problem to solve with broken ini:/ | 20:42 |
| inc0 | sbezverk kfox1111 around? | 20:43 |
| inc0 | sdake: | 20:44 |
| sdake | sup | 20:44 |
| inc0 | wanna have orch conversation? | 20:44 |
| *** lucasxu has quit IRC | 20:44 | |
| sdake | if sbezverk and kfox1111 around | 20:44 |
| *** sbezverk has quit IRC | 20:44 | |
| *** sbezverk has joined #openstack-kolla | 20:45 | |
| sbezverk | sdake: I am around but I have crappy connection on the train, so I might be delayed with replies | 20:46 |
| sdake | snakes on a train | 20:46 |
| sdake | how about kfox1111 | 20:46 |
| sbezverk | sdake: you wish ;) | 20:47 |
| inc0 | I'm not sure if kfox cares that much;) he won't use it that much anyway right? | 20:49 |
| inc0 | what I'm trying to do is to agree on signle direction and roll with it, we can't affort too much bikeshedding because we'll never do 1.0 | 20:49 |
| sbezverk | inc0: actually he does care :) and we wanted to use this opportunity to discuss the approach we take with different orchestrations | 20:50 |
| inc0 | yeah, ofc, I don't want to lock him out, I didn't schedule time for it so it's on me:( | 20:50 |
| inc0 | do we want to use ML to figure it out and put a deadline for final decision? | 20:51 |
| sbezverk | inc0: I chatted with kfox1111 yesterday and he should be around | 20:51 |
| inc0 | ok, let's just wait then | 20:52 |
| sbezverk | but as you proposed ML could br a good alternative | 20:52 |
| kfox1111 | hi. | 20:52 |
| inc0 | there ya go:P | 20:52 |
| inc0 | sorry kfox1111 ;) | 20:52 |
| sbezverk | kfox1111: good timing ;) | 20:52 |
| inc0 | soo, let's start | 20:53 |
| sdake | first off - need a dose of anestaphine from the medlock plz | 20:53 |
| inc0 | (also I wanted to apologize for assuming you don't care, I shouldn't assume, sorry) | 20:53 |
| inc0 | sdake: you're srsly into painkillers man;) | 20:53 |
| sdake | walk a mile in my shoes | 20:54 |
| inc0 | I'm not judging | 20:54 |
| inc0 | well, let's kick off discusion | 20:54 |
| inc0 | so we have few changes that works nicely together | 20:54 |
| inc0 | https://review.openstack.org/#/c/473588/ | 20:55 |
| inc0 | https://review.openstack.org/#/c/474770/ | 20:55 |
| inc0 | I see few good solutions for this problem | 20:55 |
| inc0 | my 2 personal favorites are: | 20:56 |
| inc0 | 1. run full playbook for configure + deploy in container | 20:56 |
| inc0 | 2. run configure in container, use entrypoint for 90% of orchestration and ansible for 10% that's problematic with entrypoint | 20:56 |
| sbezverk | gents: personally I do not have issue with which ever way kolla-k8s is deployed, I think it is fair for different type of orchestration to use the best way from its perspective. | 20:58 |
| sbezverk | the only thing I would care in this is not to polute kolla-k8s repo with these | 20:59 |
| inc0 | well, I'd like to have one that's good | 20:59 |
| kfox1111 | yeah. orchestration is hard and rather particular to sites. | 20:59 |
| kfox1111 | having it clearly seperate has a lot of beinifits I think. | 20:59 |
| inc0 | and manual as option always | 20:59 |
| inc0 | separate - totally yes | 20:59 |
| sbezverk | kfox1111: +1 | 20:59 |
| inc0 | but one we support for 90% of people | 20:59 |
| sbezverk | we should use kolla approach, when images are stored in 1 repo and ansible deployment in another | 21:00 |
| kfox1111 | +1 | 21:00 |
| inc0 | please lets not add kolla-kubernetes-ansible repo... | 21:00 |
| inc0 | let's do what kolla did for most part | 21:00 |
| inc0 | keep ansible in ansible dir | 21:00 |
| sbezverk | same should apply to charts and different orchestrations using them | 21:00 |
| kfox1111 | why so against extra repos? | 21:00 |
| inc0 | kfox1111: even now users are getting confusedf | 21:01 |
| kfox1111 | it has worked well for kolla-ansible and kolla-kubernetes. | 21:01 |
| inc0 | and tha'ts level where I'm ok with | 21:01 |
| sbezverk | inc0: this way kolla-k8s could explode | 21:01 |
| inc0 | let's keep it from exploding please;) | 21:01 |
| sbezverk | we can build right compartments from start | 21:02 |
| sdake | you never really know what ou need untill you are 80% done | 21:02 |
| inc0 | it works well for kolla-k8s and kolla-ansible because kolla for long time was single repo and people grew familiar with idea | 21:02 |
| sdake | that is the time to make a repo split if one is to be made | 21:02 |
| kfox1111 | kolla-kubernetes took off becaue it wasnt tied to kolla-ansible's repo too tightly. | 21:02 |
| kfox1111 | orchestration I think may end up in the same boat. | 21:02 |
| kfox1111 | there are conflicting interests there, so seperate repos allows them to make progress without stepping on each other. | 21:03 |
| sbezverk | I think it is rally good timing for that split | 21:03 |
| inc0 | I'd rather start with helm microcharts available and one way to orchestrate in house | 21:03 |
| sbezverk | we finishged 80-90% of charts | 21:03 |
| sbezverk | \but otchestration is just starting | 21:03 |
| sbezverk | so nothing to break by this split | 21:03 |
| kfox1111 | inc0: there is alreayd one way today in tree. service charts. | 21:03 |
| inc0 | sbezverk: and if we make it hard to handle for someone new to kolla-k89s | 21:03 |
| *** eaguilar has quit IRC | 21:03 | |
| inc0 | it's gonna hurt recruitment of reviewers | 21:04 |
| kfox1111 | you have a differing view (I don't disagree with) on orchestration. | 21:04 |
| inc0 | kfox1111: if service charts would be enough to handle all things we care about (upgrades and such) I'd be ok | 21:04 |
| inc0 | with putting ansible in different repo | 21:04 |
| sbezverk | inc0: I do not think anything prevents kolla-ansible folks to use kolla images | 21:04 |
| inc0 | frankly? I wouldn't bother to keep up with ansible at all | 21:04 |
| sbezverk | or confuse them ;) | 21:05 |
| inc0 | thing is, we had ansible in kolal for long time | 21:05 |
| inc0 | and split when second thing apperared | 21:05 |
| inc0 | that was good model imho | 21:05 |
| inc0 | and menas we have one functional way in tree until we grow second | 21:06 |
| inc0 | then we split both | 21:06 |
| kfox1111 | I think kolla might ahave had more contributoers sooner if it was split earlier. | 21:06 |
| sdake | kfox1111 i think you are wrong on that point | 21:06 |
| sdake | even now people are confused about the repository split | 21:06 |
| sbezverk | inc0: here is the thing if we focus on ansible then gates should be testing ansible more, but then we close door for other people to consume kolla-k8s | 21:06 |
| inc0 | and I agree with sdake. Cleanliness of split < easiness to start with project | 21:06 |
| kfox1111 | its hard to do later. its easy to do now. | 21:07 |
| sbezverk | as we would test more orchestration specific things | 21:07 |
| kfox1111 | it also make it more likely others will start using it. | 21:07 |
| inc0 | sbezverk: let's wait for these people to come first | 21:07 |
| inc0 | then we'll discuss | 21:07 |
| sdake | kfox1111 its easy to do the wrong thing now | 21:07 |
| kfox1111 | we need to be encuring others to share as much as possible. | 21:07 |
| sdake | its more costly but easier to do the correct thing later | 21:07 |
| kfox1111 | another example.... | 21:07 |
| kfox1111 | kolla-kubernetes microcharts are pretty stable at this point. | 21:07 |
| inc0 | I think for one that we should limit number of steps and configurations to get kolla-k8s up and running | 21:08 |
| kfox1111 | but we can't call that part 1.0 due to no orchestration | 21:08 |
| sbezverk | sdake: so you are saying it is easier to build house or wrong foundation ;) ??? | 21:08 |
| inc0 | and splitting repos will do exactly oposite | 21:08 |
| kfox1111 | which means people are less likely to adopt that level of components even though it is probably ready for wider use. | 21:08 |
| inc0 | kfox1111: 1.0 *has to have* orchestration | 21:08 |
| sdake | well i guess that is a third option | 21:08 |
| inc0 | you can handle migration manually | 21:08 |
| kfox1111 | orchestation should be seperate and versioned 0.1.0. microservices, 1.0.0 | 21:08 |
| sdake | rename kolla-kubernetes to kolla-microcharts and call it 1.0 | 21:08 |
| inc0 | 90% of people can't | 21:08 |
| *** jtriley has quit IRC | 21:09 | |
| kfox1111 | i'd be ok going that other route I guess. | 21:09 |
| sdake | and say that building block is done and move on to orchestration | 21:09 |
| sbezverk | inc0: helm charts now are tested but a common appraoch, which means regardless who consumes it can be sure no orchestration short cuts are taken | 21:09 |
| kfox1111 | kolla-kubernetes-core, kolla-kubernetes-common, kolla-kubernetes-microservice, something. | 21:09 |
| inc0 | and I'd say...can we just move to orchestration and keep it in tree for now? I really don't want more chaos today with repo split before we even started | 21:09 |
| kfox1111 | gaain, its easy today. | 21:10 |
| kfox1111 | do the git rename, and add kolla-kubernetes. | 21:10 |
| kfox1111 | its harder later to split a repo. | 21:10 |
| inc0 | it's easy to make, yes | 21:10 |
| inc0 | it's harder to set up openstack with it | 21:10 |
| sdake | harder to build a community aroudn it | 21:10 |
| inc0 | that's what I want to avoid | 21:10 |
| kfox1111 | why? | 21:10 |
| sdake | its a startup project in essence | 21:10 |
| sbezverk | inc0: if people start using right appraoch form the start they do not need to change their workflows | 21:11 |
| kfox1111 | some projects literally have hundreds of repo's. why is it hard to have 1 more? I really don't get it. | 21:11 |
| sbezverk | it will become natural right from the start | 21:11 |
| inc0 | kfox1111: and one repo to rule them all | 21:11 |
| kfox1111 | yeah. whats the problem? | 21:11 |
| inc0 | that's how fuel, osa and others roll - one repo and 100s of sub-repos | 21:11 |
| inc0 | we might as well have single repo... | 21:11 |
| kfox1111 | not sure we're going to see eye to eye on this one. | 21:12 |
| inc0 | having multiple repos with git submodule gives us little of gain | 21:12 |
| kfox1111 | I really think one repo's a bad idea, for the same reasons I think one kolla repo was a bad idea. | 21:12 |
| kfox1111 | I don't want a million little repos either though. | 21:12 |
| inc0 | well, what I'm saying, can we shelve multi-repo discussion and talk how to make deployment of kolla-k8s easy | 21:12 |
| inc0 | ? | 21:12 |
| kfox1111 | a happy middle ground. | 21:12 |
| sdake | i think repo split is jumping the gun - we dont even have any code to split | 21:13 |
| inc0 | exactly | 21:13 |
| sbezverk | sdake: that is the whole poiunt | 21:13 |
| sdake | ok - so how does one go about building that code? | 21:13 |
| sbezverk | when code is ready it goes where it makes sense | 21:13 |
| kfox1111 | we've shelved it for a year now. how much longer do hou want to ask those that feel more then one repo are the right way to hold their toungs? | 21:13 |
| sdake | from scratch, with no community, completely from nothing | 21:13 |
| inc0 | kfox1111: we need an orchestration layer | 21:13 |
| inc0 | if we split code prior to 1.0 fine | 21:13 |
| kfox1111 | your asking us to defacto agree with you for some undeterminate amount of time. | 21:13 |
| sbezverk | and not where it goes until something forces to go in the different repo | 21:14 |
| inc0 | but can we build an orcherstration layer *before* we split? | 21:14 |
| Pavo | evening | 21:14 |
| inc0 | what if ansible turns out to be bad idea and we'll want to move to puppet for example? | 21:14 |
| inc0 | we bury one repo and open another? | 21:14 |
| inc0 | kolla was single repo before we were 100% confident our stuff is worth while | 21:14 |
| sbezverk | inc0: no we add a new repo | 21:14 |
| inc0 | and we can say *after* it's done | 21:15 |
| sbezverk | and keep ansible repo for ansible funs ;) | 21:15 |
| kfox1111 | probably easier to burry an ansible repo then with it mixed in with all the microservices. | 21:15 |
| inc0 | sbezverk: then we'll have 5 unmiaintained repos and one working....also know as "nightmare for newcommers" | 21:15 |
| kfox1111 | we wouldn't be able to get rid of the history if we didn't want it. | 21:15 |
| kfox1111 | the whole kolla-mesos example. | 21:15 |
| kfox1111 | it was easy to get rid of when it failed. | 21:15 |
| inc0 | hey Pavo btw;) | 21:16 |
| inc0 | because we had mainline kolla | 21:16 |
| Pavo | so...... noticed some major changes in kolla since last time I deployed | 21:16 |
| kfox1111 | if it was in kolla, then it would be in the git repo, forever. | 21:16 |
| inc0 | Pavo: well....possibly;) | 21:16 |
| sdake | kfox1111 i miss the concern - other then git rm takes some time | 21:16 |
| inc0 | what I don't want is to create galaxy of kolla-k8s orch engines with single star in it | 21:17 |
| Pavo | no more kolla-build from pip install, and docs say there should be a startup-registry script in /usr/share/kolla-ansible/tools and there isn't one | 21:17 |
| kfox1111 | another concern is users having to track versions due to orchestration changes reguardless of microservice changes. | 21:17 |
| inc0 | Pavo: kolla-build is in kolla, so is start registry | 21:17 |
| inc0 | pip install kolla | 21:17 |
| sdake | kfox1111 see pavo above - he is confused by the repo split :) | 21:17 |
| inc0 | and pip install kolla-ansible | 21:18 |
| Pavo | so... its pip install kolla and not pip install kolla-ansible now? | 21:18 |
| inc0 | yes | 21:18 |
| inc0 | since ocata | 21:18 |
| Pavo | ah | 21:18 |
| kfox1111 | sdake: I'm confused by the not split. | 21:18 |
| Pavo | now I understand | 21:18 |
| kfox1111 | kolla = containers. kolla-ansible = tooling. | 21:18 |
| Pavo | huh | 21:18 |
| Pavo | now that confused me | 21:19 |
| openstackgerrit | Bertrand Lallau proposed openstack/kolla-ansible master: Enable cluster trust for Magnum service https://review.openstack.org/472658 | 21:19 |
| kfox1111 | kolla really should be renamed kolla-contiainers to avoid some confusion. | 21:19 |
| inc0 | haha | 21:19 |
| inc0 | Pavo: kolla is about building images and all that comes with it | 21:19 |
| inc0 | kolla-ansible is to deploy them | 21:19 |
| kfox1111 | yeah. and kolla-kubernets is an alternate deployment mechanism to kolla-ansible. | 21:19 |
| sbezverk | inc0: kolla-k8s is about charts and what ever in orchestration repos to deploy them ;) make sense to me | 21:20 |
| jascott1 | so kolla-ansible doesnt orchestrate? | 21:20 |
| Pavo | yeah I knew that but docs say pip install kolla-ansible | 21:20 |
| Pavo | not pip install kolla | 21:20 |
| kfox1111 | the issue is, kolla-contaienrs is common between two different deployment tools. | 21:20 |
| Pavo | extremely confused now | 21:20 |
| inc0 | technically ansible playbook for deploying kolla-k8s could land in kolla-ansible | 21:20 |
| inc0 | that would be cool confusion to add | 21:20 |
| kfox1111 | kolla-kubernetes microservices are common in the same way to orchestration. | 21:21 |
| inc0 | now you need all 3! | 21:21 |
| Pavo | wait what! | 21:21 |
| inc0 | Pavo: we're having 2 converstation in similar topic | 21:21 |
| kfox1111 | Pavo: sorry. we're discussing two different things at the moment. :/ | 21:21 |
| Pavo | now you have to have all 3, pip install kolla, pip install kolla-ansible and pip install kolla-kubernetes | 21:21 |
| inc0 | let me summarize one that got you confused | 21:21 |
| inc0 | Pavo: no, you don't need kolla-kubernetes | 21:21 |
| Pavo | ah ok lol was about to say | 21:22 |
| * Pavo shotgun to face | 21:22 | |
| inc0 | so, Pavo | 21:22 |
| sdake | imo best course of action is to rename kolla-kubernetes to kolla-helm-microcharts | 21:22 |
| kfox1111 | yeah. for your use case, you need kolla for the image,s and kolla-ansible to deploy. | 21:22 |
| inc0 | in ocata we pretty miuch split ansible playbook to separate repo | 21:22 |
| inc0 | so other tools like k8s could be created | 21:22 |
| sdake | and either a) write kolla-kubernetes or b) give up on kolla-kubernetes | 21:22 |
| kfox1111 | sdake: yeah. | 21:23 |
| sdake | however don't expect someone to just come along and fill out that repo for you for the A option | 21:23 |
| sbezverk | sdake: why just microcharts? would it not be better kolla-k8s-charts? | 21:23 |
| inc0 | imho we're worthless (apart of few specific use cases) unless we're easy to deploy | 21:23 |
| sdake | someone has to review the code | 21:23 |
| inc0 | and for that we need orch | 21:23 |
| Pavo | ah ok now I think I understand | 21:23 |
| sdake | sbezverk i htink we fundamnetallly all disagree on what exactly kolla-kubernetes s a repository should represent | 21:23 |
| Pavo | kolla and kolla-ansible | 21:24 |
| Pavo | to be able to build and deploy openstack ocata using kolla | 21:24 |
| sbezverk | sdake: sure let's clear it | 21:24 |
| inc0 | kolla and kolla-ansible is what you need Pavo | 21:24 |
| inc0 | exactly | 21:24 |
| inc0 | also if you want to upgrade netwon to ocata, all works | 21:24 |
| Pavo | where in the docs say how to install kolla then | 21:24 |
| openstackgerrit | Bertrand Lallau proposed openstack/kolla-ansible master: Magnum: Enable cluster trust customization https://review.openstack.org/472658 | 21:24 |
| Pavo | is it the same as kolla-ansible? | 21:24 |
| sbezverk | sdake: I think kolla-k8s should ONLY have charts and gates code | 21:24 |
| inc0 | Pavo: https://docs.openstack.org/developer/kolla/ | 21:25 |
| jamesbenson_ | Pavo: pip install kolla and pip install kolla-ansible | 21:25 |
| inc0 | Pavo: https://docs.openstack.org/developer/kolla-ansible | 21:25 |
| inc0 | 2 different links | 21:25 |
| sbezverk | sdake: in this case it is easier to protect and maintain | 21:25 |
| sdake | what kind of charts - deployment charts, micrcharts, service charts? | 21:25 |
| sbezverk | sdake: all charts | 21:25 |
| sdake | if all charts -why have a separate repo for orchestration | 21:25 |
| inc0 | yeah, I'm -2 in putting helm stuff in more than 1 repo | 21:25 |
| inc0 | also, what about configs? | 21:26 |
| inc0 | do we add configs to charts or orchestration? | 21:26 |
| kfox1111 | I kind of like the idea of having all the common stuff. the "low level" stuff in the architecture in one repo. | 21:26 |
| Pavo | inc0 I get it now but just a FYI there needs to be something on the https://docs.openstack.org/developer/kolla/ page to tells people this, because I was presuming that kolla-ansible was the same as it was prior | 21:26 |
| kfox1111 | microservices charts and configcharts. | 21:26 |
| sbezverk | sdake: because these two a completely separate things... like images and way to deploy them | 21:26 |
| kfox1111 | that lets us stablalize it and ship it without orchestration. | 21:26 |
| inc0 | Pavo: kolla page says that;) | 21:26 |
| sdake | sbezverk they are 3 things not 2 | 21:27 |
| kfox1111 | those wanting to self orchestrate can then garantee orchestration isn't creeping in. | 21:27 |
| inc0 | https://docs.openstack.org/developer/kolla/index.html this one in first paragraph | 21:27 |
| Pavo | where because I don't see that at alll | 21:27 |
| sdake | 1 is microcharts, another is deploy charts, another is ansible deployment | 21:27 |
| sdake | why 3, why 2, why 1? | 21:27 |
| sbezverk | sdake: sorry what do you mean deploy charts? Service charts? | 21:27 |
| sdake | service charts are an example of deployment yes | 21:28 |
| inc0 | service charts + entrypoint is quasi orch | 21:28 |
| sdake | or the deploy charts that run the ansible code | 21:28 |
| sbezverk | sdake: no reason t separate them, chart user can select which chart to use | 21:28 |
| inc0 | not fully functional tho and we don't know if it ever be fully functional | 21:28 |
| sbezverk | the charts are designed to sllow flexibility | 21:28 |
| kfox1111 | sbezverk: there's another reason to seperate it. it makes it lkook like there's a preffered way to do orchestration. | 21:29 |
| inc0 | ok another idea | 21:29 |
| inc0 | how about kolla-k8s main repo with helm charts | 21:29 |
| kfox1111 | I don't know if helm orchestration is the way to go. or ansible. or something else. | 21:29 |
| jascott1 | k8s comes with a scheduler but you can still bring your own | 21:29 |
| kfox1111 | I do know the microservices work great. | 21:29 |
| inc0 | and have charts-per-orch | 21:29 |
| inc0 | I want to run ansible in chart | 21:29 |
| kfox1111 | might be ok. but also might be confusing. | 21:30 |
| jascott1 | so init containers come out of existin grepo? | 21:30 |
| inc0 | if we split ansible from chart you're closing this avenue regardless | 21:30 |
| jascott1 | are removed from imean | 21:30 |
| inc0 | and what of entrypoint | 21:30 |
| inc0 | ? | 21:30 |
| *** sbezverk has quit IRC | 21:31 | |
| kfox1111 | init containers stay. they aren't nessisarily orchestration. | 21:31 |
| jascott1 | ? | 21:31 |
| kfox1111 | they can be used for it, as in the case of kube-entrypoint. | 21:31 |
| inc0 | they kinda are... | 21:31 |
| kfox1111 | but they also do stuff like tweak the config file to add the node specific stuff into it. | 21:31 |
| kfox1111 | sutff that no other entity can do. | 21:31 |
| jascott1 | but dont they interfere with 3rd party orch? | 21:31 |
| jascott1 | not those in particular but all of the init container together | 21:32 |
| kfox1111 | it shouild be minimized to not interfeer, but theres some stuff that just has to work that way. | 21:32 |
| kfox1111 | say, neutron ip addresses. | 21:32 |
| kfox1111 | it has to be in the config file, and only known when the pods starts. | 21:32 |
| kfox1111 | it can't be done outside of the pod. | 21:32 |
| *** sbezverk has joined #openstack-kolla | 21:32 | |
| *** pradk has quit IRC | 21:33 | |
| sbezverk | kfox1111: we use service charts for gates, so it is concidered as a part of gate code | 21:33 |
| jascott1 | so charts will never be 'pure' charts without enforced orchestration on some level | 21:34 |
| kfox1111 | doing stuff like tweaking generic config options in init-container is bad though. it means stuff done by orchestration might be undone by mistake. | 21:34 |
| sbezverk | kfox1111: we can clean them up to make even more generic | 21:34 |
| kfox1111 | sbezverk: those gates would be split out to the orchestration repo I think? | 21:34 |
| kfox1111 | jascott1: thats one way of viewing it. but its very very few places really. | 21:35 |
| sdake | that is correct - a econd type of orchestration | 21:35 |
| inc0 | ok, can I as you all a favor? | 21:35 |
| sdake | so now we have 2 new projects to populate instead of one | 21:35 |
| inc0 | can we shelve "split" conversation for between "have one working" and 1.0? | 21:35 |
| jascott1 | to me there is an argument there that we could have a 'charts' repo (like kolla is to images) and then N orchestration options. but sounds like its too tightly entangled for actual clean separation | 21:35 |
| kfox1111 | the microservice charts are designed as building blocks. you can have optional features in it and not use them. | 21:35 |
| inc0 | we need to make decision prior to 1.0 | 21:35 |
| inc0 | we need orch for 1.0 as far as I'm concerned | 21:35 |
| kfox1111 | inc0: we're having this because we're adding a second orchestration system now. :/ | 21:36 |
| kfox1111 | we punted a year now. | 21:36 |
| inc0 | kfox1111: we don't have one fully functional | 21:36 |
| kfox1111 | now we have two things and its getting worse. :/ | 21:36 |
| sdake | inc0 i think what i fundamentallly hear in this converstaion is kolla-kubernetes - the deliverable - is really not kolla-kubernetes | 21:36 |
| sdake | instead it is helm-microcharts | 21:36 |
| inc0 | kfox1111: let's drop everything to one then and decide which one? | 21:36 |
| sdake | or kolla-helm-microcharts | 21:36 |
| kfox1111 | sdake: yeah. that. | 21:36 |
| inc0 | if we decide to drop entrypoint alltogether, I'm fine | 21:37 |
| kfox1111 | its a seperate deliverable. | 21:37 |
| sbezverk | kfox1111: where service chart would go? | 21:37 |
| inc0 | if we decide to use entrypoint+ansible or just ansible, I'm fine | 21:37 |
| kfox1111 | sbezverk: an orchestration repo? | 21:37 |
| inc0 | just let's have 1 and only one orch for 1.0 | 21:37 |
| sbezverk | kfox1111: ok so we will have 1 orch repo for all orchstrations? | 21:37 |
| sdake | kolla-kubernetes-service-helm-entrypoint | 21:37 |
| kfox1111 | inc0: orchestration arch is defined by those willing to work on it. :/ | 21:37 |
| kfox1111 | inc0: I don't know if we can decide that. | 21:38 |
| inc0 | yes, and we have only few people by this table | 21:38 |
| sbezverk | sdake: it will be tiny and probably not use much other than at the gate | 21:38 |
| inc0 | and with these few people we have to make orch | 21:38 |
| inc0 | that's why we need single | 21:38 |
| inc0 | to keep our focus | 21:38 |
| sbezverk | that was why I thought it would make sense to keep it with microcharts | 21:38 |
| kfox1111 | inc0: my primary interest is in having a stable microcharts set, and ideally a stable configmap set. | 21:38 |
| sdake | inc0 i think what kfox1111 is saying is he doen't care about orch - just not on his lawn :) | 21:38 |
| *** athomas has quit IRC | 21:38 | |
| inc0 | with current amount of devs/reviewers we can't affort having more than one | 21:38 |
| kfox1111 | the imporatnce of that is on a much quicker timescale then orchestration IMO. | 21:39 |
| inc0 | kfox1111, you're free to opt out from this, just let us work in the repo | 21:39 |
| sdake | kfox1111 what inc0 said there is your plan is not strategically sound and I agree :) | 21:39 |
| inc0 | I'm thinking from perspective of recrutment of users/contributors | 21:39 |
| sdake | clearly we could make a micorcharts repo - slap 1.0 on it - but would anyone use it? | 21:40 |
| inc0 | that's important side to it too | 21:40 |
| kfox1111 | I am too. | 21:40 |
| kfox1111 | we just dont' see eye to eye on recrutment of users/contributors. | 21:40 |
| kfox1111 | you think having one thing to look at will increase it. | 21:40 |
| kfox1111 | I think it will turn users away. | 21:40 |
| sbezverk | sdake: everybody will use it, unless you plan to rewrite charts | 21:40 |
| kfox1111 | its teh same debate as kolla/kolla-ansible split. | 21:40 |
| *** schwicht has quit IRC | 21:40 | |
| inc0 | sbezverk: not everyone | 21:40 |
| sdake | sbezverk people use kolla containers because of kolla-ansible | 21:41 |
| inc0 | as much as they might want to, without easy start (and we don't have easy start), nobody will | 21:41 |
| sbezverk | inc0: are you planning develop new microcharts? | 21:41 |
| sdake | that is the reason | 21:41 |
| kfox1111 | some people will see ansible in kolla-kubernetes and start contributing. | 21:41 |
| kfox1111 | some will take one look and run aways screeming, and start a competitor. | 21:41 |
| inc0 | sbezverk: some people will. Let's be honest, microcharts aren't easy to handle at first glance | 21:41 |
| kfox1111 | if its a split repo, then they won't nessisarily just run away. | 21:42 |
| inc0 | they're often complex and counterintuitive | 21:42 |
| sdake | took me 3 months to sort out how to use microcharts | 21:42 |
| sdake | daily working on it | 21:42 |
| sbezverk | inc0: wow that is really new :( | 21:42 |
| sdake | i'm a fairly intelligent and dedicated person | 21:42 |
| jascott1 | same difference as a subfolder called `orchestration` with multiple approaches? | 21:42 |
| inc0 | I want single approach. | 21:42 |
| inc0 | but I want to have helm install openstack | 21:42 |
| inc0 | much like we have kolla-ansible deploy | 21:42 |
| kfox1111 | inc0: I know you want a single aproach. but I dout it will happen. :/ | 21:42 |
| *** emccormick has quit IRC | 21:43 | |
| inc0 | kfox1111: for today | 21:43 |
| inc0 | I'm talking short term | 21:43 |
| inc0 | one that works | 21:43 |
| kfox1111 | inc0: this is the same argument as kolla-ansible. | 21:43 |
| inc0 | before other will naturally be created | 21:43 |
| kfox1111 | if there was one true way, it would have been kolla-ansible, right? | 21:43 |
| kfox1111 | no. because a different way became available. kubernetes. | 21:43 |
| inc0 | yes, exactly, don't forget that we pretty miuch finished ansible before k8s was created | 21:43 |
| inc0 | that had 3 main roles: | 21:44 |
| kfox1111 | all I'm saying is, different people have different interests. | 21:44 |
| inc0 | 1. proven our approach works | 21:44 |
| kfox1111 | some such as att want pure helm. | 21:44 |
| inc0 | (images and such) | 21:44 |
| sdake | the deployment toolign is what bootstraps teh building block (e.g. microcharts are bootstrapped by some kubernetes deployment tech) | 21:44 |
| sdake | the building block alone is useless | 21:44 |
| kfox1111 | some want ansible because they know it/ like it/ can develop in it. | 21:44 |
| inc0 | 2. provided template for other (mounts and stuff) | 21:44 |
| kfox1111 | some want other tools. because they are already invested it that other thing. | 21:44 |
| inc0 | 3. recruited community | 21:44 |
| kfox1111 | I kind of doubt we will ever see just one way to orchestrat it. | 21:45 |
| sbezverk | sdake: I would say orch tools are useless without well defined building blocks | 21:45 |
| inc0 | kfox1111: yes we started with ansible because that allowed us to move fast, but we need one to move fast | 21:45 |
| inc0 | sbezverk: yes, they need to come together | 21:45 |
| inc0 | and you need orch *for* building blocks too | 21:45 |
| inc0 | to confirm they're working as intended in all the scenrios | 21:46 |
| inc0 | scenarios* | 21:46 |
| kfox1111 | inc0: technically, most of the contributaions so far have been twoards helm+entrypoint. | 21:46 |
| kfox1111 | your proposing a second orchestration strategy. | 21:46 |
| sbezverk | inc0: that is why still in my mind this separation building block from workers placing these blocks make sense.. | 21:46 |
| inc0 | then let's do helm+entrypoint, I'm fine with that | 21:46 |
| sdake | helm+entrypooint doesn't work | 21:46 |
| *** ansmith has joined #openstack-kolla | 21:46 | |
| sdake | beene there | 21:46 |
| sdake | done that | 21:46 |
| inc0 | unless that ^ | 21:46 |
| sbezverk | sdake: tell it to the gate ;) | 21:46 |
| kfox1111 | I'm not convinced helm+entrypoint's a good idea. just saying different contributors have different ideas on how to take things. | 21:46 |
| sdake | ok fellas, i've hit the limit of my tolerance for pain today - ttyl | 21:46 |
| sbezverk | if you do something wrong does not mean the tool is qrong ;)\ | 21:47 |
| kfox1111 | I can't say what contributors will contribute. | 21:47 |
| inc0 | sbezverk: we can stick to entrypoint as far as I'm concerned | 21:47 |
| inc0 | or entrypoint with some ansible, that's actually my favorite | 21:47 |
| kfox1111 | if inc0 wasnts to work on an ansible container and sbezverk (not putting words in your mouth. just an example) wants to contribute to heml+entrypoint, its those devs choices. | 21:47 |
| inc0 | but *one* | 21:47 |
| kfox1111 | I can't say which one we should allow. | 21:47 |
| inc0 | well for now we need ansible regardless for configs | 21:48 |
| jascott1 | or rather disallow | 21:48 |
| openstackgerrit | sean mooney proposed openstack/kolla master: introduces support for the OVS DPDK dataplane https://review.openstack.org/342354 | 21:48 |
| kfox1111 | I don't think we really have a good idea on which way is best. | 21:48 |
| inc0 | and we still don't have good idea for multinode mariadb or easy upgrades | 21:48 |
| kfox1111 | I think it should maybe incubate a bit more. | 21:48 |
| kfox1111 | try out multiple things. | 21:48 |
| kfox1111 | right. | 21:48 |
| inc0 | yes...and when we decide on *one* that works | 21:48 |
| inc0 | then we split | 21:48 |
| inc0 | and allow more than one to take from the one | 21:48 |
| kfox1111 | so we keep all the crap in git forever? | 21:49 |
| sbezverk | kfox1111: what about incubate with 2 repos, 1 -chars and 2nd -orchestration? | 21:49 |
| kfox1111 | or we keep it out of tree, prove out one, and use that. | 21:49 |
| jascott1 | thats what its for :) | 21:49 |
| sbezverk | in this case we can see where it goes with orch repo | 21:49 |
| openstackgerrit | sean mooney proposed openstack/kolla-ansible master: This change allow setting kernel commandline args https://review.openstack.org/408876 | 21:49 |
| inc0 | no, we split with first fully functional implementation and prior to 1.0 | 21:49 |
| openstackgerrit | sean mooney proposed openstack/kolla-ansible master: introduce contrib playbook for ovs-dpdk https://review.openstack.org/408872 | 21:49 |
| inc0 | how about that? | 21:49 |
| kfox1111 | sbezverk: that works for me. | 21:49 |
| sbezverk | maybe it will stay just ansible forever | 21:49 |
| sbezverk | inc0: would it work for you? | 21:49 |
| inc0 | I wouldn't be surprised if it works well | 21:49 |
| kfox1111 | inc0: then we're carrying history of junk forever. | 21:49 |
| kfox1111 | inc0: it works well for those who are ansible folks. | 21:50 |
| jascott1 | yeah thats git | 21:50 |
| inc0 | so you want to create new repo? | 21:50 |
| kfox1111 | inc0: it will be poison to those against ansible. :/ | 21:50 |
| inc0 | kfox1111: unless ansible runs in job | 21:50 |
| inc0 | which is idea I like most | 21:50 |
| kfox1111 | yeah. I like that idea the most too. | 21:50 |
| inc0 | then poeple who doesn't like ansible won't even need to look at | 21:50 |
| inc0 | it | 21:50 |
| inc0 | however splitting repos will make it much harder | 21:50 |
| sbezverk | inc0: kfox1111: + 1 for second orch repo | 21:51 |
| inc0 | because if we want to put ansible in jopb then we need helm chart for it | 21:51 |
| inc0 | so we'll have helm chart for orch | 21:51 |
| kfox1111 | inc0: I think it simplifies the arch if its split? | 21:51 |
| inc0 | I'm -1 for second repo today. | 21:51 |
| kfox1111 | inc0: think of it this way... | 21:51 |
| inc0 | I need containe rimage with helm and ansible in it | 21:52 |
| inc0 | as dir contents | 21:52 |
| inc0 | today its simple ADD | 21:52 |
| kfox1111 | kolla-kubernetes-orchestration has the operator code/etc. | 21:52 |
| inc0 | with multi repo it's nightmare | 21:52 |
| kfox1111 | kolla-kubernetes-microcharts has the helm charts. | 21:52 |
| kfox1111 | I think the aproach your talking in the end is, | 21:52 |
| kfox1111 | a set of containers with ansible in it. these read k8s 3rd party resources. | 21:53 |
| inc0 | how do I create container image with both helm charts and ansible code? | 21:53 |
| inc0 | if for example I'd like to call helm from ansible | 21:53 |
| jascott1 | carefully | 21:53 |
| inc0 | ? | 21:53 |
| kfox1111 | so user helm installs kolla/operator, then kubectl apply -f mycloudconfig? | 21:53 |
| kfox1111 | which that microchart pulls in containers built from the -operator repo. | 21:53 |
| kfox1111 | helm code would be in microservices then, from that repo. | 21:54 |
| kfox1111 | source for the orchestration stuff would be in operators repo. | 21:54 |
| kfox1111 | it can be built by the user, or skipped entirely and use upstream repo's. | 21:54 |
| kfox1111 | images I mean. | 21:55 |
| inc0 | ehh, I really don't like that at thius stage of project | 21:55 |
| inc0 | srsly, it's hard to handle today, let's not make it harder | 21:55 |
| sbezverk | inc0: but it looks like it is the best time to do it | 21:55 |
| inc0 | I wholehearly disagree | 21:55 |
| inc0 | it's easier time to do it | 21:56 |
| inc0 | not best | 21:56 |
| sbezverk | since code is just on the process of being developed | 21:56 |
| sbezverk | you will not need to change later | 21:56 |
| inc0 | easiest != best in this case imho | 21:56 |
| kfox1111 | its going to be much harder to unravel the gate stuff if we build it together now and try and rip it apart later. | 21:56 |
| kfox1111 | really hard. :? | 21:56 |
| inc0 | well, if we rip apart config generation today | 21:56 |
| inc0 | it's gonna make gates not functional at all | 21:57 |
| inc0 | until we're pretty much done | 21:57 |
| inc0 | if we work in tree we can move iteratively | 21:57 |
| inc0 | to a point of separation | 21:57 |
| kfox1111 | I'm not talking about moving the templates out of the microservices repo. | 21:57 |
| inc0 | which, if we do it correctly, will also be point of functionality | 21:57 |
| kfox1111 | thats a temporary fork until those templates move to helm. | 21:57 |
| kfox1111 | so its in the right place mostly already. | 21:58 |
| inc0 | and orchestration for deployment? | 21:58 |
| inc0 | in gate we test functional openstack | 21:58 |
| kfox1111 | the tests as is mostly test the functionality of the microservice charts. | 21:58 |
| inc0 | wich will include orch engine | 21:58 |
| inc0 | but add upgrades to gate | 21:58 |
| inc0 | add HA of mariadb to gate | 21:58 |
| kfox1111 | the new code being discussed is testing an operator that manages that process itself. | 21:58 |
| inc0 | and it grows to functional orch | 21:58 |
| kfox1111 | my 2 cents on mariadb? | 21:59 |
| inc0 | and if we rip entrypoint from gate, you won't have that | 21:59 |
| *** jamesbenson_ has quit IRC | 21:59 | |
| inc0 | run mariadb out of k8s?;) | 21:59 |
| kfox1111 | my crystal ball is something like: | 21:59 |
| kfox1111 | make a microservice for mariadb that allows host storage passthrough/host pinning. | 21:59 |
| kfox1111 | and 2, make an etcd-operator like mariadb-operator that manages instances of that microservice chart. | 22:00 |
| inc0 | no floating around? | 22:00 |
| sbezverk | inc0: let's try to add things without removing, like genconfig or entry point gate jpobs | 22:00 |
| kfox1111 | no need to float, when you can loose individual nodes in the cluster. | 22:00 |
| inc0 | sbezverk: that's easier with single repo and I agree | 22:00 |
| kfox1111 | if you want to float, ou just run one. no cluster. | 22:00 |
| inc0 | with single repo we can keep adding code while keeping gates green | 22:01 |
| kfox1111 | I guess maybe I shoudl word it this way... | 22:01 |
| inc0 | and have separation in mind | 22:01 |
| *** jamesbenson has joined #openstack-kolla | 22:01 | |
| kfox1111 | microservices are about building blocks that provide a standard api. | 22:01 |
| kfox1111 | HOW they implement the api is kind of up to the containers pulled in. | 22:02 |
| inc0 | so at some point, when we have confirmation that we can separate (which to me is - when we can deploy and upgrade prod openstac) | 22:02 |
| kfox1111 | that chunk of functionality is very vendor/tool neutral. | 22:02 |
| kfox1111 | so we can get a lot of contributors to it. | 22:02 |
| inc0 | we also have code to separate and keep us functional | 22:02 |
| jascott1 | aaaaaaand thats an hour on "repos". bbiab | 22:02 |
| *** jascott1 has quit IRC | 22:02 | |
| *** jascott1 has joined #openstack-kolla | 22:03 | |
| kfox1111 | the implementatino of standard job X, using ansible, is the more contentious part? image implementation code in a seperate repo would be less contentious? | 22:03 |
| kfox1111 | well.... I guess my concern is, the api for orchestration bits will never be very clean. :/ | 22:04 |
| openstackgerrit | Bertrand Lallau proposed openstack/kolla-ansible master: Fix outdated barbican-api-paste.ini file https://review.openstack.org/469939 | 22:04 |
| kfox1111 | I don't know man. I think its a bad idea to mix orchestration on low level components. | 22:05 |
| kfox1111 | I've remained silent for a long time, and now its coming to a head. | 22:05 |
| *** jascott1 has quit IRC | 22:05 | |
| kfox1111 | I don't want to fight over it much . but think its hurting the project. | 22:05 |
| *** schwicht has joined #openstack-kolla | 22:06 | |
| kfox1111 | how about this compromize though.... | 22:06 |
| *** jamesbenson has quit IRC | 22:06 | |
| kfox1111 | keep all the orchestration stuff under /orchestration/ansible/ including gate tests and docker files? | 22:06 |
| kfox1111 | the remaining issue is how do we release 1.0 of the microservices before we have orchestration ready? | 22:07 |
| kfox1111 | I don't think openstack release team can allow that without seperate repo's. | 22:07 |
| inc0 | kfox1111: only reason I need dockerfile in top dir is because docker doesn't support ADD ../../helm | 22:07 |
| inc0 | we need orch for 1.0 period. | 22:08 |
| kfox1111 | microservices should e versioned seperately. | 22:08 |
| inc0 | because we need upgrades tested and proven workign | 22:08 |
| kfox1111 | isn't there any way around that? that is painful. | 22:08 |
| kfox1111 | alternate build scripts or something. | 22:08 |
| inc0 | arguably each microservice could be versioned separately from each other | 22:08 |
| kfox1111 | or use kolla's build system. | 22:08 |
| kfox1111 | inc0: and I think that's a good idea too. | 22:09 |
| inc0 | kolla build system won't work :( | 22:09 |
| *** shardy has quit IRC | 22:09 | |
| kfox1111 | but I have seen pushback at my site due to 'kolla-kubernetes not being 1.0' | 22:09 |
| inc0 | I need kolla-k8s code in it | 22:09 |
| kfox1111 | kolla build wont work today. but can't it easily be adapted? | 22:09 |
| inc0 | well, it doesn't give us anything | 22:09 |
| kfox1111 | circular logic. | 22:10 |
| kfox1111 | you don't think it gives us something. I do. | 22:10 |
| inc0 | we can write hacky shell that will tarball /helm dir and put it in dockerfile contex | 22:10 |
| inc0 | I mean it doesn't solve problem we have here | 22:10 |
| kfox1111 | ah. that. yeah. | 22:10 |
| inc0 | we need fresh dev code with helm charts inside built container | 22:11 |
| kfox1111 | the only one we need for that is helm-repo's I think. | 22:11 |
| kfox1111 | but thats optional too. | 22:11 |
| kfox1111 | you should be able to pull from tarballs.openstack.org stuff we push. | 22:11 |
| inc0 | well we *can* have dockerfile in kolla-k8s and expect people to build | 22:11 |
| kfox1111 | with nothing on the system what so ever, I want: | 22:11 |
| inc0 | and have docker image built for every release | 22:11 |
| kfox1111 | helm repo add kolla tarballs.o.o/.... | 22:12 |
| kfox1111 | helm install kolla/xxxx to work. | 22:12 |
| kfox1111 | so... | 22:12 |
| inc0 | that's possible and I totally want that | 22:12 |
| kfox1111 | wait. we don't need the charts in the container. | 22:12 |
| inc0 | we can achieve that with having orch chart | 22:12 |
| kfox1111 | we just need it loadable via a volume. | 22:12 |
| kfox1111 | or curlable. | 22:12 |
| inc0 | hmm | 22:13 |
| kfox1111 | thats what we do with helm-repo. | 22:13 |
| inc0 | but we can't bank on tarballs for everything | 22:13 |
| inc0 | dev+gate | 22:13 |
| kfox1111 | helm pacakges are just tarballs. | 22:13 |
| sbezverk | kfox1111: how about suing helm repo container? | 22:13 |
| kfox1111 | so.. helm has a build in server. | 22:13 |
| inc0 | so let me put it this way, with image that has helm charts in it | 22:13 |
| sbezverk | it is already there and I used it in the past | 22:14 |
| inc0 | we can do kubectl run kolla-helm-repo\ | 22:14 |
| kfox1111 | you can install packages right from your dev system. | 22:14 |
| inc0 | that will build and create helm repo automagically, that'd be perfect | 22:14 |
| *** salv-orl_ has joined #openstack-kolla | 22:14 | |
| kfox1111 | building the packages requires a shell script anyway. | 22:15 |
| inc0 | right...which means we can do it in job | 22:15 |
| kfox1111 | I guess it doesn't much matter if its shipped as a container a tarball of tarballs, or whatever else. | 22:15 |
| inc0 | git checkout kolla-kubernetes; helm install helm/orchestration/compute-kit | 22:15 |
| kfox1111 | checkout won't work by itself. | 22:16 |
| kfox1111 | you need a build step. | 22:16 |
| inc0 | nto for simple chart | 22:16 |
| kfox1111 | complute-kit wont be simple. | 22:16 |
| inc0 | so let say compute-kit chart won't need kolla-common | 22:16 |
| kfox1111 | and if you want it to use the local charts, many will need building. | 22:16 |
| inc0 | but what it will do is run pod where it builds | 22:16 |
| kfox1111 | if you want to use remote charts, no need to git checkout at all. | 22:17 |
| inc0 | and serves built repo for helm | 22:17 |
| inc0 | both works together well | 22:17 |
| sbezverk | that is what helm-repo container is for | 22:17 |
| sbezverk | to service charts remotely | 22:17 |
| kfox1111 | sbezverk: yeah. | 22:18 |
| *** salv-orlando has quit IRC | 22:18 | |
| inc0 | helm-repo container needs helm chart code | 22:18 |
| kfox1111 | hostPath | 22:18 |
| inc0 | assumes you have helm charts in one of k8s nodes | 22:18 |
| inc0 | why not docker build . --tag kolla-orch | 22:18 |
| inc0 | and ADD helm; ADD ansible | 22:19 |
| inc0 | only requirement for that is Dockerfile in root dir | 22:19 |
| kfox1111 | ideally we would have a k8s volume type that was basically "use this docker image as a volume" | 22:19 |
| kfox1111 | mixing the charts and the ansible code seeems anti k8s to me. | 22:20 |
| inc0 | yeah I think it would be great addition to k8s | 22:20 |
| kfox1111 | mixes two different things in a fatter container. | 22:20 |
| inc0 | "dear k8s, make this image available to yourself" | 22:20 |
| kfox1111 | yeah. | 22:20 |
| kfox1111 | might not be too hard to write that plugin. | 22:20 |
| kfox1111 | maybe or k8s dev friends would pick that one up? | 22:20 |
| inc0 | well to make image availabie you need docker underneath support it | 22:21 |
| inc0 | anyway, we digress | 22:21 |
| kfox1111 | docker supports volume containers. | 22:21 |
| inc0 | you'd need to make qow shapshot not just plain volume | 22:22 |
| kfox1111 | so, for ansible, you need a way to get fresh charts easily. | 22:22 |
| inc0 | I mean technically you can just do "volume mount /" | 22:22 |
| kfox1111 | you can get that from helm serv running on the dev system? | 22:22 |
| kfox1111 | and an arg to the helm chart for the ansible stuff pointing at the right repo? | 22:22 |
| inc0 | btw helm charts are just tarballs right...can we start publishing them in tarballs.o.o? | 22:23 |
| inc0 | simpe file server shoudl suffice right? | 22:23 |
| kfox1111 | its like a yum repo. | 22:23 |
| kfox1111 | a set of tarballs, and an index file. | 22:23 |
| kfox1111 | should be able to push them pretty easily I think. | 22:23 |
| kfox1111 | yeah. a helm server is just http serving those files. | 22:24 |
| inc0 | http://tarballs.openstack.org/kolla/images/ that would work? | 22:24 |
| inc0 | I can write quick change to infra to start publishing microcharts on merge. | 22:25 |
| kfox1111 | I'd suggest: | 22:25 |
| kfox1111 | http://tarballs.openstack.org/kolla-kubernetes/ under helm/<openstackreleasename> | 22:25 |
| sbezverk | kfox1111: do you have link to docker volume containers? | 22:25 |
| sbezverk | I heard about the concept but never seen implementation | 22:25 |
| inc0 | ok I'll work on that | 22:25 |
| inc0 | but besides that, can we make one liner that will build charts and make them available in k8s? | 22:26 |
| inc0 | tools/build_all --push | 22:26 |
| sbezverk | inc0: push where? | 22:27 |
| kfox1111 | inc0: the gate already copies stuff there. should be pretty easy to get it extended. | 22:27 |
| inc0 | it's rly easy | 22:27 |
| inc0 | sbezverk: I'd say some registry somewhere | 22:27 |
| sbezverk | inc0: can you explain why help-repo does not work for you. | 22:27 |
| inc0 | much like kolla-ansible does | 22:27 |
| inc0 | and kubectl run pod-with-repo | 22:27 |
| sbezverk | it has a service which you can reference from any container in the cluster | 22:27 |
| inc0 | sbezverk: I want built charts there | 22:28 |
| inc0 | for kolla | 22:28 |
| sbezverk | it build all charts | 22:28 |
| kfox1111 | sbezverk: https://www.digitalocean.com/community/tutorials/how-to-work-with-docker-data-volumes-on-ubuntu-14-04 | 22:28 |
| sbezverk | and services them on request | 22:28 |
| sbezverk | from what I see it is exactly what you are asking | 22:28 |
| sbezverk | kfox1111: thanks will check | 22:28 |
| kfox1111 | sbezverk: you basically create an instance of a blank container, and then you use --volume-from to pull it into the real container. | 22:28 |
| inc0 | sbezverk: where do you have helm-repo iumage? | 22:29 |
| sbezverk | inc0: to deploy a chart all you need is to use url in your ansible thingy | 22:29 |
| sbezverk | pointing it to helm-repo service | 22:29 |
| sbezverk | that is it | 22:29 |
| inc0 | how do you deploy helm-repo? | 22:30 |
| *** jascott1 has joined #openstack-kolla | 22:30 | |
| kfox1111 | jascott1: you interested in k8s development? :) | 22:30 |
| *** ducttape_ has joined #openstack-kolla | 22:30 | |
| jascott1 | i am! | 22:30 |
| sbezverk | inc0: if you want I can build PoC for you ;) | 22:30 |
| kfox1111 | jascott1: we have a feature request for k8s. maybe you might be interested in it. | 22:30 |
| sbezverk | inc0: there is a chart | 22:30 |
| sbezverk | to deploy it\ | 22:31 |
| inc0 | of image? I have it here;) https://review.openstack.org/#/c/474770/ | 22:31 |
| kfox1111 | the idea is, k8s has volumes. It would be cool if you could use an image as a volume. | 22:31 |
| kfox1111 | volume: | 22:31 |
| kfox1111 | image: | 22:31 |
| kfox1111 | name: kolla/foo:bar | 22:31 |
| sbezverk | kfox1111: it will be less flexible | 22:32 |
| sbezverk | in terms of updates | 22:32 |
| jascott1 | what special powers would it get? (same as mounting a disk image with same contents?) | 22:32 |
| sbezverk | of charts | 22:32 |
| inc0 | I'd say something better | 22:32 |
| inc0 | kubectl image create my-local-docker-image:latest | 22:32 |
| inc0 | and poof image is available for pods to use | 22:32 |
| kfox1111 | jascott1: yeah. very similar to the git volume type. but would use the docker thats already there to do pulling/mounting. | 22:32 |
| inc0 | sort of glance for k8s | 22:32 |
| jascott1 | all kubernetes projects are in `stability and security fixes` atm | 22:33 |
| kfox1111 | inc0: no, they wont suppor tthat. | 22:33 |
| jascott1 | according to technosophos | 22:33 |
| sbezverk | inc0: you will need to sell it to k8s community ;) | 22:33 |
| kfox1111 | images are out of scope of k8s. | 22:33 |
| *** ducttap__ has quit IRC | 22:33 | |
| kfox1111 | hub.docker.com/quay/artefactory/etc all can be used instead, | 22:33 |
| kfox1111 | and you can load your own repos into k8s too. | 22:34 |
| kfox1111 | but, what may be in scope, is using an image as a volume. | 22:34 |
| sbezverk | kfox1111: right, we have troubles with netwokring enhancements, they will kick us out for images things | 22:34 |
| inc0 | kfox1111: btw I wonder if docker run -v /path/to/image:/ whatever-image would work | 22:34 |
| kfox1111 | inc0: thats what hostPath does. | 22:34 |
| *** gfidente|afk has quit IRC | 22:35 | |
| kfox1111 | and what --volume-from I think does. | 22:35 |
| *** schwicht has quit IRC | 22:35 | |
| kfox1111 | one other possibility.... | 22:38 |
| kfox1111 | it might work with just a sidecar. | 22:38 |
| kfox1111 | can you run docker enough in a container to pull/dump an image? | 22:38 |
| kfox1111 | still, I guess having it be an actual k8s volume plugin lets it cache the image in the real docker's image cache. | 22:39 |
| kfox1111 | so pod restart is fast. | 22:39 |
| *** salv-orl_ has quit IRC | 22:41 | |
| *** StephenWang1991 has joined #openstack-kolla | 22:41 | |
| Pavo | I am still concerned with no instructions in docs saying to install just kolla lol | 22:43 |
| inc0 | Pavo: well...we must've missed that one;) since it's one line | 22:44 |
| Pavo | I might be blind though but I can not find that anywhere in the docs | 22:44 |
| inc0 | pip install kolla;) | 22:44 |
| Pavo | yeah, I did that and now have kolla-build now but really don't see it anywhere in the docs just a FYI | 22:44 |
| *** renmak_ has quit IRC | 22:45 | |
| inc0 | Pavo: wanna make commit?:) | 22:45 |
| sbezverk | inc0: kfox1111: so do we split repo, or not or we discuss more on ML? | 22:45 |
| Pavo | inc0 remember someone was gonna teach me how to do that | 22:45 |
| Pavo | lol | 22:45 |
| inc0 | sbezverk: can we not please? And discuss when we have prototype orch ready | 22:45 |
| Pavo | never did though | 22:45 |
| inc0 | at least prototype that doesn't look awfully hacky | 22:46 |
| jascott1 | Pavo are you talking about contributing in general? I can help if you have questions | 22:46 |
| *** StephenWang1991 has quit IRC | 22:46 | |
| Pavo | jascott1 just submitting issues I see in docs | 22:46 |
| Pavo | for now | 22:46 |
| Pavo | contributing in general would probably help though | 22:47 |
| jascott1 | its nice to go ahead and submit a patch for what you think it should be. then people can see that and weigh in, probably alter it a bit and then merge and done. | 22:47 |
| Pavo | jascott1 PM? | 22:48 |
| jascott1 | sure | 22:48 |
| inc0 | I gtg guys, talk to you later | 22:49 |
| kfox1111 | inc0: wait. | 22:50 |
| kfox1111 | can we agree to keep it in the same repo just for now, before 1.0, | 22:50 |
| kolla-slack | <inc0> Ok | 22:50 |
| kfox1111 | if its all kept to /orchestration/ansible/*, including gate tests, build stuff, the works? | 22:50 |
| kfox1111 | then it should be much easier to split later. | 22:50 |
| *** renmak_ has joined #openstack-kolla | 22:50 | |
| kfox1111 | I don't really like it, but seems like a reasonable middle ground? | 22:51 |
| kolla-slack | <inc0> Yes, I'm ok with that | 22:51 |
| kfox1111 | sbezverk: you good with that? | 22:51 |
| sbezverk | kfox1111: yep at least it is some way of organizing things | 22:51 |
| kfox1111 | ok. cool. | 22:51 |
| kolla-slack | <inc0> How do you want to run gates with no orch at all tho? | 22:52 |
| sbezverk | inc0: each orch will need to proivde gate job I think | 22:52 |
| kolla-slack | <inc0> Anyway we can figure this out as we go | 22:52 |
| sbezverk | which should mimic tests we run currently at very least | 22:53 |
| kolla-slack | <inc0> Well what I'm saying is our gate today have prothesis orch | 22:53 |
| kolla-slack | <inc0> Which we want to get rid of and replace with something better | 22:53 |
| sbezverk | inc0: our gates now have globally available jobs bash is everywhere | 22:54 |
| sbezverk | ansible is not ;)\ | 22:54 |
| jascott1 | anyone have that openstack doc link handy? the one about setting up gerrit etc | 22:54 |
| kolla-slack | <inc0> So I'd gladly get rid of prothesis orch while creating proper one | 22:54 |
| jascott1 | i bet the PTL know that link right? | 22:54 |
| kolla-slack | <inc0> I'm on my way to car so not rly;) | 22:55 |
| jascott1 | oh i found it thanks | 22:55 |
| kolla-slack | <inc0> Sergey with today's gate doing upgrade is pain | 22:55 |
| kolla-slack | <inc0> Configs need to be moved to helm | 22:56 |
| kolla-slack | <inc0> And entrypoint shouldn't be in microchart repo | 22:56 |
| kolla-slack | <inc0> So current gates will not work with split | 22:57 |
| kolla-slack | <inc0> I think shell gate should be replaced with something we'll suggest for prod | 22:58 |
| kolla-slack | <inc0> No reason to support them once we have prod ready orch | 22:58 |
| kfox1111 | inc0: the orchestration stuff will need its own jgate job anyway. | 22:59 |
| kfox1111 | can just point that job entrypoint into /orchestration/ansible | 23:00 |
| kfox1111 | its just fine for that job to call back into the other gate scripts though. | 23:00 |
| *** rwallner has quit IRC | 23:02 | |
| sbezverk | inc0: the only conecr I have is doing gate job for a specific orch kind of give preference to it. using something orch agnostic makes sure all orchs will get the same level | 23:03 |
| sbezverk | of testing | 23:03 |
| kfox1111 | +1 | 23:07 |
| kfox1111 | lets not gut the gate jobs we have already. they work ok. lets add new ones to test the orchestration. | 23:08 |
| sbezverk | kfox1111: totally agree | 23:11 |
| *** mmehan has quit IRC | 23:11 | |
| *** ducttape_ has quit IRC | 23:12 | |
| *** rwallner has joined #openstack-kolla | 23:15 | |
| *** kolla-slack1 has joined #openstack-kolla | 23:18 | |
| *** kolla-slack has quit IRC | 23:18 | |
| *** Pavo has quit IRC | 23:19 | |
| jascott1 | have we documented what needs to be done for orchestration? ive heard HA db, queue etc | 23:21 |
| kfox1111 | no. :/ | 23:22 |
| jascott1 | id love to work on an operator for a piece of the puzzle but not sure what it needs to do (because of my lack of operator experience) | 23:23 |
| *** tonanhngo has quit IRC | 23:26 | |
| *** jamesbenson has joined #openstack-kolla | 23:27 | |
| *** rwallner has quit IRC | 23:31 | |
| *** jamesbenson has quit IRC | 23:31 | |
| kfox1111 | jascott1: that would be sweet. :) | 23:32 |
| kfox1111 | I'd really like to see an operator that works like etcd-operator, but for galera. | 23:32 |
| kfox1111 | I think that would go a really long way. :) | 23:32 |
| jascott1 | i was looking into that | 23:34 |
| jascott1 | i would beg that it could be done in go as it would be easier to follow the existing operator(s) | 23:34 |
| jascott1 | thats the argument I will use anyway :) | 23:34 |
| jascott1 | there exists one as a sample already | 23:35 |
| kfox1111 | works for me. :) | 23:35 |
| *** tonanhngo has joined #openstack-kolla | 23:36 | |
| kfox1111 | not sure, ut there may not be significant amounts of code differences between the two. | 23:36 |
| kfox1111 | might be able to use the same codebase for both etcd and galera | 23:36 |
| jascott1 | i think I could get a skeleton up that follows the operator pattern and we could plug in 'what it needs to do' | 23:36 |
| kfox1111 | (kind of what Iwas suggesting trove do...) | 23:37 |
| kfox1111 | cool. | 23:37 |
| jascott1 | fun even! | 23:37 |
| kfox1111 | :) | 23:37 |
| kfox1111 | jascott1: I did get flannel running on my system from that chart and static pod helm. | 23:37 |
| *** tonanhngo_ has joined #openstack-kolla | 23:37 | |
| kfox1111 | so totally viable workflow. :) | 23:38 |
| jascott1 | nice! | 23:38 |
| kfox1111 | next I wana try a k8s control plane. | 23:38 |
| kfox1111 | :) | 23:38 |
| kfox1111 | I think you could make a chart that does kube-apiserver,kube-conroller-manager,kube-scheduler,flannel,kube-dns,kube-dashboard. :) | 23:39 |
| *** tonanhngo has quit IRC | 23:40 | |
| *** tonanhngo_ has quit IRC | 23:42 | |
| *** ducttape_ has joined #openstack-kolla | 23:51 | |
| *** ducttape_ has quit IRC | 23:55 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!