| *** jtriley has joined #kolla | 00:11 | |
| *** diogogmt has joined #kolla | 00:18 | |
| *** weiyu_ has joined #kolla | 00:20 | |
| *** shadower has quit IRC | 00:23 | |
| *** shadower has joined #kolla | 00:23 | |
| *** dolpher has joined #kolla | 00:25 | |
| *** weiyu_ has quit IRC | 00:25 | |
| *** weiyu has joined #kolla | 00:27 | |
| *** bmace has quit IRC | 00:39 | |
| *** weiyu has quit IRC | 00:46 | |
| *** bmace has joined #kolla | 00:53 | |
| *** jasonsb has joined #kolla | 01:01 | |
| openstackgerrit | Sam Yaple proposed stackforge/kolla: Update Ansible docs to remove requirements https://review.openstack.org/209716 | 01:04 |
|---|---|---|
| *** jasonsb has quit IRC | 01:04 | |
| *** jasonsb_ has joined #kolla | 01:04 | |
| *** echoingu_ has quit IRC | 01:06 | |
| *** dolpher1 has joined #kolla | 01:11 | |
| *** dolpher has quit IRC | 01:13 | |
| *** weiyu has joined #kolla | 01:15 | |
| *** dims_ has joined #kolla | 01:21 | |
| *** erkules_ has joined #kolla | 01:22 | |
| *** dims has quit IRC | 01:22 | |
| *** erkules has quit IRC | 01:24 | |
| *** sdake has joined #kolla | 01:27 | |
| *** sdake_ has joined #kolla | 01:30 | |
| openstackgerrit | Sam Yaple proposed stackforge/kolla: Reduce dependencies on the dest nodes https://review.openstack.org/208451 | 01:32 |
| *** sdake has quit IRC | 01:34 | |
| *** SourabhP has quit IRC | 01:39 | |
| *** echoingumesh has joined #kolla | 01:41 | |
| openstackgerrit | Sam Yaple proposed stackforge/kolla: Updated some requirements about docker https://review.openstack.org/209722 | 01:43 |
| *** unicell has quit IRC | 01:52 | |
| *** unicell has joined #kolla | 01:53 | |
| sdake_ | samyaple did you see tripelo's plans | 01:55 |
| sdake_ | pretty cool going to use puppet to configure kola containers using tripleo ;-) | 01:56 |
| sdake_ | It hink using exclusvey config-external | 01:56 |
| sdake_ | so we can possibly just drop config-internal entirely | 01:57 |
| SamYaple | sdake_: yea thats awesome | 01:57 |
| SamYaple | i say Slower and ryan talking about it in the channel a few weeks ago | 01:58 |
| *** unicell has quit IRC | 01:58 | |
| SamYaple | looking forward to helping with that if possible | 01:58 |
| SamYaple | not a big puppet guy though | 01:58 |
| SamYaple | sdake_: the containerized deps is ready to test | 01:58 |
| *** achanda has joined #kolla | 02:00 | |
| *** dims has joined #kolla | 02:03 | |
| sdake_ | cool | 02:05 |
| sdake_ | dinner time bbl | 02:05 |
| *** dims_ has quit IRC | 02:05 | |
| *** dims_ has joined #kolla | 02:06 | |
| *** dims has quit IRC | 02:09 | |
| *** sdake_ has quit IRC | 02:09 | |
| openstackgerrit | Swapnil Kulkarni (coolsvap) proposed stackforge/kolla: Add ansible support for Cinder https://review.openstack.org/208868 | 02:17 |
| *** zhiwei has joined #kolla | 02:17 | |
| *** unicell has joined #kolla | 02:18 | |
| *** echoingumesh has quit IRC | 02:19 | |
| *** jtriley has quit IRC | 02:19 | |
| *** echoingumesh has joined #kolla | 02:20 | |
| *** unicell1 has joined #kolla | 02:22 | |
| *** dims_ has quit IRC | 02:22 | |
| *** unicell has quit IRC | 02:22 | |
| *** diogogmt has quit IRC | 02:22 | |
| *** echoingumesh has quit IRC | 02:24 | |
| *** achanda has quit IRC | 03:44 | |
| *** coolsvap|away is now known as coolsvap | 03:58 | |
| *** jasonsb_ has quit IRC | 04:01 | |
| *** jasonsb has joined #kolla | 04:03 | |
| *** weiyu has quit IRC | 04:24 | |
| *** achanda has joined #kolla | 04:50 | |
| *** achanda has quit IRC | 05:25 | |
| *** weiyu_ has joined #kolla | 05:28 | |
| *** achanda has joined #kolla | 05:31 | |
| *** SourabhP has joined #kolla | 05:50 | |
| *** dims has joined #kolla | 06:03 | |
| *** dims has quit IRC | 06:09 | |
| *** zhiwei has quit IRC | 07:03 | |
| *** juggler_ has joined #kolla | 07:14 | |
| *** juggler has quit IRC | 07:17 | |
| *** zhiwei has joined #kolla | 07:33 | |
| *** pbourke has quit IRC | 07:33 | |
| *** pbourke has joined #kolla | 07:33 | |
| *** jmccarthy has quit IRC | 07:36 | |
| *** jmccarthy has joined #kolla | 07:36 | |
| *** SourabhP has quit IRC | 07:47 | |
| *** sdake has joined #kolla | 07:54 | |
| sdake | evening folks | 07:54 |
| *** weiyu_ has quit IRC | 07:58 | |
| *** weiyu_ has joined #kolla | 08:04 | |
| *** tfukushima has joined #kolla | 08:12 | |
| *** achanda has quit IRC | 08:14 | |
| *** tfukushima has quit IRC | 08:15 | |
| *** tfukushima has joined #kolla | 08:17 | |
| *** achanda has joined #kolla | 08:18 | |
| *** tfukushima has quit IRC | 08:22 | |
| *** athomas has joined #kolla | 08:26 | |
| *** achanda has quit IRC | 08:29 | |
| *** weiyu_ has quit IRC | 08:39 | |
| *** achanda has joined #kolla | 08:40 | |
| *** weiyu_ has joined #kolla | 08:43 | |
| *** achanda has quit IRC | 08:51 | |
| *** athomas has quit IRC | 09:08 | |
| *** inc0 has joined #kolla | 09:13 | |
| inc0 | good morning | 09:15 |
| *** inc0 has quit IRC | 09:21 | |
| *** inc0 has joined #kolla | 09:21 | |
| *** athomas has joined #kolla | 09:26 | |
| *** inc0 has quit IRC | 09:27 | |
| *** inc0_ has joined #kolla | 09:27 | |
| openstackgerrit | weiyu proposed stackforge/kolla: Ceilometer config-ceilometer.sh file without backslash https://review.openstack.org/209473 | 09:30 |
| *** inc0_ has quit IRC | 09:32 | |
| *** inc0 has joined #kolla | 09:38 | |
| coolsvap | pbourke, around | 09:49 |
| *** dims has joined #kolla | 09:49 | |
| pbourke | coolsvap: hi | 09:49 |
| coolsvap | pbourke, do you have steps to get the ansible env running, i am not able to run it from yesterday | 09:49 |
| coolsvap | i think its some config issue | 09:50 |
| pbourke | coolsvap: im using harmw's vagrantfile env | 09:50 |
| pbourke | works very nicely | 09:50 |
| *** gfidente has joined #kolla | 09:50 | |
| coolsvap | let me set it up | 09:57 |
| *** inc0 has quit IRC | 10:00 | |
| *** dolpher1 has quit IRC | 10:01 | |
| *** inc0 has joined #kolla | 10:04 | |
| coolsvap | pbourke, i think similar issue will be there in https://review.openstack.org/#/c/205528/15/ansible/roles/heat/tasks/config.yml like you said for cinder | 10:07 |
| pbourke | looks like it | 10:08 |
| coolsvap | i am not able to install hostmanager plugin for vagrant | 10:10 |
| *** inc0 has quit IRC | 10:11 | |
| *** Haomeng has quit IRC | 10:19 | |
| *** Haomeng has joined #kolla | 10:19 | |
| *** weiyu_ has quit IRC | 10:28 | |
| *** weiyu_ has joined #kolla | 10:28 | |
| *** tfukushima has joined #kolla | 10:29 | |
| *** weiyu_ has quit IRC | 10:37 | |
| *** weiyu_ has joined #kolla | 10:44 | |
| *** inc0 has joined #kolla | 10:44 | |
| inc0 | hello | 10:44 |
| *** zhiwei has quit IRC | 10:45 | |
| inc0 | anyone ran kolla with kilo openstack? | 10:45 |
| pbourke | inc0: parts of it | 10:47 |
| inc0 | pbourke, do you have dockerfiles with kilo maybe? | 10:50 |
| inc0 | this should be as simple as changing repos, but never know | 10:50 |
| pbourke | inc0: we just have buildconfs to pin to the 2015.1.0 tag | 10:50 |
| inc0 | also, it will be Oracle linux right? | 10:50 |
| inc0 | aaa...source install | 10:50 |
| inc0 | even better | 10:50 |
| pbourke | yup :) | 10:51 |
| inc0 | I'll play with that, thanks:) | 10:51 |
| *** dims_ has joined #kolla | 10:51 | |
| pbourke | inc0: added a reply on https://review.openstack.org/#/c/209430/ | 10:53 |
| *** dims has quit IRC | 10:53 | |
| pbourke | does it make sense? | 10:53 |
| inc0 | yeah, could you remove this file in separate patch tho? Or I can do it | 10:54 |
| inc0 | anyway, I'll +1 this one | 10:56 |
| pbourke | inc0: the patch is still needed though afaik | 10:56 |
| pbourke | as it hasn't been merged in upstream horizon yet | 10:56 |
| inc0 | I think its fix released status, so it was merged | 10:57 |
| inc0 | but I'll check | 10:57 |
| coolsvap | pbourke, are you using virtualbox with vagrant? | 11:00 |
| pbourke | coolsvap: yes | 11:00 |
| coolsvap | hmm alright, i might need to tweak the vagrantfile a bit | 11:01 |
| *** sdake has quit IRC | 11:07 | |
| pbourke | has anyone come across this problem with the horizon dashboard https://i.imgur.com/KcDSzRv.png | 11:10 |
| pbourke | can't click anything, styling seems to be messed up | 11:10 |
| pbourke | ok found it, collectstatic is not being run | 11:19 |
| *** rhallisey has joined #kolla | 11:21 | |
| *** rhallisey has quit IRC | 11:27 | |
| *** rhallisey has joined #kolla | 11:27 | |
| *** jtriley has joined #kolla | 11:29 | |
| *** jtriley has quit IRC | 11:33 | |
| inc0 | anyone had any problems with neutron l3 agent? it doesn't seem to add record to db | 11:48 |
| inc0 | and server breaks | 11:48 |
| rhallisey | what's the error inc0 ? | 11:48 |
| inc0 | neutron-l3-agent is running, but table neutron/agents is empty | 11:48 |
| rhallisey | cause I'm seeing something similar with cinder | 11:49 |
| inc0 | and neutron-server breaks with "no l3 agent found" | 11:49 |
| rhallisey | ERROR cinder DBConnectionError: (OperationalError) (2002, "Can't connect to local MySQL server through socket '/var/lib/mysql/mysql.sock' (2)") None None | 11:49 |
| inc0 | problem is, it looks like it connects to db | 11:49 |
| rhallisey | ya it's weird cause cinder creates user and stuff but says in never connects | 11:50 |
| *** diogogmt has joined #kolla | 11:54 | |
| *** weiyu_ has quit IRC | 11:58 | |
| openstackgerrit | Paul Bourke proposed stackforge/kolla: Fix missing config in horizon source image https://review.openstack.org/209415 | 12:01 |
| openstackgerrit | Paul Bourke proposed stackforge/kolla: Add horizon source for oraclelinux https://review.openstack.org/209430 | 12:03 |
| openstackgerrit | Paul Bourke proposed stackforge/kolla: Add horizon source for oraclelinux https://review.openstack.org/209430 | 12:03 |
| *** weiyu_ has joined #kolla | 12:06 | |
| *** jtriley has joined #kolla | 12:08 | |
| openstackgerrit | Paul Bourke proposed stackforge/kolla: Implement Murano source images https://review.openstack.org/209905 | 12:10 |
| openstackgerrit | Paul Bourke proposed stackforge/kolla: Implement Murano source images https://review.openstack.org/209905 | 12:12 |
| *** unicell has joined #kolla | 12:20 | |
| *** unicell1 has quit IRC | 12:21 | |
| *** weiyu_ has quit IRC | 12:23 | |
| *** dims has joined #kolla | 12:24 | |
| *** Haomeng|2 has joined #kolla | 12:25 | |
| *** weiyu has joined #kolla | 12:26 | |
| *** dims_ has quit IRC | 12:27 | |
| inc0 | SamYaple, ping | 12:27 |
| *** weiyu has quit IRC | 12:27 | |
| SamYaple | pong | 12:27 |
| inc0 | did you have any problems with neutron setup on multinode? | 12:27 |
| SamYaple | define problem | 12:27 |
| inc0 | it dies without traceback, and only error I see is AgentsNotFound | 12:28 |
| SamYaple | yea its been doing that for the past few days | 12:28 |
| inc0 | L3 agent is running but it never added anything to DB | 12:28 |
| SamYaple | something upstream changed | 12:28 |
| *** Haomeng has quit IRC | 12:28 | |
| inc0 | meh | 12:28 |
| inc0 | I hate this project | 12:29 |
| rhallisey | lols | 12:29 |
| SamYaple | i hate the binary deploy | 12:29 |
| inc0 | been debugging this for last 4 hrs | 12:29 |
| SamYaple | i want the source to work but dont have time | 12:29 |
| SamYaple | from source would be so much more stable to test against | 12:29 |
| inc0 | yeah, kilo | 12:30 |
| inc0 | do they even have CI? | 12:30 |
| *** diogogmt has quit IRC | 12:31 | |
| *** weiyu has joined #kolla | 12:36 | |
| SamYaple | we are the CI | 12:37 |
| SamYaple | ! | 12:37 |
| rhallisey | ya haha | 12:37 |
| *** dwalsh has joined #kolla | 12:44 | |
| *** jtriley has quit IRC | 12:46 | |
| *** vbel has joined #kolla | 12:59 | |
| vbel | Good morning / evening everyone | 12:59 |
| SamYaple | morning vbel | 12:59 |
| vbel | https://bugs.launchpad.net/kolla/+bug/1482214 I wonder if anyone could run keystone recently | 13:00 |
| openstack | Launchpad bug 1482214 in kolla "ansible does not run bootstrap for keystone" [Undecided,New] | 13:00 |
| vbel | quite annoying because it should actually work - run_once is ignored as other flags | 13:00 |
| vbel | ansible-playbook 1.9.2 | 13:01 |
| SamYaple | vbel: i have no problem running keystone | 13:01 |
| vbel | I have also found 'detach' does not work as | 13:01 |
| SamYaple | let me doublecheck what its doing | 13:01 |
| SamYaple | vbel: there is a bug about detach | 13:02 |
| SamYaple | check bootstrap.yml | 13:02 |
| SamYaple | it should have a link | 13:02 |
| vbel | ok, so we use docker wait instead | 13:02 |
| *** athomas has quit IRC | 13:02 | |
| SamYaple | yea docker module bug upstream | 13:02 |
| SamYaple | its been fixed | 13:02 |
| vbel | bootstrap.yml: | 13:03 |
| vbel | 23 - debug: msg="starting bootstrap container" | 13:03 |
| vbel | 24 run_once: True | 13:03 |
| vbel | 25 | 13:03 |
| vbel | 26 - include: start.yml | 13:03 |
| vbel | 27 vars: | 13:03 |
| vbel | 28 run_me_once: True | 13:03 |
| vbel | 29 when: database|changed | 13:03 |
| SamYaple | vbel: "run_me_once" ? | 13:03 |
| vbel | it runs on all hosts and without bootstrapping flags - therefore no tables are inside of keystone | 13:03 |
| vbel | sorry, run_once, I am hacking a bit :) | 13:03 |
| SamYaple | oh | 13:04 |
| SamYaple | :/ | 13:04 |
| vbel | wondered if keywords should not be vars | 13:04 |
| SamYaple | well it works in the repo as ive just tested | 13:04 |
| SamYaple | you can post your hacked playbooks but i make no promises | 13:04 |
| vbel | I used kolla latest | 13:04 |
| vbel | no modifications | 13:05 |
| SamYaple | run_me_once isn't in the playbooks | 13:05 |
| vbel | - name: Creating database runs on first node | 13:05 |
| vbel | - name: Creating database user and setting permissions runs on first node | 13:06 |
| vbel | but the above runs everywhere | 13:06 |
| *** athomas has joined #kolla | 13:06 | |
| vbel | something to do with multilevel includes? | 13:07 |
| vbel | TASK: [keystone | Copying over config(s)] ************************************* | 13:08 |
| vbel | ok: [ndb01] | 13:08 |
| vbel | ok: [ndb02] | 13:08 |
| vbel | TASK: [keystone | Creating database] ****************************************** | 13:08 |
| vbel | changed: [ndb01] | 13:08 |
| vbel | TASK: [keystone | Creating database user and setting permissions] ************* | 13:08 |
| vbel | changed: [ndb01] | 13:08 |
| vbel | TASK: [keystone | Starting container(s)] ************************************** | 13:08 |
| vbel | changed: [ndb02] | 13:08 |
| vbel | changed: [ndb01] | 13:08 |
| vbel | TASK: [keystone | Starting PID container(s)] ********************************** | 13:08 |
| SamYaple | pastebin please | 13:08 |
| vbel | skipping: [ndb01] | 13:08 |
| vbel | skipping: [ndb02] | 13:08 |
| vbel | 1 sec | 13:08 |
| vbel | yep | 13:08 |
| vbel | http://paste.openstack.org/show/411166/ | 13:11 |
| SamYaple | which version of ansible | 13:12 |
| vbel | ^ with clean checkout. ansible-playbook 1.9.2 | 13:12 |
| vbel | you see that creating db and user does run once, but bootstrap run on both hosts. And more - the containers don't run bootstrapping (no db tables created) | 13:14 |
| vbel | the same with -vvvvvv (after user/db cleanup): http://paste.openstack.org/show/411170/ | 13:17 |
| SamYaple | yea i see the same thing, with mine bootstraping works, but i see it start on both nodes | 13:17 |
| vbel | I will try to debug this locally, with run_me_really_once :) | 13:19 |
| *** jpeeler has quit IRC | 13:20 | |
| SamYaple | vbel: can you file a bug? | 13:20 |
| vbel | yes - https://bugs.launchpad.net/kolla/+bug/1482214 | 13:21 |
| openstack | Launchpad bug 1482214 in kolla "ansible does not run bootstrap for keystone" [Undecided,New] | 13:21 |
| SamYaple | thanks | 13:21 |
| *** thumpba has joined #kolla | 13:23 | |
| inc0 | also, it seems that nova bootstrap runs on compute nodes as well | 13:25 |
| inc0 | isn't it? | 13:25 |
| SamYaple | its probably like that for everything inc0 | 13:26 |
| SamYaple | this isn;t related to only keystone im sure | 13:26 |
| *** thumpba_ has joined #kolla | 13:26 | |
| inc0 | it is, but I don't think that's good for this case, and no, not related to keystone | 13:26 |
| inc0 | just random useless comment, you should get used to these after spending that much time in my presence | 13:26 |
| SamYaple | well vbel i was going to remove this start.yml abstraction anyway. now is as good a time as any | 13:28 |
| vbel | thanks Sam, I will check if any workarounds are possible in the meantime | 13:29 |
| *** thumpba has quit IRC | 13:30 | |
| SamYaple | cool | 13:30 |
| SamYaple | youre probably right about the calling multiple includes | 13:30 |
| *** thumpba_ has quit IRC | 13:30 | |
| SamYaple | we could change this with pre/post tasks, but ive been trying to keep those out of the playbooks so they remain simple for people to read | 13:30 |
| *** jpeeler has joined #kolla | 13:33 | |
| *** thumpba has joined #kolla | 13:36 | |
| *** jtriley has joined #kolla | 13:41 | |
| vbel | Interesting, even if I remove run_once from start.yml and add it like below it still does not run once | 13:47 |
| vbel | 23 - include: start.yml | 13:47 |
| vbel | 24 run_once: True | 13:47 |
| vbel | 25 when: database|changed | 13:47 |
| SamYaple | you cant run_once on includes | 13:49 |
| vbel | I see | 13:50 |
| *** jtriley_ has joined #kolla | 13:55 | |
| *** sdake has joined #kolla | 13:55 | |
| sdake | morning | 13:57 |
| *** jtriley has quit IRC | 13:57 | |
| SamYaple | morning sdake | 13:57 |
| *** jtriley_ has quit IRC | 13:59 | |
| sdake | hey sam | 14:00 |
| coolsvap | morning | 14:02 |
| sdake | hey cool | 14:02 |
| openstackgerrit | Sam Yaple proposed stackforge/kolla: Updated some requirements about docker https://review.openstack.org/209722 | 14:04 |
| SamYaple | sdake: https://review.openstack.org/#/c/208451/ | 14:04 |
| SamYaple | its ready to try | 14:04 |
| vbel | SamYaple: it is not only run_once, also when is ignored: | 14:09 |
| vbel | 23 - include: start.yml | 14:09 |
| vbel | 24 vars: | 14:09 |
| vbel | 25 run_once: True | 14:09 |
| vbel | 26 when: inventory_hostname == "xyz" | 14:09 |
| coolsvap | harmw, around? | 14:09 |
| SamYaple | vbel: no when works fine | 14:09 |
| SamYaple | theres probably something else going on there | 14:10 |
| SamYaple | i mean when works fine with includes in geeneral | 14:10 |
| SamYaple | it may be broken there | 14:10 |
| vbel | SamYaple: yes, ignore the sample above | 14:12 |
| *** sdake has quit IRC | 14:14 | |
| SamYaple | oh jeez | 14:14 |
| SamYaple | in openstack-meeting-4 they are tlaking abou splitting out glance into microservices | 14:14 |
| SamYaple | :/ | 14:14 |
| rhallisey | oh rlly | 14:19 |
| *** Ti-mo- has quit IRC | 14:23 | |
| *** Ti-mo has joined #kolla | 14:24 | |
| *** dims has quit IRC | 14:25 | |
| *** dims has joined #kolla | 14:26 | |
| *** weiyu has quit IRC | 14:33 | |
| *** jtriley has joined #kolla | 14:36 | |
| *** diogogmt has joined #kolla | 14:38 | |
| *** weiyu has joined #kolla | 14:38 | |
| *** weiyu has quit IRC | 14:41 | |
| *** rhallisey has quit IRC | 14:42 | |
| SamYaple | well you know they just split out glance_store as a seperate library | 14:50 |
| *** banix has joined #kolla | 14:55 | |
| *** achanda has joined #kolla | 14:59 | |
| *** inc0 has quit IRC | 15:06 | |
| *** achanda has quit IRC | 15:09 | |
| openstackgerrit | Paul Bourke proposed stackforge/kolla: Add a data vol image for the arista Neutron plugin https://review.openstack.org/207545 | 15:11 |
| *** jtriley has quit IRC | 15:11 | |
| pbourke | anyone using horizon lately? | 15:12 |
| pbourke | https://bugs.launchpad.net/kolla/+bug/1482196 | 15:12 |
| openstack | Launchpad bug 1482196 in kolla "Horizon dashboard not loading properly, can't click on anything" [Undecided,New] | 15:12 |
| pbourke | if anyone knows even what horizon.auth is or does it would help | 15:13 |
| SamYaple | sounds like it provied auth for horizon | 15:14 |
| * SamYaple is being helpful | 15:14 | |
| *** jtriley has joined #kolla | 15:15 | |
| pbourke | ah, it was so simple! | 15:15 |
| *** kevsi has joined #kolla | 15:16 | |
| *** diogogmt has quit IRC | 15:16 | |
| *** diogogmt has joined #kolla | 15:17 | |
| *** rhallisey has joined #kolla | 15:18 | |
| vbel | SamYaple: http://paste.openstack.org/show/411194/ - this use case show that var is known but does not go into final run_once | 15:19 |
| SamYaple | yea vbel i think its because we are already running on an included playbook | 15:20 |
| *** blahRus has joined #kolla | 15:24 | |
| vbel | SamYaple: you know what is my workaround? put 'run_once: True' in service related bootstrap.yml :) | 15:26 |
| vbel | SamYaple: wait a sec, retry | 15:27 |
| vbel | SamYaple: sorry, wrong test | 15:31 |
| *** sdake has joined #kolla | 15:31 | |
| *** daneyon has quit IRC | 15:32 | |
| *** daneyon has joined #kolla | 15:33 | |
| *** dwalsh has quit IRC | 15:33 | |
| SamYaple | the start.yml is going to continue to be a problem | 15:34 |
| SamYaple | i dont see any choice but to do it "right" and duplicate that docker module block | 15:34 |
| SamYaple | other than the duplicate code, is there any objections? | 15:34 |
| SamYaple | pbourke vbel ^ | 15:35 |
| vbel | nope | 15:35 |
| pbourke | i haven't been following this issue closely enough | 15:35 |
| SamYaple | pbourke: just in geeneral | 15:35 |
| SamYaple | remove the abstraction layer around start.yml | 15:35 |
| SamYaple | basically call the docker module without having the start.yml include | 15:35 |
| vbel | so at least we can have "starting keystone" instead of double "starting container(s)" | 15:36 |
| SamYaple | yea | 15:36 |
| SamYaple | that i hated | 15:36 |
| vbel | +2 | 15:36 |
| SamYaple | i mean this was always the way it had to go, but i couldnt think of an easy way to abstract | 15:36 |
| SamYaple | im going to hit up #ansible for a minute to ask | 15:37 |
| vbel | you can reference that paste from me | 15:37 |
| pbourke | im looking at start.yml and can't see the difference between the two blocks | 15:37 |
| vbel | with host pid and without | 15:37 |
| vbel | ? | 15:38 |
| SamYaple | pbourke: dont worry about that | 15:38 |
| SamYaple | i just mean in geeneral | 15:38 |
| SamYaple | we are doing wierd ansible things there | 15:38 |
| sdake | i really dislike the "starting container" skipoped thing | 15:38 |
| pbourke | let me read the bug and scroll back | 15:39 |
| SamYaple | so we are agreeing to duplicate teh docker block? | 15:39 |
| sdake | i'm not familiar with the dockerblock in question | 15:39 |
| pbourke | I understand about 25% of what you guys are talking about :p | 15:39 |
| vbel | since we duplicate it anyway by passing quite some parameters | 15:39 |
| SamYaple | yea a little bit | 15:39 |
| sdake | are you talking about container start duplicatd twice? | 15:40 |
| SamYaple | sdake: kinda, inderectly | 15:40 |
| sdake | the reaosn is to get pid=host - unfortunately that is necessary | 15:40 |
| SamYaple | basically instead of calling ../../start.yml im just going to switch it to the dockre module directly | 15:40 |
| sdake | another way to do it is specify container_pid in every module | 15:41 |
| SamYaple | so we duplicate 10 params each time | 15:41 |
| sdake | right | 15:42 |
| sdake | its pretty terrible, the biggest problem with it is it starts every container twice | 15:42 |
| sdake | or says starting container skipped rather | 15:43 |
| vbel | if it is pid: "host" then we could combine blocks with pid: "{{ 'host' if container_pid else '' }}" | 15:49 |
| pbourke | vbel: can you explain the purpose of the two blocks in start.yml | 15:51 |
| pbourke | im not seeing it | 15:51 |
| *** dwalsh has joined #kolla | 15:52 | |
| vbel | I think it is to run one if variable container_pid is set or run another if the variable is not set. Both differ in pid: 'host' | 15:52 |
| pbourke | ah yeah its for libvirt | 15:53 |
| pbourke | and the two "starting containers" is one for bootstrap and one for the main container right? | 15:53 |
| vbel | I would try to combine it with pid: "{{ 'host' if container_pid else '' }}" | 15:54 |
| vbel | no, they run both but one is skipped :) | 15:54 |
| pbourke | I see | 15:55 |
| pbourke | how will moving this block into each role fix the issue | 15:55 |
| SamYaple | vbel: the issue is in the docker module | 15:56 |
| SamYaple | it doesnt accept '' as valid | 15:56 |
| vbel | if we provide pid: "{{ 'host' if container_pid else var_that_is_None }}" | 15:59 |
| vbel | it is None by default, wonder how to tell ansible it | 15:59 |
| pbourke | trying to find the source for the ansible module | 16:04 |
| pbourke | try null? | 16:07 |
| pbourke | https://groups.google.com/forum/#!topic/ansible-project/pfaxQNPpst4 | 16:07 |
| vbel | anyway it does not help with run_once | 16:10 |
| *** echoingumesh has joined #kolla | 16:13 | |
| *** sdake has quit IRC | 16:14 | |
| pbourke | vbel: do we know why its ignoring run_once | 16:15 |
| SamYaple | vbel: you cant | 16:16 |
| SamYaple | ansible wont eval it as a bool | 16:16 |
| SamYaple | or none | 16:16 |
| SamYaple | you have to set it to None in the module | 16:16 |
| SamYaple | its fixed upstream | 16:16 |
| SamYaple | ill just do the work or duplicating the docker module | 16:17 |
| SamYaple | there really is no better solution | 16:17 |
| SamYaple | at least taht i can think of i mean | 16:17 |
| *** SourabhP has joined #kolla | 16:18 | |
| *** unicell has quit IRC | 16:18 | |
| *** unicell has joined #kolla | 16:19 | |
| *** unicell has quit IRC | 16:20 | |
| vbel | I have submitted a bug https://github.com/ansible/ansible/issues/11876 for run_once | 16:20 |
| openstackgerrit | Merged stackforge/kolla: Update the documentation for Ansible https://review.openstack.org/209474 | 16:22 |
| *** dwalsh has quit IRC | 16:22 | |
| harmw | coolsvap: you have 1 minute :) | 16:22 |
| *** SourabhP_ has joined #kolla | 16:22 | |
| vbel | I would put 'starting non-PID containters' and 'starting PID containers' to avoid further questions when running this double block :) | 16:23 |
| SamYaple | vbel: im ust removing all of that | 16:23 |
| *** SourabhP has quit IRC | 16:24 | |
| *** SourabhP_ is now known as SourabhP | 16:24 | |
| vbel | ok | 16:24 |
| *** dwalsh has joined #kolla | 16:24 | |
| *** echoingumesh has quit IRC | 16:25 | |
| SamYaple | the code would be alot cleaner if i removed the ability to specify a private registry per service, but rather do it on a global scale | 16:25 |
| SamYaple | will anyone actually do that you think? | 16:25 |
| SamYaple | eh ill not be lazy and keep it | 16:27 |
| pbourke | I can't ever see it being used | 16:27 |
| openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Change build.py to match the release https://review.openstack.org/210015 | 16:27 |
| vbel | per service? yes, we had such use case, e.g. running some kollaglue containers with ours if something did not work :) But such hacks can be done locally anyway, so I would clean up too | 16:30 |
| *** SourabhP has quit IRC | 16:30 | |
| SamYaple | you think? so youre saying if someone _really_ wants to do it they should tweak the playbooks? | 16:30 |
| vbel | yes | 16:31 |
| SamYaple | fair enough | 16:31 |
| vbel | they would be 'developers' | 16:31 |
| SamYaple | i was thinking, business uses the kollaclue containers _except_ for the keystone contaienr where they do something different | 16:31 |
| SamYaple | but i suppose at that point they should probably role thier own everything | 16:31 |
| SamYaple | roll* | 16:31 |
| rhallisey | SamYaple, would you say master==latest | 16:35 |
| rhallisey | I'll just tag back to latest and move ansible to latest | 16:35 |
| SamYaple | yes, but having a latest tag is a bad thing for our project | 16:35 |
| SamYaple | we discussed this a bit at the midcycle and everyone agreed latest caused confusion | 16:35 |
| rhallisey | well technically when you're building locally it is latest | 16:35 |
| SamYaple | and was mostly broken | 16:35 |
| rhallisey | well pushing latest | 16:36 |
| SamYaple | right, but the people who git clone will end up pulling from a broken image | 16:36 |
| rhallisey | I say we just don't push the latest tag | 16:36 |
| rhallisey | given that discussion | 16:36 |
| SamYaple | if we untag latest in dockerhub im ok with that decision | 16:36 |
| rhallisey | 'latest' would be defined as you just built it | 16:37 |
| SamYaple | agreed | 16:37 |
| rhallisey | pulls will be the 'latest' release | 16:37 |
| rhallisey | but not tagged as latest | 16:37 |
| rhallisey | ok | 16:37 |
| *** dwalsh has quit IRC | 16:37 | |
| SamYaple | right i would think we would have a "liberty" tag that would be the latest liberty tested image | 16:37 |
| SamYaple | same for the other branches | 16:38 |
| rhallisey | kk | 16:38 |
| rhallisey | well the problem with defaulting ansible to latest is then it won't work outa the box | 16:41 |
| rhallisey | ehh.. | 16:41 |
| SamYaple | it will on a tagged release | 16:41 |
| rhallisey | may just need a doc | 16:41 |
| SamYaple | it _cant_ work out of the box for master | 16:41 |
| SamYaple | but a tagged release will be fine | 16:42 |
| SamYaple | and so will stable branches | 16:42 |
| rhallisey | ok I'll leave it as is then | 16:42 |
| SamYaple | no it should be changed to latest again | 16:42 |
| SamYaple | the liberty2 iamges cant work with teh master playbooks | 16:42 |
| rhallisey | ya that's true | 16:42 |
| SamYaple | but you switch to the liberty2 tag and itll be fine | 16:42 |
| SamYaple | tagged git == tagged specifed | 16:43 |
| SamYaple | master git == latest tag | 16:43 |
| SamYaple | so we switch the tags, release, then switch back | 16:43 |
| *** unicell has joined #kolla | 16:43 | |
| rhallisey | kk makes sense | 16:43 |
| SamYaple | stable will alway be pointed at a stable tag | 16:43 |
| SamYaple | always* | 16:43 |
| openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Change ansible to search for latest tag https://review.openstack.org/210015 | 16:45 |
| SamYaple | eh for readability and expected results lets just get rid of this abstractedness for the playbooks | 16:45 |
| SamYaple | i hate violating dry but idk | 16:46 |
| SamYaple | overall i think it will just be a blemish on the playbooks to do the abstraction. everyone else just deals with the DRY violation and thats ansible "best practice" | 16:47 |
| SamYaple | I tried to do something here and i dont think its working out well. ill just do it the long way | 16:48 |
| rhallisey | SamYaple, don't understand your comment | 16:49 |
| SamYaple | rhallisey: just revert the commit that changed the tag in the first place | 16:49 |
| SamYaple | i also updated all the compose file tags as well | 16:50 |
| SamYaple | just do `git revert f4d6fc6193b0d0a981433ad849c614130a13041a` | 16:50 |
| SamYaple | https://github.com/stackforge/kolla/commit/f4d6fc6193b0d0a981433ad849c614130a13041a | 16:50 |
| rhallisey | oh | 16:51 |
| openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Revert "Merge "Updated all tags for Liberty2"" https://review.openstack.org/210027 | 16:57 |
| rhallisey | ok that patch will revert it | 16:58 |
| openstackgerrit | Sam Yaple proposed stackforge/kolla: Removed bootstrap.yml and start.yml abstraction https://review.openstack.org/210041 | 17:16 |
| SamYaple | pbourke: vbel rhallisey https://review.openstack.org/#/c/210041/ | 17:17 |
| SamYaple | thats just for keystone, if everyone is ok with that change i will roll it out everywhere | 17:17 |
| rhallisey | interesting | 17:18 |
| SamYaple | I was just trying to avoid specifying those 10 docker options over and over again, but it is what it is | 17:19 |
| SamYaple | i mean they are variabled out, so the enduser won't have to specify it 30 times | 17:19 |
| rhallisey | sam can you review again https://review.openstack.org/#/c/210015/ | 17:21 |
| SamYaple | wasnt that covered in the revert? | 17:22 |
| SamYaple | oh i suppose not | 17:22 |
| *** achanda has joined #kolla | 17:30 | |
| *** sdake has joined #kolla | 17:33 | |
| *** tfukushima has quit IRC | 17:36 | |
| *** tfukushima has joined #kolla | 17:37 | |
| *** tfukushima has quit IRC | 17:39 | |
| rhallisey | jpeeler, can you ack https://review.openstack.org/#/c/210015/ | 17:41 |
| rhallisey | 1 lane change super easy | 17:41 |
| jpeeler | done | 17:43 |
| rhallisey | thx | 17:43 |
| sdake | morning again guys | 17:50 |
| *** thumpba has quit IRC | 17:54 | |
| rhallisey | morning | 17:56 |
| *** thumpba has joined #kolla | 17:57 | |
| *** echoingumesh has joined #kolla | 17:58 | |
| *** Haomeng has joined #kolla | 18:00 | |
| *** Haomeng|2 has quit IRC | 18:03 | |
| *** tfukushima has joined #kolla | 18:04 | |
| openstackgerrit | Merged stackforge/kolla: Change ansible to search for latest tag https://review.openstack.org/210015 | 18:06 |
| *** thumpba_ has joined #kolla | 18:14 | |
| openstackgerrit | Merged stackforge/kolla: Revert "Merge "Updated all tags for Liberty2"" https://review.openstack.org/210027 | 18:15 |
| *** dwalsh has joined #kolla | 18:16 | |
| *** thumpba has quit IRC | 18:17 | |
| SamYaple | hey guys, im goign to bed | 18:18 |
| SamYaple | would really appreciate reviews on https://review.openstack.org/#/c/208451/ and https://review.openstack.org/#/c/210041/ | 18:18 |
| *** daneyon has quit IRC | 18:20 | |
| *** daneyon has joined #kolla | 18:20 | |
| *** jasonsb has quit IRC | 18:40 | |
| *** jasonsb has joined #kolla | 18:40 | |
| *** jasonsb has quit IRC | 18:45 | |
| openstackgerrit | Merged stackforge/kolla: Reduce dependencies on the dest nodes https://review.openstack.org/208451 | 18:45 |
| *** dwalsh has quit IRC | 18:55 | |
| *** tfukushima has quit IRC | 19:04 | |
| *** achanda has quit IRC | 19:05 | |
| *** diga has quit IRC | 19:05 | |
| *** achanda has joined #kolla | 19:05 | |
| *** diga has joined #kolla | 19:06 | |
| *** dwalsh has joined #kolla | 19:10 | |
| jpeeler | according to this blog post: https://labs.ctl.io/docker-hub-top-10/ , we're fourth in rank for number of repos (really should be images i guess) on docker hub | 19:18 |
| sdake | rockin ;-) | 19:20 |
| rhallisey | cool | 19:21 |
| sdake | thats pretty funny actually :) | 19:21 |
| *** jasonsb has joined #kolla | 19:22 | |
| harmw | you see... this is why we should've named our repo "sdakeshouseofpleasure" | 19:22 |
| rhallisey | lol | 19:22 |
| bmace | lol | 19:23 |
| jpeeler | does anybody know if we actually need a network node for multi-node deployment? | 19:32 |
| jpeeler | this makes me think yes, but i still know nothing about ansible: https://github.com/stackforge/kolla/blob/master/ansible/inventory/multinode#L13 | 19:32 |
| sdake | network node no | 19:33 |
| sdake | each node networks (with openvswitch) | 19:34 |
| sdake | for linuxbridgey ou may nee a network node | 19:34 |
| *** thumpba_ has quit IRC | 19:34 | |
| sdake | i'd justrecommend using openvswitch if you dont have a strong preference - then it should just work multinode correctly | 19:34 |
| *** thumpba has joined #kolla | 19:35 | |
| jpeeler | ansible seems to query it | 19:39 |
| jpeeler | and i'm still getting this eth0 error from yesterday | 19:39 |
| harmw | lol, $job is featured as a AWS case study with some video interview | 19:40 |
| harmw | cool stuff | 19:41 |
| *** jruano has joined #kolla | 19:41 | |
| harmw | jpeeler: ansible is just able to assign one node with the networking bits (server), it's not required | 19:41 |
| harmw | afaik | 19:41 |
| harmw | sdake: something interesting came out of yesterday's meeting? | 19:42 |
| jpeeler | harmw: i'm actually using your vagrant setup - the intention is to execute ansible from the operator node, right? | 19:42 |
| harmw | yep, just ssh into the operator node and run ansible | 19:42 |
| jpeeler | i'm getting this strange error: fatal: [network01] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute u'ansible_eth0'", 'failed': True} | 19:43 |
| harmw | hehe | 19:43 |
| harmw | you do need to edit the Ansible config :) | 19:43 |
| harmw | you're running Fedora in the Vagrant vms | 19:43 |
| harmw | which is not booted with a parameter to setup oldskool interface naming, so it uses enp0s3 or something similar | 19:44 |
| harmw | and you should configure that in etc/kolla/defaults.yml or general.yml | 19:44 |
| jpeeler | i thought your setup also booted centos, but i definitely switched it to centos | 19:44 |
| jpeeler | so eth0 is there | 19:44 |
| harmw | oh ok | 19:44 |
| harmw | fair enough | 19:44 |
| harmw | are you setting the interface name correctly? | 19:45 |
| jpeeler | in the networking options? | 19:46 |
| harmw | yea | 19:46 |
| jpeeler | they seemed sane, i haven't changed them | 19:46 |
| jpeeler | harmw: what does the net interfaces look like for your operator node? | 19:48 |
| *** thumpba has quit IRC | 19:48 | |
| harmw | operator node isn't interesting, it's the compute and support nodes your interested in | 19:48 |
| harmw | thats where all the fun happens :) | 19:48 |
| jpeeler | well they are all configured the same i thought | 19:48 |
| *** thumpba has joined #kolla | 19:48 | |
| harmw | enp0s8 and enp0s9 is what I have, on all nodes | 19:48 |
| harmw | yep | 19:49 |
| jpeeler | i have eth0, eth1, and eth2 | 19:50 |
| jpeeler | eth2 the private interface | 19:50 |
| harmw | enp0s3 is there aswell, so that makes 3 | 19:51 |
| jpeeler | and eth0 and eth1 connected to the bridge. does that sound right? i'm doing this using the libvirt provider, hence the difference | 19:51 |
| harmw | eth0 should be hostonly, so 10.something | 19:51 |
| harmw | eth1 will be the private network shared with all vagrant virtual machines | 19:52 |
| harmw | and eth2 should be bridged to the public network | 19:52 |
| harmw | (I'm using virtualbox btw) | 19:52 |
| jpeeler | networking is not my forte | 19:53 |
| jpeeler | i have the private address on eth2, know that for sure | 19:53 |
| harmw | ah | 19:53 |
| *** thumpba has quit IRC | 19:53 | |
| jpeeler | i think eth0 and eth1 are both bridged in the same way | 19:53 |
| harmw | lets see | 19:54 |
| harmw | which address is assigned to each one of those interfaces? | 19:54 |
| jpeeler | http://paste.fedoraproject.org/252420/43889085/ | 19:54 |
| jpeeler | i did that on the operator node, but the others are similar | 19:54 |
| harmw | eth0 and eth1 are connected to the same network? | 19:54 |
| harmw | and the address on eth2, where'd that come from? | 19:55 |
| *** thumpba has joined #kolla | 19:55 | |
| harmw | or rather, those, since there are 2 addresses :) | 19:55 |
| jpeeler | harmw: yeah, not sure why there are two addresses... but it was manually assigned as i couldn't find any libvirt equivalent to dhcp there | 19:56 |
| *** thumpba has quit IRC | 19:56 | |
| jpeeler | harmw: want to see the Vagrantfile or diff? | 19:56 |
| *** thumpba has joined #kolla | 19:57 | |
| harmw | hm | 19:57 |
| harmw | the point of one of those 3 interfaces it doesn't need an IP, it's just a bridge to get out to the (your) public net on which you can plant your floating ips | 19:57 |
| harmw | so all vagrant nodes are using static ips? | 19:58 |
| harmw | vagrant ssh operator | 19:58 |
| harmw | thats how you'd get in to the operator node, it should be able to tell you on which IP it's connected | 19:58 |
| harmw | so that way you can find out which interface is the hostonly one | 19:59 |
| jpeeler | vagrant ssh operator actually uses eth0 | 19:59 |
| harmw | cool | 19:59 |
| jpeeler | 192.168.121.232 | 19:59 |
| jpeeler | is that right? | 19:59 |
| harmw | sure, probably :) | 19:59 |
| harmw | mine is using a different range, but thats dhcp | 19:59 |
| harmw | and mostly Virtualbox magic | 19:59 |
| jpeeler | well the 192.168.121.232 is dhcp libvirt stuff | 20:00 |
| harmw | excellent | 20:00 |
| harmw | so eth0 is getting an IP from dhcp? | 20:00 |
| jpeeler | yes, along with eth1 too | 20:00 |
| harmw | ok | 20:00 |
| harmw | eth2 is probably your bridged interface, not needing an address | 20:01 |
| *** jruano has quit IRC | 20:01 | |
| jpeeler | it's assigned directly though | 20:01 |
| harmw | hm | 20:01 |
| jpeeler | if this helps, http://paste.fedoraproject.org/252422/91038143/ | 20:01 |
| jpeeler | i tried to replicate what you did | 20:01 |
| * jpeeler thought this would be a simple image replacement | 20:02 | |
| jpeeler | not the case | 20:02 |
| jpeeler | diff if that's helpful http://paste.fedoraproject.org/252423/88910581/ | 20:02 |
| *** dwalsh has quit IRC | 20:04 | |
| harmw | your not building those network nodes properly, that looks like it won't achieve what you are after | 20:04 |
| harmw | am I missing the networking stuff for networks 2 and 3? | 20:04 |
| jpeeler | i thought only one node was needed? i only created it because the ansible multihost ping reported it missing | 20:05 |
| harmw | depends, it's currently looking for it since it's specified in multihost inventory | 20:06 |
| harmw | but you could change that | 20:06 |
| harmw | anyway, I only see one network created | 20:06 |
| harmw | and it's public | 20:06 |
| jpeeler | i know i'm network stupid, but i'm not seeing the difference from what you did | 20:08 |
| harmw | np | 20:08 |
| harmw | you posted that diff | 20:09 |
| harmw | L38 | 20:09 |
| harmw | sets up a network | 20:09 |
| harmw | right | 20:09 |
| harmw | and L32+33 disable two networks, right? | 20:09 |
| jpeeler | i disabled the virtual box way | 20:10 |
| jpeeler | i tried to replicate the private ip assignment in each node config | 20:10 |
| jpeeler | for example, line 92 | 20:10 |
| harmw | I saw that, but is that creating networks in the way we want them? | 20:11 |
| jpeeler | probably not... | 20:11 |
| harmw | I don't know Vagrant that much :) (and prefer building such a thing upfront) | 20:11 |
| harmw | you don't happen to have access to virtualbox do you :p | 20:11 |
| *** achanda_ has joined #kolla | 20:12 | |
| *** achanda has quit IRC | 20:12 | |
| jpeeler | well i mean, i was hoping to get this to work on libvirt. it's the fedora way | 20:12 |
| *** thumpba_ has joined #kolla | 20:12 | |
| harmw | yea well, libvirt can't bridge on wireless so I sadly choose vbox :( | 20:12 |
| harmw | but can't you duplicate L38(+) to create those other 2 networks there? | 20:13 |
| jpeeler | i think i could, would that work? | 20:14 |
| harmw | give it a go, it'll atleast match the virtualbox stuff :) | 20:15 |
| jpeeler | ok thanks for looking at it. it'd really rock to get this going | 20:16 |
| *** thumpba has quit IRC | 20:16 | |
| harmw | sure np, let me know if I can be of any more help :) | 20:18 |
| sdake | samyaple you around? | 20:19 |
| jpeeler | bleh private_network seems to require ip assignment. might have to shelve this for now | 20:22 |
| harmw | jpeeler: not if you configure it as dhcp, as is currently done | 20:22 |
| harmw | would that work? | 20:22 |
| jpeeler | i don't think the libvirt provider plugin supports that option | 20:22 |
| sdake | hey jpeeler you ar estuck on ansible eth0? | 20:23 |
| jpeeler | which is the problem | 20:23 |
| jpeeler | sdake: yes, but it's probably related to my vagrant networking setup | 20:23 |
| sdake | The reason that error would occur is becasue eth0 is not your interface name on your control node | 20:23 |
| sdake | ok cool | 20:23 |
| *** dims_ has joined #kolla | 20:23 | |
| harmw | ah, yes, that was what got this started in the first place :p | 20:23 |
| jpeeler | harmw: https://github.com/pradels/vagrant-libvirt#networks | 20:24 |
| *** dims_ has quit IRC | 20:24 | |
| openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Glance api and Glace registy shoudn't share tags https://review.openstack.org/210109 | 20:24 |
| openstackgerrit | Ryan Hallisey proposed stackforge/kolla: Glance api and Glance registry shouldn't share tags https://review.openstack.org/210109 | 20:24 |
| rhallisey | lol that spelling | 20:24 |
| openstackgerrit | Merged stackforge/kolla: Fedora - Install from source for Gnocchi https://review.openstack.org/208795 | 20:25 |
| openstackgerrit | Merged stackforge/kolla: Fedora - Install from source for Swift https://review.openstack.org/208789 | 20:25 |
| openstackgerrit | Merged stackforge/kolla: Adds push to build script https://review.openstack.org/208266 | 20:25 |
| *** dims_ has joined #kolla | 20:25 | |
| *** dims has quit IRC | 20:26 | |
| *** echoingumesh has quit IRC | 20:27 | |
| *** echoingumesh has joined #kolla | 20:28 | |
| *** echoingumesh has quit IRC | 20:32 | |
| openstackgerrit | Merged stackforge/kolla: Updated some requirements about docker https://review.openstack.org/209722 | 20:37 |
| harmw | rhallisey: is there even a docker_glance_registry_tag set? | 20:37 |
| rhallisey | ya | 20:37 |
| rhallisey | the line above it | 20:38 |
| harmw | oh yes | 20:38 |
| harmw | yep | 20:38 |
| openstackgerrit | Merged stackforge/kolla: Fedora - Install from source for Designate https://review.openstack.org/209336 | 20:38 |
| harmw | just noticed | 20:38 |
| harmw | damn :P | 20:38 |
| openstackgerrit | Merged stackforge/kolla: Ceilometer config-ceilometer.sh file without backslash https://review.openstack.org/209473 | 20:38 |
| rhallisey | harmw, no worries | 20:38 |
| rhallisey | I though the same thing when I saw it | 20:38 |
| rhallisey | I was like maybe it's suppose to be that way | 20:38 |
| harmw | it's not broken because both got the same value, oh well | 20:39 |
| harmw | nice catch :) | 20:39 |
| sdake | bmace you around | 20:39 |
| rhallisey | harmw, ya it works. Thanks :) | 20:39 |
| harmw | btw, what's you guys' oppinion about curling stuff and piping to bash? | 20:40 |
| harmw | re my comment in https://review.openstack.org/209722 | 20:40 |
| harmw | srw-rw---- 1 root root 0 6 aug 20:14 /var/run/docker.sock | 20:41 |
| harmw | wtf | 20:41 |
| harmw | docker should be doing root:docker per default | 20:41 |
| sdake | harmw tb fair those are docker instructions for intalling their packages | 20:41 |
| sdake | but what we should do in that document is link to the upstream docker install directions | 20:41 |
| harmw | I know they are sdake | 20:41 |
| bmace | sdake here | 20:41 |
| sdake | bmace i've been thinking about ha of the deploy node forabout 45 minutes | 20:41 |
| harmw | I'm more about the general trend to do that more, or so it feels | 20:42 |
| sdake | I wanted to run my thoughts by you | 20:42 |
| bmace | sdake sure | 20:42 |
| sdake | so basically to do ha we need one storage environment for the configuration files | 20:42 |
| sdake | and then when ansible runs it would pull the configuration files from some central store | 20:43 |
| sdake | so kolla ansible would basically be stateless except for the database or etcd or whatever | 20:43 |
| bmace | well, that or replicate the config data, sure | 20:43 |
| sdake | yes the config files must be replicated, but I dont think we want to do that directly in kolla | 20:43 |
| sdake | we want kolla to read from some third party data storage tool the config | 20:44 |
| sdake | and offer an opportnity to load new config stuff in | 20:44 |
| sdake | doing replication correctly is very hard, although in kolla's case it could be versioned so it could be easy | 20:45 |
| bmace | it would be nice to allow for the properties to be read in a variety of ways, sure. | 20:45 |
| harmw | jpeeler: is your dockerd running with proper permissions on the .sock? | 20:46 |
| *** echoingumesh has joined #kolla | 20:46 | |
| jpeeler | harmw: where? | 20:46 |
| bmace | there are varied degrees of complexity of ha. simple active / passive can also do very simple file replication / sync, though it does end up being sort of an odd thing for kolla to worry about directly. | 20:46 |
| harmw | operator | 20:46 |
| harmw | I didn't check the others yet | 20:47 |
| sdake | i'm thinking the anible modules instead of reading files could read from a replicated data store | 20:47 |
| bmace | lots of environments may be using etcd or some other replicated store and it would be great to take advantage of that. | 20:47 |
| jpeeler | harmw: i actually destroyed everything hoping i could recreate with a new private network config, but i didn't see a way to change from what i had | 20:48 |
| bmace | could what they read from be configurable? then if you want simple files, have at it, if you have some other mechanism, you can always write your own sort of interface between that service and how we do the file reads? | 20:48 |
| jpeeler | harmw: you think docker would be related to the ansible error i was getting? | 20:48 |
| harmw | bmace: that sort of thing does drag in several dependencies though, like having a zookeeper+etcd or something similar containers | 20:48 |
| harmw | (not that I don't like the idea of a key-value store containers to Kolla btw) | 20:48 |
| harmw | jpeeler: nah, it was ansible complaining | 20:49 |
| bmace | that is why i'm suggesting, especially for the ansible management nodes, leave it variable / at the distro or users discretion. | 20:49 |
| harmw | I wanted to launch vagrant myself, now running into this | 20:49 |
| harmw | bmace: yea, sounds like how we'd want to do that (for now though, haven't given it much thought yet) | 20:49 |
| sdake | i dont think we want zookeeper | 20:50 |
| harmw | jpeeler: but docker is running? | 20:50 |
| sdake | but etcd is dependency free | 20:50 |
| harmw | jpeeler: forget that, I meant is it accessible with the vagrant accuont :) | 20:51 |
| bmace | just because there is a lot of different technology in different environments i think it is odd for us to pick one. if we want to have sort of pre-built code to handle etcd and just flat files, that is probably good, as long as it isn't too onerous for people to contribute code to use other technologies. | 20:51 |
| sdake | so the way the file merging works today is it reads from the filesysstem | 20:51 |
| sdake | this code is a python module | 20:51 |
| jpeeler | harmw: i do believe it was, but i'll have to redeploy to further debug. do you think the networking options need changing or no? i couldn't change the private network to not be assigned by IP | 20:53 |
| sdake | bmace: merge-configs https://github.com/stackforge/kolla/blob/master/ansible/library/merge_configs.py | 20:53 |
| *** dwalsh has joined #kolla | 20:53 | |
| harmw | jpeeler: weird, tthough it feels like it's mostly libvirt related... | 20:53 |
| harmw | is there no libvirt-specific doc on Vagrant? | 20:53 |
| jpeeler | the github repo i linked earlier are the only docs i could find | 20:54 |
| harmw | ah, yes, Ill look into that | 20:54 |
| bmace | sdake: yup, i have looked at that code before. so it seems possible to support some other file read mechanism. I don't know where ConfigParser comes from. it takes the file path so that would need to handle something else, or before the ConfigParser takes it, we could pull it out of some other storage medium and dump it into a normal file. | 20:55 |
| sdake | well hte merging onto the local disk is fine | 20:56 |
| sdake | its the reading that needs to come from a remote storage environment | 20:56 |
| bmace | the comments in there seem off? isn't if this is the config to be pulled into the container the destination should be the /opt/kolla/blah (or i hear that is being possibly changed to /var/lib/kolla/blah) | 20:58 |
| sdake | this file merges a bunch of config files from /etc into /opt/kolla | 20:59 |
| sdake | the container then loads that /opt/kollaconfig/file.ini into the container | 21:00 |
| sdake | so instead of merging a bunch of config files on hard disk it could do it from etc or db | 21:00 |
| sdake | i'm not sure if this module could be parameterized to different backends | 21:00 |
| sdake | but get the concern about wanting to support something besides just etcd | 21:00 |
| bmace | right, which is why the docs are a bit off, for our environment | 21:00 |
| harmw | jpeeler: can you verify if docker is accessible on lets say the operator node? | 21:01 |
| bmace | well, it isn't this file that does the real read, the file path / name of the source files is passed to the ConfigParser, which is why i suggest either ConfigParser would need to change, or we would need to have something in the middle, before the ConfigParser code, that pulls down the data from the alternate data source and drops it into files, in /tmp for example, and cleans them up after | 21:01 |
| *** jtriley has quit IRC | 21:02 | |
| sdake | well lets say its path/file like /etc/kolla/nova.conf | 21:02 |
| sdake | if your data store is configured for etcd, the merge config would go read nova.conf rom /etc/kolla/nova.conf | 21:03 |
| *** britthou_ has quit IRC | 21:05 | |
| bmace | which wouldn't work terribly well, since it needs to talk to etcd to read the data? :) | 21:05 |
| sdake | sorry read from etcd://etc/kolla/nova.conf | 21:05 |
| openstackgerrit | Harm Weites proposed stackforge/kolla: Start docker without -H to make chgrp work on socket https://review.openstack.org/210122 | 21:06 |
| *** rhallisey has quit IRC | 21:06 | |
| harmw | ^ Vagrant specific guys | 21:07 |
| *** thumpba_ has quit IRC | 21:08 | |
| sdake | bmace so think of the merge as haveing a file:// processor or etcd:// processor or mysqld:// processor | 21:11 |
| sdake | when the files get built the source data store is passed in | 21:12 |
| sdake | rather hwen merge gets called | 21:12 |
| jpeeler | harmw: redeploying, but i'm pretty sure the docker permissions were fine | 21:15 |
| bmace | sdake: that is fine, regardless of the path it still looks like it is config.read() that reads in the source files to merge them.. so either ConfigParser handles those protocols to find / read them or they need to be dumped into regular files before hand. | 21:18 |
| sdake | that alone nearly gives us full ha | 21:20 |
| sdake | because we kick ha problem to the data storage mechanism | 21:20 |
| sdake | and etcd and mysql both have solutiosn for that | 21:20 |
| jpeeler | harmw: so on support01, docker ps works fine and such | 21:20 |
| bmace | yup, pick your own ha is nice. | 21:21 |
| sdake | the only other thing we need to ensure is that multiple ansible deploys don't trigger at the same time | 21:21 |
| bmace | whatever you are using in house? either we have code for it already or just add a little yourself and poof | 21:21 |
| sdake | bmace I do think it means we need to possibly figure out how to containerize etcd and ansiblize deploy it in ha mode so its easy for users to use atleast 1 data store | 21:22 |
| bmace | it would be nice. "control" over who is doing the deploy could even be managed that way | 21:23 |
| sdake | agree | 21:23 |
| bmace | some sort of lock held in etcd | 21:23 |
| sdake | although there would need to be a reset the lock mode ;) | 21:24 |
| bmace | yup, a command to say.. i know better than you, wipe the lock :) | 21:24 |
| sdake | fantastic we just designed ha in 1 hour that is easymode | 21:24 |
| sdake | you asked 2 months ago sorry it took me so long ;) | 21:24 |
| bmace | teamwork! :) | 21:25 |
| sdake | you asked before i understood what the constraints were | 21:25 |
| *** dwalsh has quit IRC | 21:27 | |
| harmw | jpeeler: weird, my docker sock file didn't get chgrpd to docker grp | 21:29 |
| jpeeler | harmw: well, i am using a different image. had to for libvirt support | 21:29 |
| harmw | nonetheless, the service file would be the same | 21:29 |
| harmw | so: weird | 21:29 |
| harmw | :) | 21:29 |
| harmw | :libvirt__dhcp_enabled | 21:29 |
| harmw | can't you use that? | 21:29 |
| jpeeler | well it also is marked as true by default? | 21:34 |
| harmw | in that case, you only need something like 'private_network' | 21:34 |
| jpeeler | i've tried that, it reports that it requires usage of ip | 21:34 |
| jpeeler | "* An IP is required for a private network." | 21:38 |
| harmw | ok | 21:39 |
| harmw | sucks | 21:39 |
| harmw | maybe 0.0.0.0 works | 21:40 |
| harmw | hm, bootstrap_nova isn't progressing : | 21:40 |
| harmw | :| | 21:40 |
| jpeeler | harmw: i can try that i guess. but maybe i should do this on just one node... | 21:46 |
| harmw | hehe true | 21:47 |
| harmw | you can just vagrant up compute01 afaik | 21:47 |
| harmw | the operator is heavy on requirements, the others aren't | 21:47 |
| harmw | luckily :p | 21:47 |
| jpeeler | oh cool, didn't know that. still quite the beginner | 21:49 |
| jpeeler | well it didn't accept 0.0.0.0 as the ip | 21:50 |
| jpeeler | i mean did | 21:50 |
| harmw | ah | 21:50 |
| harmw | ok | 21:50 |
| jpeeler | will let you know how the interfaces look in a sec | 21:50 |
| jpeeler | why do you say only the operator is heavy though? it looks like they all execute the bootstrap script | 21:51 |
| harmw | the operator is yum installing quite a lot | 21:51 |
| harmw | the other don't have to since they won't be running anything difficult | 21:51 |
| harmw | (uhm, right? can't even remember, lol) | 21:52 |
| harmw | ah great, ansible is done deploying OS :) | 21:52 |
| harmw | $ neutron agent-list | 21:53 |
| harmw | publicURL endpoint for network service not found | 21:53 |
| harmw | whoops | 21:53 |
| *** banix has quit IRC | 21:57 | |
| sdake | harmw you get that with master? | 21:59 |
| sdake | i hope the images i pushed are not busted | 21:59 |
| harmw | no, my images are 13 days old :) | 22:00 |
| harmw | ansible code is fresh pull though | 22:00 |
| harmw | but things locked up when I first launched nova, it freezed/timedout whatever | 22:01 |
| harmw | I'm only starting services per tag, from ansible | 22:01 |
| harmw | we need a nuke button on this btw :p | 22:01 |
| harmw | multihost-nuke | 22:01 |
| harmw | jpeeler: | 22:06 |
| harmw | fatal: [support01] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute u'ansible_enp0s8'", 'failed': True} | 22:06 |
| harmw | fatal: [support01] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute u'ansible_enp0s8'", 'failed': True} | 22:07 |
| harmw | interesting to see that all of a sudden | 22:07 |
| harmw | haproxy | 22:07 |
| jpeeler | yeah that's what i saw! | 22:09 |
| *** diga has quit IRC | 22:10 | |
| jpeeler | except ansible_eth0 instead | 22:11 |
| harmw | jpeeler: you mind filing a bug on this? | 22:11 |
| harmw | lets see if I can fix this | 22:11 |
| jpeeler | well no, but i don't understand what the problem is | 22:12 |
| harmw | undefined variable :) | 22:12 |
| jpeeler | can i file it in a few? i gotta go take care of something | 22:12 |
| jpeeler | maybe 10 min | 22:12 |
| harmw | yea totaly | 22:12 |
| *** diga has joined #kolla | 22:14 | |
| *** diga has quit IRC | 22:15 | |
| *** diga has joined #kolla | 22:17 | |
| jpeeler | harmw: the undefined variable has nothing to do with vagrant, right? | 22:23 |
| harmw | ofcourse not :) | 22:23 |
| jpeeler | just checking! | 22:23 |
| harmw | :) np | 22:23 |
| harmw | hm, looks like the haproxy is a little screwed | 22:25 |
| jpeeler | https://bugs.launchpad.net/kolla/+bug/1482406 | 22:26 |
| openstack | Launchpad bug 1482406 in kolla "ansible deploy yielding undefined variable " [Undecided,New] | 22:26 |
| harmw | could you make that specific to haproxy, if it's not already? | 22:26 |
| jpeeler | i hadn't | 22:27 |
| jpeeler | how did you determine that was from haproxy? | 22:27 |
| harmw | ansible told me :) | 22:27 |
| harmw | it says up front which play it's going to run | 22:27 |
| harmw | but perhaps you need a -v for that | 22:27 |
| harmw | (most of the times I run with -vv) | 22:28 |
| jpeeler | right, just not used to reading ansible output | 22:28 |
| harmw | np | 22:28 |
| jpeeler | it is in the haxproxy task | 22:28 |
| jpeeler | this past hour has made me feel like i know nothing heh | 22:28 |
| sdake | folks can you discuss https://blueprints.launchpad.net/kolla/+spec/deploy-ha plz | 22:28 |
| harmw | it tells you it'll be doing something with configuration stuff | 22:28 |
| harmw | from haproxy | 22:28 |
| jpeeler | yeah i see now | 22:28 |
| harmw | so just take a look into that task | 22:28 |
| harmw | the task at hand (config.yml) will tell you it's mangling some template | 22:29 |
| harmw | and *that* fails | 22:29 |
| harmw | because that variable is missing | 22:29 |
| sdake | what is the variable | 22:29 |
| jpeeler | harmw: so now i have one interface with no IP, is that what we wanted? | 22:29 |
| harmw | - server {{ hostvars[host]['ansible_hostname'] }} {{ hostvars[host]['ansible_' + api_interface]['ipv4']['address'] }}:{{ keystone_admin_port }} check inter 2000 rise 2 fall 5 | 22:30 |
| harmw | + server {{ hostvars[ansible_hostname] }} {{ hostvars[inventory_hostname]['ansible_' + api_interface]['ipv4']['address'] }}:{{ keystone_admin_port }} check inter 2000 rise 2 fall 5 | 22:30 |
| harmw | thats times 100.000 sdake :) | 22:30 |
| harmw | (or maybe less) | 22:30 |
| harmw | but after that it fails to boot, because the config file is broken | 22:30 |
| sdake | jpeeler yes that is whaty ou want | 22:30 |
| harmw | jpeeler: excellent, yes | 22:30 |
| harmw | sdake: you know if anything was added to keepalived container these last 2 weeks? | 22:33 |
| harmw | ERROR: missing KEEPALIVED_HOST_PRIORITIES | 22:33 |
| harmw | could be fixed already though | 22:33 |
| sdake | harmw i hae not see that | 22:33 |
| sdake | harmw which imaags are ou building | 22:33 |
| sdake | rather which images did you use | 22:33 |
| harmw | 2 weeks old, self build | 22:33 |
| sdake | use git blame on that line | 22:34 |
| sdake | i think ansible wasn't done 2 weeks ago tbh for keepalived | 22:34 |
| sdake | that error looks like a config-internal operation, so if your running ansible thatshould not be printing | 22:34 |
| harmw | yep, but it did :) | 22:35 |
| sdake | harmw if your using master of kolla it pulls liberty2 images | 22:35 |
| harmw | anyway, I'll rebuild | 22:35 |
| harmw | ansible is setup to pull from my own repo | 22:35 |
| harmw | (inside vagrant) | 22:35 |
| harmw | which is made up of self build images) | 22:35 |
| *** Haomeng|2 has joined #kolla | 22:37 | |
| sdake | i see, so if your using latest for your pull for vagrant that should be fine | 22:38 |
| sdake | master uses liberty2 | 22:38 |
| sdake | so when people pull kolla from git they don't have to build images | 22:38 |
| *** britthouser has joined #kolla | 22:38 | |
| *** Haomeng has quit IRC | 22:40 | |
| *** dims_ has quit IRC | 22:50 | |
| harmw | right, well the fix is probably real simple but my devenv is giving me a hard time :p better luck tomorrow | 23:04 |
| harmw | night guys | 23:04 |
| harmw | (what I just pasted isn't the proper fix btw) | 23:05 |
| *** dims_ has joined #kolla | 23:10 | |
| *** blahRus has quit IRC | 23:24 | |
| *** echoingumesh has quit IRC | 23:28 | |
| *** echoingumesh has joined #kolla | 23:29 | |
| *** echoingumesh has quit IRC | 23:33 | |
| *** sdake has quit IRC | 23:38 | |
| *** weiyu has joined #kolla | 23:56 | |
| *** jasonsb has quit IRC | 23:58 | |
| *** achanda has joined #kolla | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
