| kfox1111 | embargo lifted.... https://spectreattack.com/ :/ | 00:01 |
|---|---|---|
| *** daidv has quit IRC | 00:24 | |
| *** goldyfruit has joined #openstack-kolla | 00:26 | |
| harlowja | we all screwed | 00:32 |
| harlowja | lol | 00:32 |
| *** ipsecguy has quit IRC | 00:41 | |
| *** duonghq has joined #openstack-kolla | 00:48 | |
| openstackgerrit | zhongshengping proposed openstack/kolla-ansible stable/pike: Prevent cinder conf from copy nfs_shares to unrelated nodes https://review.openstack.org/531043 | 00:49 |
| openstackgerrit | zhongshengping proposed openstack/kolla-ansible stable/ocata: Prevent cinder conf from copy nfs_shares to unrelated nodes https://review.openstack.org/531044 | 00:50 |
| *** goldyfruit has quit IRC | 00:50 | |
| kfox1111 | heh. | 00:55 |
| *** kevinz has quit IRC | 01:31 | |
| openstackgerrit | Xinliang Liu proposed openstack/kolla master: Fix MariaDB bootstrap for 10.1 version https://review.openstack.org/529199 | 01:35 |
| openstackgerrit | zhongshengping proposed openstack/kolla-ansible master: Kuryr not starting https://review.openstack.org/531050 | 01:35 |
| *** threestrands has joined #openstack-kolla | 01:55 | |
| *** zhurong has joined #openstack-kolla | 02:13 | |
| *** Daviey has quit IRC | 02:14 | |
| *** jascott1 has joined #openstack-kolla | 02:22 | |
| jascott1 | "no agent of type Open vSwitch agent registered on host blah-node-2" | 02:23 |
| inc0 | ovs - eternal nemesis of jascott1 | 02:23 |
| inc0 | Jeffrey4l: thanks for taking care about todays meeting, I fell sick over night and just slept through | 02:24 |
| jascott1 | inc0 which service is it complaining about? | 02:24 |
| inc0 | well, you'll need to check logs of neutron-openvswitch-agent | 02:25 |
| jascott1 | inc0 not seeing anything in there. seems like Im missing a service on compute nodes | 02:28 |
| inc0 | that'd be service | 02:28 |
| jascott1 | inc0 is it 'neutron-openvswitch-agent-network-' or 'openvswitch-vswitchd-network' ? | 02:30 |
| *** ljjjustin has joined #openstack-kolla | 02:31 | |
| inc0 | vswitchd - actual ovs process, neutron agent is what's missing | 02:32 |
| inc0 | although agent might be missing because it broke because vswitchd is borked | 02:32 |
| jascott1 | ok this guy 'neutron-openvswitch-agent-network' is not running on compute node | 02:33 |
| inc0 | it should | 02:33 |
| inc0 | so should vswitchd | 02:34 |
| jascott1 | ok thanks | 02:35 |
| jascott1 | inc0 but not ovsdb right? | 02:35 |
| inc0 | ovsdb to | 02:36 |
| inc0 | too | 02:36 |
| jascott1 | oh i remember now (vacation is a mofo) I had labeled everything kolla_node and changed that selector to that. mustve lost that somewhere | 02:39 |
| *** daidv has joined #openstack-kolla | 02:43 | |
| openstackgerrit | Justin Scott proposed openstack/kolla-kubernetes master: WIP deploy with ansible inside orchestration container https://review.openstack.org/520847 | 02:44 |
| openstackgerrit | Dai Dang Van proposed openstack/kolla-ansible master: Support policy.yaml file for Aodh https://review.openstack.org/514565 | 02:49 |
| jascott1 | oh i had it in arboreal and it never made it over :/ | 02:51 |
| *** caowei has joined #openstack-kolla | 02:52 | |
| openstackgerrit | Justin Scott proposed openstack/kolla-kubernetes master: WIP deploy with ansible inside orchestration container https://review.openstack.org/520847 | 02:54 |
| *** masber has joined #openstack-kolla | 02:55 | |
| masber | good afternoon, is there a way to tell openstack-kolla to deploy vlan network instead of vxlan? | 02:55 |
| jascott1 | masber you mean kolla-ansible? | 02:56 |
| masber | kolla-ansible yes sorry | 02:56 |
| jascott1 | np just making sure :) | 02:56 |
| jascott1 | so many projects these days | 02:57 |
| masber | the reason I need vlan network is because I am looking forward using SR-IOV | 02:57 |
| masber | which does not work with vxlan | 02:57 |
| jascott1 | my guess is somewhere in /etc/kolla/globals.yml | 02:58 |
| inc0 | masber: I think we have guide for that | 03:00 |
| inc0 | let me check | 03:00 |
| inc0 | jascott1: a bit more to that | 03:00 |
| masber | thank you | 03:01 |
| inc0 | https://docs.openstack.org/kolla-ansible/latest/reference/networking-guide.html | 03:02 |
| inc0 | you might need also write vlan ranges to ml2.conf | 03:03 |
| inc0 | some notes here too https://docs.openstack.org/kolla-ansible/latest/reference/ironic-guide.html | 03:05 |
| inc0 | hmm...I'll need to double check all that | 03:05 |
| inc0 | masber: let me know if you run into problems and I'll write proper docs for it afterwards | 03:06 |
| masber | inc0, so I understand that the only thing I need to do is to add this neutron_tenant_network_types: "vxlan,vlan,flat" to the gloobals.yml file? | 03:07 |
| masber | ok, let me try | 03:08 |
| masber | inc0, is ironic still in techpreview? | 03:08 |
| masber | *tech-preview | 03:08 |
| inc0 | I think it's fairly stable now | 03:09 |
| inc0 | so about just globals I can't see it in docs but you might need to write template override for vlan ranges and whatnot | 03:09 |
| inc0 | config override* | 03:09 |
| masber | ok | 03:11 |
| *** unicell has quit IRC | 03:14 | |
| *** coolsvap has joined #openstack-kolla | 03:26 | |
| openstackgerrit | Chason Chan proposed openstack/kolla master: Migrate neutron-vpnaas-agent to neutron-l3-agent https://review.openstack.org/530208 | 03:30 |
| *** strgout has joined #openstack-kolla | 03:35 | |
| *** unicell has joined #openstack-kolla | 03:40 | |
| strgout | ok so i started asking about this last night and there was no beer involved at all. | 03:46 |
| strgout | i'm not getting how you populate the local registery if i'm not bulding the images myself. | 03:46 |
| strgout | shouldn't kolla-ansible pull -i inventory populate the local docker reg? | 03:48 |
| *** janonymous has joined #openstack-kolla | 03:51 | |
| adrianreza | strgout, if you want skip building your own images you can download old images from https://tarballs.openstack.org/kolla/images/ and then deploy your local insecure registry | 04:09 |
| adrianreza | strgout, this is the example | 04:11 |
| adrianreza | https://www.irccloud.com/pastebin/cy0bZrrM/ | 04:11 |
| adrianreza | strgout, and then in your /etc/kolla/globals.yml add docker_registry: "10.10.10.1:4000", docker_namespace: "lokolla" in docker section | 04:13 |
| *** ansmith has quit IRC | 04:18 | |
| strgout | hey sorry didn't see this. So do all the docs assume your going to build your own images in all cases? | 04:20 |
| *** caowei has quit IRC | 04:28 | |
| *** strgout has quit IRC | 04:57 | |
| *** caowei has joined #openstack-kolla | 05:14 | |
| *** janki has joined #openstack-kolla | 05:43 | |
| openstackgerrit | Qi Peng proposed openstack/kolla-ansible master: kolla-ansible bootstrap-servers failed because of docker restart failed https://review.openstack.org/531077 | 05:50 |
| *** markmcclain has quit IRC | 05:51 | |
| *** markmcclain has joined #openstack-kolla | 05:54 | |
| openstackgerrit | Justin Scott proposed openstack/kolla-kubernetes master: WIP deploy with ansible inside orchestration container https://review.openstack.org/520847 | 06:26 |
| *** namnh has joined #openstack-kolla | 06:36 | |
| *** zhurong has quit IRC | 06:47 | |
| *** zhurong has joined #openstack-kolla | 06:52 | |
| *** masuberu has joined #openstack-kolla | 06:59 | |
| xinliang | Hi , CI seems broken. | 07:00 |
| xinliang | many jobs failed looks the same: http://logs.openstack.org/95/529795/1/check/kolla-ansible-centos-source/e358e45/job-output.txt.gz | 07:00 |
| xinliang | cp: cannot stat ‘/var/lib/docker/volumes/kolla_logs/_data/*’: No such file or directory | 07:01 |
| *** jaosorior has quit IRC | 07:01 | |
| *** egonzalez has joined #openstack-kolla | 07:02 | |
| openstackgerrit | Chason Chan proposed openstack/kolla master: Migrate neutron-vpnaas-agent to neutron-l3-agent https://review.openstack.org/530208 | 07:05 |
| *** jaosorior has joined #openstack-kolla | 07:11 | |
| *** jaosorior has quit IRC | 07:13 | |
| *** jaosorior has joined #openstack-kolla | 07:14 | |
| *** jaosorior has quit IRC | 07:14 | |
| *** jaosorior has joined #openstack-kolla | 07:14 | |
| *** kristaps_ has quit IRC | 07:15 | |
| namnh | hi everyone, i am using docker version 17, and i created a local registry like this command [1]. But docker on target node cannot push/pull image from the local registry, it allways raise an error like this [2]. is there anyone encounter this error. | 07:19 |
| namnh | [1] http://paste.openstack.org/show/637583/ | 07:19 |
| namnh | [2] http://paste.openstack.org/show/637585/ | 07:19 |
| namnh | i am trying to deploy kolla on multi nodes. | 07:20 |
| openstackgerrit | Merged openstack/kolla-ansible master: Update READEME https://review.openstack.org/530424 | 07:20 |
| *** sbezverk has quit IRC | 07:21 | |
| *** threestrands has quit IRC | 07:25 | |
| *** jrist has quit IRC | 07:29 | |
| *** pcaruana has joined #openstack-kolla | 07:32 | |
| xinliang | Jerrey4l: CI seems broken | 07:33 |
| xinliang | many jobs failed looks the same: http://logs.openstack.org/95/529795/1/check/kolla-ansible-centos-source/e358e45/job-output.txt.gz | 07:33 |
| xinliang | cp: cannot stat ‘/var/lib/docker/volumes/kolla_logs/_data/*’: No such file or directory | 07:33 |
| *** cah_link has joined #openstack-kolla | 07:34 | |
| *** unicell has quit IRC | 07:35 | |
| *** cah_link has quit IRC | 07:37 | |
| *** cah_link has joined #openstack-kolla | 07:37 | |
| *** tovin07_ has joined #openstack-kolla | 07:53 | |
| egonzalez | namnh, need to configure insecure registry to point your registry | 07:54 |
| egonzalez | namnh, the port exposed by docker is 443 -p 443:443 | 07:56 |
| egonzalez | not 50000 | 07:56 |
| openstackgerrit | Justin Scott proposed openstack/kolla-kubernetes master: WIP deploy with ansible inside orchestration container https://review.openstack.org/520847 | 07:57 |
| *** ljjjustin has quit IRC | 08:02 | |
| namnh | egonzalez: yead, i will try it. | 08:06 |
| openstackgerrit | zhouya proposed openstack/kolla-ansible master: Support assigning HA traffic to dedicated interface https://review.openstack.org/481503 | 08:09 |
| *** shardy has joined #openstack-kolla | 08:16 | |
| daidv | egonzalez, Hi | 08:17 |
| *** kiennt26 has joined #openstack-kolla | 08:20 | |
| egonzalez | daidv, hi | 08:20 |
| daidv | egonzalez, thank for your review, I want to dicuss with u a little bit more. | 08:23 |
| egonzalez | daidv, what review? | 08:24 |
| daidv | egonzalez, #link https://review.openstack.org/#/c/514565/ | 08:24 |
| hrw | rwellum: one step at time. I managed to understand how to build (kolla) so time for deploy (k-a). | 08:25 |
| hrw | morning | 08:25 |
| egonzalez | daidv, if i put an empty /etc/kolla/config/aodh/somethingwrong.json, it will copy to aodh hosts breaking the service | 08:26 |
| daidv | 1. I'm using first_found to only get yaml or json policy file because we can not merged them | 08:26 |
| daidv | egonzalez, no, we just find policy.json or policy.yaml | 08:26 |
| egonzalez | right, missed that part | 08:26 |
| daidv | For your question about aodh_policy.results, when we don't have policy.json and policy.yaml, aodh_policy.results = [] | 08:27 |
| daidv | So, I think we can use it to check if users are customizing policies. | 08:28 |
| egonzalez | daidv, how about aodh_policy.results.0.stat.exists? | 08:30 |
| daidv | egonzalez, It will raise an error in case user didn't put any policy file | 08:31 |
| egonzalez | shouldnt, because policies are optionals | 08:31 |
| daidv | because as I said above, aodh.results will be empty list, we can call aodh_policy.results.0.stat.exists | 08:31 |
| egonzalez | gotcha | 08:31 |
| daidv | Ok, the last point is about optional": true | 08:32 |
| daidv | In old flow, we are using "optional": true to pass the case that user didn't have policy.json | 08:33 |
| *** b_bezak has joined #openstack-kolla | 08:33 | |
| daidv | but now, we need to know aodh_policy_file (policy.json or policy.yaml) to copy from source to dest | 08:33 |
| *** b_bezak has quit IRC | 08:34 | |
| daidv | So I think we shouldn't use optional: true any more, and just check if aodh_policy_file is defined | 08:34 |
| daidv | egonzalez, is it Ok? | 08:35 |
| egonzalez | ok, then remove optional to ensure the file is there to copy. We do not have jinja trimmers in other json files, please remove them, smth like http://paste.openstack.org/show/637657/ | 08:36 |
| egonzalez | other lgtm | 08:36 |
| daidv | egonzalez, ah, I have test with that but using an "{%-" with an "-" is better for us | 08:38 |
| daidv | it with clear white spaces for us | 08:38 |
| daidv | s/with/will | 08:38 |
| egonzalez | yeah, but we do not have it in other json files, just for consistency with other roles | 08:39 |
| openstackgerrit | Chason Chan proposed openstack/kolla master: Migrate neutron-vpnaas-agent to neutron-l3-agent https://review.openstack.org/530208 | 08:39 |
| daidv | egonzalez, I think we can do the same thing with other roles. | 08:40 |
| daidv | which are supporting policy files. | 08:40 |
| daidv | and we only it for check if policy_file is defined in json files. | 08:41 |
| *** cduchesne has quit IRC | 08:43 | |
| daidv | egonzalez, Ah other thing, I just see your comment about order of "Check if policies shall be overwritten" | 08:43 |
| egonzalez | for me is harder to review if there is a missing minus in the middle of a trimmer and know is going to render correctly the json (only my preferences, not others, not a hard requirement) | 08:43 |
| daidv | I need to move it up, because we need to know which policy format are using before we render config file with [oslo_policy] policy_file = which_one | 08:45 |
| openstackgerrit | Merged openstack/kolla-ansible stable/ocata: Prevent cinder conf from copy nfs_shares to unrelated nodes https://review.openstack.org/531044 | 08:46 |
| egonzalez | ahh, got it, ok then | 08:46 |
| *** cduchesne has joined #openstack-kolla | 08:48 | |
| openstackgerrit | Merged openstack/kolla-ansible stable/pike: Prevent cinder conf from copy nfs_shares to unrelated nodes https://review.openstack.org/531043 | 08:49 |
| daidv | egonzalez, so can I make your recommendation in a follow up patch? I will use that patch as an example to complete my blue print ASAP with other similar path for other roles? | 08:49 |
| egonzalez | daidv, i prefer in this one as is the first of a series, then replicate the final patch to all roles, not having to look into 2 different changes | 08:50 |
| *** b_bezak has joined #openstack-kolla | 08:52 | |
| *** b_bezak has quit IRC | 08:52 | |
| *** b_bezak has joined #openstack-kolla | 08:53 | |
| *** b_bezak has quit IRC | 08:54 | |
| *** b_bezak has joined #openstack-kolla | 08:55 | |
| daidv | egonzalez, Ah, I tested your suggestion about trimmer, but if with "{%-" we will have better output format. | 08:55 |
| daidv | #link http://paste.openstack.org/show/637657/ | 08:55 |
| *** b_bezak has quit IRC | 08:59 | |
| *** b_bezak has joined #openstack-kolla | 09:03 | |
| openstackgerrit | Martin André proposed openstack/kolla-ansible master: Use local cirros image if available https://review.openstack.org/498449 | 09:06 |
| daidv | egonzalez, Oops, my bad, you're right, it is not really need. Thank for your support, I will update my patch now. | 09:13 |
| *** jrist has joined #openstack-kolla | 09:13 | |
| openstackgerrit | Dai Dang Van proposed openstack/kolla-ansible master: Support policy.yaml file for Aodh https://review.openstack.org/514565 | 09:16 |
| *** ktibi has joined #openstack-kolla | 09:19 | |
| openstackgerrit | Tudosoiu Marian proposed openstack/kolla-ansible master: Related-Bug: 1682153 https://review.openstack.org/531115 | 09:22 |
| openstack | bug 1682153 in kolla-ansible "mariadb_recovery is prone to data loss" [Critical,Confirmed] https://launchpad.net/bugs/1682153 - Assigned to Tudosoiu Marian (mtudosoiu) | 09:22 |
| hrw | can someone take a look at https://review.openstack.org/#/c/530897/ one? it checks is installed python-openstackclient is at least 3.12 | 09:22 |
| openstackgerrit | Merged openstack/kolla-ansible master: Add tempurl to swift pipeline https://review.openstack.org/527677 | 09:23 |
| openstackgerrit | Merged openstack/kolla-ansible master: Fix failure handling in prechecks https://review.openstack.org/527095 | 09:25 |
| *** tovin07_ has quit IRC | 09:25 | |
| openstackgerrit | Taeha Kim proposed openstack/kolla-ansible master: Use local cirros image if available https://review.openstack.org/498449 | 09:26 |
| *** mgoddard has joined #openstack-kolla | 09:26 | |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla-ansible stable/pike: Remove the unnecessary register.yml from upgrade.yml for Nova https://review.openstack.org/528799 | 09:27 |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla-ansible stable/ocata: Remove extra configuration item in solum.conf.j2 https://review.openstack.org/472647 | 09:28 |
| openstackgerrit | Merged openstack/kolla-ansible master: fix the tempest failure for external ceph https://review.openstack.org/530892 | 09:39 |
| *** kiennt26 has quit IRC | 09:49 | |
| *** pbourke has quit IRC | 10:10 | |
| *** namnh has quit IRC | 10:27 | |
| openstackgerrit | Tudosoiu Marian proposed openstack/kolla-ansible master: Related-Bug: 1682153 https://review.openstack.org/531122 | 10:35 |
| openstack | bug 1682153 in kolla-ansible "mariadb_recovery is prone to data loss" [Critical,Fix committed] https://launchpad.net/bugs/1682153 - Assigned to Tudosoiu Marian (mtudosoiu) | 10:35 |
| *** duonghq has quit IRC | 10:37 | |
| egonzalez | daidv, missed one json file, otherwise will +2ed | 10:44 |
| *** markmcclain has quit IRC | 10:51 | |
| *** markmcclain has joined #openstack-kolla | 10:59 | |
| *** robbbe has joined #openstack-kolla | 11:03 | |
| openstackgerrit | Michal Rostecki proposed openstack/kolla-ansible master: dev mode: Add support for magnum https://review.openstack.org/529323 | 11:05 |
| openstackgerrit | Michal Rostecki proposed openstack/kolla-ansible master: dev mode: Add support for keystone https://review.openstack.org/529324 | 11:05 |
| openstackgerrit | Michal Rostecki proposed openstack/kolla-ansible master: [WIP] dev mode: Add support for neutron https://review.openstack.org/526466 | 11:05 |
| *** nhlfr has joined #openstack-kolla | 11:05 | |
| *** jstramka89 has joined #openstack-kolla | 11:14 | |
| jstramka89 | hello folks | 11:15 |
| *** semoac has quit IRC | 11:15 | |
| jstramka89 | trying to deploy ocata tag 4.0.0 but get this error : http://paste.openstack.org/raw/637818/ | 11:17 |
| jstramka89 | any ideas? | 11:17 |
| jstramka89 | seems problem with docker version | 11:17 |
| egonzalez | jstramka89, what python docker version have installed, pip freeze | grep -i docker | 11:24 |
| jstramka89 | egonzalez: docker-py==1.10.6 | 11:24 |
| -openstackstatus- NOTICE: zuul seems to have gotten stuck and will probably need a restart, please be patient | 11:27 | |
| *** openstackstatus has quit IRC | 11:28 | |
| *** openstack has quit IRC | 11:28 | |
| *** openstack has joined #openstack-kolla | 13:10 | |
| *** ChanServ sets mode: +o openstack | 13:10 | |
| *** janonymous has quit IRC | 13:11 | |
| cah_link | ok. i have found it in /var/lib/kolla/venv/bin/glance-wsgi-api | 13:11 |
| *** b_bezak has joined #openstack-kolla | 13:13 | |
| *** n0isyn0ise has joined #openstack-kolla | 13:33 | |
| *** pengdake has joined #openstack-kolla | 13:53 | |
| *** nhlfr has quit IRC | 13:57 | |
| *** mrostecki has joined #openstack-kolla | 13:59 | |
| *** tvignaud has quit IRC | 14:01 | |
| kfox1111 | morning. | 14:03 |
| *** diranerpi has quit IRC | 14:12 | |
| n0isyn0ise | hi and happy new year everyone | 14:17 |
| ktibi | Can we configure AvailabilityZone for a compute with kolla ? | 14:18 |
| n0isyn0ise | I'm experiencing some trouble trying to deploy kolla-ansible 6.0.0.0b2, neutron_l3_agent is restarting in loop, and neutron-server.log complains about privsep helper command | 14:18 |
| n0isyn0ise | did anyone else experience this trouble ? | 14:18 |
| kolla-slack1 | <aslancimbom100> n0isyn0ise please paste the exact log entries for someone to have a good idea of the problem | 14:20 |
| n0isyn0ise | there it is : https://pastebin.com/eRVui80B | 14:24 |
| *** vincent_vdk has left #openstack-kolla | 14:35 | |
| *** goldyfruit has joined #openstack-kolla | 14:36 | |
| *** gema has quit IRC | 14:38 | |
| *** pmisiak has joined #openstack-kolla | 14:42 | |
| pmisiak | hi guys, i'm thinking about migration from Mitaka to Pike | 14:43 |
| pmisiak | what is a suggested path? | 14:43 |
| pmisiak | is it any doc describing openstack release update? | 14:43 |
| pmisiak | can i jump directly from mitaka to pike? | 14:44 |
| -openstackstatus- NOTICE: zuul has been restarted, all queues have been reset. please recheck your patches when appropriate | 14:47 | |
| *** goldyfruit has quit IRC | 14:47 | |
| egonzalez | ktibi, yep, you can do it with nova commands, there is not any limitation, maybe need to add nova filters to the scheduler if they arent in the nova defaults | 14:58 |
| egonzalez | pmisiak, only can jump from release to release, cannot do multiple version jump | 15:03 |
| *** masuberu has quit IRC | 15:04 | |
| openstackgerrit | Juan Badia Payno proposed openstack/kolla master: Fluentd plugins rpm installation on CentOS https://review.openstack.org/529806 | 15:05 |
| *** goldyfruit has joined #openstack-kolla | 15:06 | |
| pmisiak | egonzalez: thanks, what is the workflow? I build containers for next release and do kolla upgrade ? | 15:10 |
| pmisiak | and which kolla-ansible version I should use? I suppose for the next release? | 15:10 |
| pmisiak | so I take kolla-ansible and kolla for neuton, build containers and run kolla upgrade? | 15:11 |
| egonzalez | n0isyn0ise, logs from l3 agent? | 15:11 |
| egonzalez | pmisiak, update kolla and kolla-ansible to next release, build next release containers, pull on the hosts, make appropiate changes (read release notes), db backups and stuff like that, kolla-ansible upgrade, test, repeat again with next release | 15:13 |
| *** jbrooks has quit IRC | 15:13 | |
| *** jbrooks has joined #openstack-kolla | 15:15 | |
| egonzalez | n0isyn0ise, i dont see any error in n-server or n-l3-agent, what versions you using, distro, install_type? | 15:15 |
| pmisiak | egonzalez: great, which release notes I should read? OpenStack release notes? or Kolla release notes? or both? | 15:15 |
| egonzalez | pmisiak, i'd do both since we only notify our changes, not other projects changes (ie, neutron changed some config) | 15:17 |
| pmisiak | egonzalez: I suppose that you took account of config changes developing kolla for a particular openstack release | 15:21 |
| egonzalez | pmisiak, yep, but we do not control custom configs or overrides by users | 15:21 |
| pmisiak | egonzalez: yes, sure :) | 15:21 |
| pmisiak | egonzalez: perfectly fine | 15:22 |
| n0isyn0ise | egonzalez: i finally succeed by re-deploying and didn't meet the issue anew, so sorry for wrong alert | 15:24 |
| pmisiak | egonzalez: If i want do multiple upgrades during one maintenance windows can I just do intermediate upgrades on controllers? And on computes only the last one? | 15:42 |
| openstackgerrit | Dai Dang Van proposed openstack/kolla-ansible master: Support policy.yaml file for Aodh https://review.openstack.org/514565 | 15:45 |
| *** pengdake has quit IRC | 15:53 | |
| *** jascott1 has quit IRC | 16:02 | |
| *** egonzalez has quit IRC | 16:07 | |
| *** pmisiak has quit IRC | 16:14 | |
| *** b_bezak has quit IRC | 16:19 | |
| *** cah_link has quit IRC | 16:20 | |
| *** mgoddard has quit IRC | 16:21 | |
| *** snuffkin has quit IRC | 16:22 | |
| *** b_bezak has joined #openstack-kolla | 16:24 | |
| *** b_bezak has quit IRC | 16:25 | |
| *** b_bezak has joined #openstack-kolla | 16:25 | |
| *** b_bezak has quit IRC | 16:30 | |
| ktibi | Anyone have already test kolla with 3 nodes controllers ? because when I try to shutdown one, the VIP is on the remaining two. | 16:31 |
| *** zshi has quit IRC | 16:38 | |
| *** ChanServ sets mode: -r | 16:49 | |
| *** mcnanci has joined #openstack-kolla | 16:49 | |
| n0isyn0ise | ktibi: unless I misunderstood, the VIP is used for that, HA cluster | 16:51 |
| *** pcaruana has quit IRC | 16:52 | |
| *** kristaps has joined #openstack-kolla | 16:53 | |
| hrw | https://marcin.juszkiewicz.com.pl/2018/01/04/today-i-was-fighting-with-nova-no-idea-who-won/ | 16:54 |
| *** jstramka89 has quit IRC | 16:54 | |
| *** wojdec has joined #openstack-kolla | 16:57 | |
| *** wojdec has quit IRC | 17:01 | |
| *** wojdec has joined #openstack-kolla | 17:02 | |
| inc0 | good mornign | 17:07 |
| hrw | inc0: read ^^ in free time | 17:09 |
| inc0 | hahah | 17:09 |
| rwellum | Good read hrw | 17:12 |
| hrw | thx | 17:12 |
| kfox1111 | morning. | 17:13 |
| inc0 | hrw: that reminds me my presentation from vancouver summit - deep dive into live migration | 17:15 |
| inc0 | this cpu features thingy is painful | 17:15 |
| hrw | inc0: link to video? | 17:16 |
| inc0 | https://www.openstack.org/videos/vancouver-2015/dive-into-vm-live-migration | 17:16 |
| inc0 | basically unless you have exactly homogenous env (hardware wise) and not have access to host-model or selecting cpu features from libvirt, you're screwed | 17:17 |
| *** wojdec1 has joined #openstack-kolla | 17:21 | |
| hrw | inc0: o! I see dulek as cospeaker ;D | 17:22 |
| inc0 | yup | 17:22 |
| inc0 | you work together now right? | 17:22 |
| hrw | same company | 17:22 |
| hrw | never met | 17:22 |
| inc0 | cool guy, you should meet, you're from around Gdansk as well right? | 17:23 |
| hrw | Szczecin | 17:23 |
| *** wojdec has quit IRC | 17:23 | |
| hrw | ~5h by car | 17:23 |
| inc0 | ah, well, ok, far away;) | 17:23 |
| * hrw off | 17:23 | |
| inc0 | good night | 17:23 |
| hrw | Nova won. | 17:23 |
| inc0 | haha | 17:24 |
| inc0 | or lost | 17:24 |
| inc0 | we don't know | 17:24 |
| inc0 | what we know is that you lost;) | 17:24 |
| inc0 | there will be other battles in this endless war, fear not | 17:24 |
| *** mdorman has joined #openstack-kolla | 17:26 | |
| *** ntpttr_laptop has joined #openstack-kolla | 17:29 | |
| *** wojdec has joined #openstack-kolla | 17:33 | |
| *** wojdec1 has quit IRC | 17:35 | |
| *** mgoddard has joined #openstack-kolla | 17:40 | |
| *** ntpttr_laptop has quit IRC | 17:49 | |
| *** ntpttr_laptop has joined #openstack-kolla | 17:52 | |
| *** ntpttr_laptop has quit IRC | 17:53 | |
| *** david-lyle has quit IRC | 18:00 | |
| *** david-lyle has joined #openstack-kolla | 18:01 | |
| *** jascott1 has joined #openstack-kolla | 18:13 | |
| *** gkadam has joined #openstack-kolla | 18:23 | |
| *** shardy has quit IRC | 18:25 | |
| *** dardelean has joined #openstack-kolla | 18:35 | |
| *** numans_afk is now known as numans | 18:36 | |
| kfox1111 | ok... got fluent-bit to forward to elasticsearch managed by elasticsearch-operator. :) | 18:39 |
| *** dardelean has quit IRC | 18:40 | |
| *** mchlumsky has quit IRC | 18:40 | |
| *** DosIK has quit IRC | 18:42 | |
| *** mchlumsky has joined #openstack-kolla | 18:46 | |
| *** n0cat has quit IRC | 18:49 | |
| *** robbbe has quit IRC | 18:52 | |
| *** mchlumsky has quit IRC | 18:57 | |
| jascott1 | excellent! | 18:59 |
| *** dardelean has joined #openstack-kolla | 19:01 | |
| *** dardelean has quit IRC | 19:02 | |
| *** dardelean has joined #openstack-kolla | 19:03 | |
| *** mchlumsky has joined #openstack-kolla | 19:04 | |
| *** wojdec1 has joined #openstack-kolla | 19:05 | |
| *** wojdec has quit IRC | 19:06 | |
| *** mchlumsky_ has joined #openstack-kolla | 19:07 | |
| *** dardelean has quit IRC | 19:07 | |
| *** wojdec has joined #openstack-kolla | 19:09 | |
| *** mchlumsky has quit IRC | 19:09 | |
| *** wojdec1 has quit IRC | 19:10 | |
| *** dardelean has joined #openstack-kolla | 19:11 | |
| rwellum | So you'd helm install two new containers kfox1111 - FB and ES? | 19:11 |
| *** dardelean has quit IRC | 19:11 | |
| *** cduchesne has quit IRC | 19:11 | |
| *** dardelean has joined #openstack-kolla | 19:12 | |
| rwellum | kfox1111: be cool if you end up with a doc on how you created a new operator... | 19:12 |
| rwellum | Then jascott1 can use it to create a deploy operator :) | 19:12 |
| *** cduchesne has joined #openstack-kolla | 19:14 | |
| *** dardelean has quit IRC | 19:16 | |
| *** jascott1 has quit IRC | 19:26 | |
| openstackgerrit | Christopher Nipper proposed openstack/kolla-ansible master: Add support for Neutron SRIOV agent container https://review.openstack.org/487610 | 19:26 |
| openstackgerrit | Christopher Nipper proposed openstack/kolla-ansible master: Add details about SRIOV enablement to the networking guide https://review.openstack.org/498112 | 19:26 |
| *** jascott1 has joined #openstack-kolla | 19:26 | |
| *** jascott1 has quit IRC | 19:28 | |
| *** jascott1 has joined #openstack-kolla | 19:29 | |
| *** jascott1 has quit IRC | 19:30 | |
| *** mdorman_ has joined #openstack-kolla | 19:30 | |
| *** mdorman has quit IRC | 19:31 | |
| *** mdorman_ is now known as mdorman | 19:31 | |
| *** jascott1 has joined #openstack-kolla | 19:33 | |
| *** scheuk has joined #openstack-kolla | 19:34 | |
| *** dardelean has joined #openstack-kolla | 19:36 | |
| *** dardelean has quit IRC | 19:37 | |
| *** dardelean has joined #openstack-kolla | 19:38 | |
| *** unicell has joined #openstack-kolla | 19:39 | |
| *** dardelean has quit IRC | 19:42 | |
| kfox1111 | rwellum: there isn't too much to creating an operator. | 19:43 |
| kfox1111 | 1. you have some code that runs in a container. | 19:43 |
| kfox1111 | 2. you create a crd definition. | 19:43 |
| kfox1111 | 3. you write some code that watches for your crd's and do something when one gets created/changed. | 19:43 |
| *** mgoddard has quit IRC | 19:45 | |
| kfox1111 | #2 looks something like this: https://github.com/kfox1111/charts/blob/kube-dhcp/incubator/kube-dhcp/templates/kubedhcp-crd.yaml | 19:45 |
| openstackgerrit | Christopher Nipper proposed openstack/kolla-ansible master: Add details about SRIOV enablement to the networking guide https://review.openstack.org/498112 | 19:48 |
| *** gkadam has quit IRC | 19:49 | |
| *** jascott1 has quit IRC | 20:02 | |
| *** jascott1 has joined #openstack-kolla | 20:02 | |
| *** jascott1 has quit IRC | 20:03 | |
| *** jascott1 has joined #openstack-kolla | 20:04 | |
| *** jascott1 has quit IRC | 20:08 | |
| *** jascott1 has joined #openstack-kolla | 20:10 | |
| *** markmcclain has quit IRC | 20:10 | |
| *** markmcclain has joined #openstack-kolla | 20:11 | |
| *** jascott1 has quit IRC | 20:13 | |
| *** dardelean has joined #openstack-kolla | 20:25 | |
| *** dardelean has quit IRC | 20:26 | |
| *** mgoddard has joined #openstack-kolla | 20:28 | |
| *** dardelean has joined #openstack-kolla | 20:32 | |
| *** dardelean has quit IRC | 20:32 | |
| *** dardelean has joined #openstack-kolla | 20:33 | |
| *** jascott1 has joined #openstack-kolla | 20:34 | |
| *** dardelean has quit IRC | 20:37 | |
| *** jascott1 has quit IRC | 20:38 | |
| *** jascott1 has joined #openstack-kolla | 20:41 | |
| *** jascott1 has quit IRC | 20:43 | |
| *** ktibi has quit IRC | 20:45 | |
| *** kristaps has quit IRC | 20:46 | |
| *** jascott1 has joined #openstack-kolla | 20:47 | |
| openstackgerrit | Mark Goddard proposed openstack/kolla-ansible master: Fix ironic inspector public endpoint https://review.openstack.org/528278 | 20:48 |
| *** wojdec has quit IRC | 20:48 | |
| *** mchlumsky has joined #openstack-kolla | 20:48 | |
| *** mchlumsky_ has quit IRC | 20:50 | |
| *** jascott1 has quit IRC | 20:50 | |
| *** jascott1 has joined #openstack-kolla | 20:54 | |
| *** jascott1 has quit IRC | 20:56 | |
| *** jascott1 has joined #openstack-kolla | 20:58 | |
| *** mgoddard has quit IRC | 21:05 | |
| kfox1111 | interesting... github seems to be starting to security scan code in repos hosted there... | 21:12 |
| *** jascott1 has quit IRC | 21:29 | |
| *** jascott1 has joined #openstack-kolla | 21:29 | |
| *** jascott1 has quit IRC | 21:30 | |
| *** jascott1 has joined #openstack-kolla | 21:30 | |
| *** jascott1 has quit IRC | 21:34 | |
| *** jascott1 has joined #openstack-kolla | 21:35 | |
| *** threestrands has joined #openstack-kolla | 21:35 | |
| *** threestrands has quit IRC | 21:35 | |
| *** threestrands has joined #openstack-kolla | 21:35 | |
| *** jascott1 has quit IRC | 21:37 | |
| *** jascott1 has joined #openstack-kolla | 21:38 | |
| rwellum | kfox1111: are you having to patch your clouds for the CPU bugs this week? | 21:39 |
| kfox1111 | inc: https://review.openstack.org/#/c/528724/ | 21:40 |
| kfox1111 | rwellum: as soon as possible, yeah. | 21:40 |
| *** jascott1 has quit IRC | 21:42 | |
| *** sbezverk has joined #openstack-kolla | 21:43 | |
| *** dmellado has quit IRC | 21:43 | |
| *** dmellado has joined #openstack-kolla | 21:46 | |
| inc0 | kfox1111 rwellum playbook orchestration incoming soon:) | 21:52 |
| inc0 | my personal request, let's merge it asap in it's rough, but functional, state, rewrite quickstart to become managable | 21:53 |
| inc0 | and then gradually improve it both on gate and general quality fields | 21:53 |
| rwellum | inc0: +1 sounds good. | 21:53 |
| inc0 | reason I'm asking for it is we need onboarding mechanism | 21:53 |
| inc0 | for new people | 21:53 |
| rwellum | This works inc0 : curl https://raw.githubusercontent.com/RichWellum/k8s/master/ko.py | python - ens3 ens4 --image_version pike -cni weave | 21:54 |
| rwellum | jk tbh | 21:54 |
| rwellum | Prefer a playbook and looking forward to playing with it. | 21:54 |
| inc0 | at the end it's going to be playbook running in container | 21:55 |
| rwellum | Now that's interesting - what do we gain? | 21:56 |
| inc0 | helm install compute-kit | 21:56 |
| inc0 | that will start a job on k8s | 21:57 |
| inc0 | which will build all microcharts | 21:57 |
| inc0 | setup all that's needed | 21:57 |
| inc0 | and run ansible-playbook to deploy thing | 21:57 |
| rwellum | Are you feeling better btw inc0 ? I'm in a conference and I swear 50% are hacking coughs and sneezes into the air... | 21:57 |
| kfox1111 | inc0: sure. though still want to see the seperation before I'm good with it. | 21:58 |
| inc0 | think of it as ansible-driven ioeratir | 21:58 |
| inc0 | operator | 21:58 |
| rwellum | The first part - helm install compute-kit - that works already? | 21:58 |
| kfox1111 | perfectly happy with an ansible driven operator. :) | 21:58 |
| inc0 | kfox1111: there is little overlap really since we don't have orch at all now;) | 21:58 |
| inc0 | well not fully | 21:58 |
| kfox1111 | inc0: so, simple enough to rename it operator/xxxx | 21:58 |
| inc0 | you still need to create secrets/configmaps | 21:58 |
| inc0 | build charts | 21:58 |
| kfox1111 | the charts should be prebuilt into the container I think. | 21:59 |
| kfox1111 | no reason to build them every time. | 21:59 |
| inc0 | after that helm install compute-kit will just throw it all with entrypoint | 21:59 |
| inc0 | ansible will be more sane workload and will do all the manual thingys | 21:59 |
| kfox1111 | would rather it not use the service charts. those have always been a bit sketchy. | 21:59 |
| inc0 | well, container has to be built | 21:59 |
| rwellum | kfox1111: +1 - this is the longest step in the whole process. | 21:59 |
| inc0 | whether it's us -> dockerhub or user | 22:00 |
| kfox1111 | yeah. but the charts don't need to be rebuilt on every launch. | 22:00 |
| inc0 | well, or both | 22:00 |
| inc0 | yeah I agree | 22:00 |
| inc0 | we can have periodic job much like kolla jobs | 22:00 |
| inc0 | that builds orch image and pushes to dockerhub | 22:00 |
| inc0 | then really you barely need to do anything besides running one yaml we document;) | 22:01 |
| kfox1111 | it should basically be a ansibleification of whats in tests/bin/ceph_workflow.sh | 22:01 |
| kfox1111 | +1. | 22:01 |
| inc0 | curl docs.openstack.org/kolla-k8s-compute-kit | kubectl create | 22:01 |
| inc0 | kfox1111: yeah pretty much | 22:01 |
| kfox1111 | like elasticsearc-operator or rook-operator. | 22:01 |
| inc0 | and making it non gate specific | 22:01 |
| kfox1111 | helm install rook-operator, | 22:01 |
| inc0 | yup | 22:02 |
| kfox1111 | kubectl create -f mycluster.yaml | 22:02 |
| kfox1111 | done. :) | 22:02 |
| inc0 | you can create globals.yml like you do for kolla-ansible and create configmap | 22:02 |
| inc0 | to control your os installation | 22:02 |
| inc0 | and bam, it installs | 22:02 |
| kfox1111 | globals.yaml overrides should probably be in mycluster.yaml. | 22:02 |
| kfox1111 | some flag saying: | 22:03 |
| inc0 | well whatever you want to pass it, it's fine | 22:03 |
| kfox1111 | ansible-genconfig: | 22:03 |
| kfox1111 | globals.yaml stuff here | 22:03 |
| kfox1111 | then there is only one document the operator has to pull. | 22:03 |
| inc0 | well, no reason we need both globals and cloud.yml | 22:03 |
| kfox1111 | right. | 22:03 |
| inc0 | just squash it to one file/configmap | 22:03 |
| inc0 | to govern all the things | 22:03 |
| kfox1111 | well, kind of depends what we're talking about I guess. | 22:04 |
| rwellum | Yeah that's always been confusing/overlapping | 22:04 |
| kfox1111 | globals.yml is mostly about config of the openstack ini stuff. | 22:04 |
| inc0 | in any case, that's for later | 22:04 |
| inc0 | now yes | 22:04 |
| kfox1111 | while cloud.yaml was just k8s specific archetecture bits. | 22:04 |
| inc0 | but if we squash it it also can do things like enable_cinder: true | 22:04 |
| kfox1111 | the k8s specific stuff is absolutelyl critical. | 22:04 |
| kfox1111 | the rest, is optional. | 22:04 |
| kfox1111 | you could generate config entirely yourself and skip the rest. | 22:05 |
| inc0 | I'm thinking of globals.yml in kolla-ansible standard - that also affects deployment | 22:05 |
| inc0 | however we name file, I don't care | 22:05 |
| kfox1111 | hmm.. well, for operator I don't really care so much. since it is a clean seperation at the microservice level. | 22:06 |
| kfox1111 | so wathever the devs think. | 22:06 |
| inc0 | yeha we'll figure it out | 22:06 |
| kfox1111 | kk. | 22:06 |
| kfox1111 | just be careful. we spent a lot of work making everything very customizable via the cloud.yaml stuff. | 22:07 |
| kfox1111 | far more I think then kolla-ansible is. | 22:07 |
| openstackgerrit | Merged openstack/kolla-kubernetes master: Cinder is missing the v3 API endpoints. https://review.openstack.org/528724 | 22:07 |
| kfox1111 | so please don't just revert to the kolla-ansible way. | 22:07 |
| inc0 | ofc | 22:08 |
| inc0 | I'm saying we can do it | 22:08 |
| rwellum | ^^^ Woot thanks inc0 kfox1111 (for the merge) | 22:08 |
| inc0 | idea I mean | 22:08 |
| kfox1111 | yup. just giving you a heads up. | 22:08 |
| inc0 | idea being one file to rule dem all | 22:08 |
| kfox1111 | the config layout is probably one of the easiest parts of it. so not too big a deal until closer to 1.0. | 22:08 |
| kfox1111 | yup. | 22:09 |
| inc0 | I think we can arrive to "edit globals.yml, fill it with options -> kubectl create configmap --from-file globals.yml -> kubectl create -f deply_me_openstack.yml -> wait for it -> wait for it -> this is how you spawn vm" | 22:10 |
| *** markmcclain has quit IRC | 22:10 | |
| *** markmcclain has joined #openstack-kolla | 22:11 | |
| rwellum | That is using ansible behind the scenes? | 22:11 |
| inc0 | yeah | 22:12 |
| inc0 | deploy me openstack would run a job from image published in hub | 22:12 |
| inc0 | image will have microcharts built and ansible playbook | 22:12 |
| inc0 | ansible playbook will read configmap and deploy openstack accordingly | 22:12 |
| kfox1111 | inc0: no configmap. | 22:14 |
| kfox1111 | kubectl create -f myopenstack.yaml | 22:14 |
| kfox1111 | shoudl be kind=OpenStackComputeKit or something like that. | 22:15 |
| inc0 | well, configmap can be part of it;) but ok, however we model input file, up to us | 22:15 |
| kfox1111 | under a spec section can be all the config. | 22:15 |
| inc0 | that works as well | 22:16 |
| kfox1111 | up to you, but I'm just telling you how operators work. | 22:16 |
| inc0 | then gate scenario would be just this one ifle | 22:16 |
| inc0 | file | 22:16 |
| kfox1111 | yeah. | 22:16 |
| kfox1111 | should look something like this: | 22:18 |
| kfox1111 | https://github.com/upmc-enterprises/elasticsearch-operator/blob/master/example/example-es-cluster-minikube.yaml | 22:18 |
| inc0 | wfm | 22:19 |
| *** markmcclain has quit IRC | 22:19 | |
| kfox1111 | the operator then looks at the current state, the requested state in the doc, and launches/upgrades helm charts from within its own container as needed to get the desired state. | 22:20 |
| *** markmcclain has joined #openstack-kolla | 22:20 | |
| kfox1111 | aka, dump out the spec section to an ansible compatible format and run asible on the roles. | 22:20 |
| inc0 | yeah somethign like that | 22:21 |
| *** ipsecguy_ has joined #openstack-kolla | 22:21 | |
| *** ManoX has joined #openstack-kolla | 22:27 | |
| *** jistr has quit IRC | 22:29 | |
| *** dardelean has joined #openstack-kolla | 22:33 | |
| openstackgerrit | Justin Scott proposed openstack/kolla-kubernetes master: Deploy with ansible inside orchestration container https://review.openstack.org/520847 | 22:34 |
| inc0 | yay | 22:35 |
| inc0 | let the festivities commence! | 22:35 |
| kfox1111 | still needs stuff moved to orchestration | 22:35 |
| *** dardelean has quit IRC | 22:38 | |
| *** spiette has joined #openstack-kolla | 22:40 | |
| *** dardelean has joined #openstack-kolla | 22:46 | |
| *** jascott1 has joined #openstack-kolla | 22:46 | |
| jascott1 | kfox1111 are you saying move 'ansible' dir or just the new piece? It uses the same var file so seems we would want to move it all... | 22:50 |
| *** dardelean has quit IRC | 22:50 | |
| kfox1111 | just the new pieces. | 22:53 |
| kfox1111 | the existing ansible dir is for genconfig only. | 22:53 |
| kfox1111 | and temporary. (hopefully) | 22:53 |
| kfox1111 | we reached an agreement a while ago that rather then add more to it, making it harder for someone to tackle genconfig reworking, | 22:54 |
| kfox1111 | we would keep orchestration code seperate. | 22:54 |
| kfox1111 | if you wanted to move the genconfig stsuff from ansible to genconfig/ansible or something like that for consistency, that would be ok too. | 22:55 |
| jascott1 | the orchestration piece uses the existing ansible to generate config and then creates configmaps from the resulting files | 23:01 |
| kfox1111 | can it just exec out to do the genconfig? | 23:02 |
| kfox1111 | will make it easier to replace with helm for genconfig when that gets done too. | 23:02 |
| jascott1 | its currently sharing the var file from ansible/ | 23:02 |
| jascott1 | it looks like one can pass in a var file now so maybe that would work | 23:03 |
| kfox1111 | whats it need from there? | 23:03 |
| jascott1 | i will have to check but not much iirc | 23:03 |
| kfox1111 | k. | 23:03 |
| inc0 | ansible isn't great about reusing code from outside your role | 23:06 |
| inc0 | so moving orch out of ansible could potentially end up with lots of duplication, and that'd be bad | 23:06 |
| *** dave-mccowan has joined #openstack-kolla | 23:07 | |
| jascott1 | looks like mostly interfaces and IPs | 23:07 |
| kfox1111 | inc0: today, you can do all of genconfig with an external command. why would that be hard to do via ansible calling the external command? | 23:08 |
| kfox1111 | jascott1: those should come from the CRD. so can skip them? | 23:08 |
| inc0 | what I'm saying is it's easier to keep genconfig command operable while having other tasks for deploy in same role | 23:09 |
| inc0 | than moving deploy and genconfig totally disconnected | 23:09 |
| inc0 | because deploy won't be able to access genconfigs default and vice versa | 23:09 |
| kfox1111 | inc0: but harder in other ways, like if someone pops up and is willing to work on genconfig to not ansible, now they have to troll through a lot of non genconfig code. | 23:10 |
| inc0 | so potentially duplication if particular value would be needed in both - for example in keystone endpoint scenrario | 23:10 |
| inc0 | it can be config.yml file in role neutron | 23:10 |
| kfox1111 | inc0: its going to not work anyway once its not an ansible file. so don't assume out of the box it will work. | 23:10 |
| inc0 | and only look at config.yml ignoring deploy.yml | 23:10 |
| kfox1111 | user should be pushing config through the CRD, the operator conversts that to an ansible vars that the ansible operator reads then drives genconfig / helm via that. | 23:11 |
| inc0 | yeah, but what I'm saying is if we separate roles completely (put it into different dir) this process ^ will potentially need to generate same var twice to 2 different places | 23:12 |
| inc0 | ansible is stupid and can't do things like inheritance of roles | 23:12 |
| kfox1111 | unavoidable I think. | 23:12 |
| inc0 | avoidable if we will keep role neutron as singular role | 23:13 |
| inc0 | with multiple sub-parts | 23:13 |
| kfox1111 | we have a case where openstack ini's and kube need to know some things and they are currently in two different languages. | 23:13 |
| inc0 | one for config and another for deploy | 23:13 |
| kfox1111 | helm and ansible. | 23:13 |
| inc0 | right, but we don't need to add 3rd way | 23:13 |
| inc0 | because ansible and ansible is same language | 23:13 |
| inc0 | I just don't think it will make any harder moving genconfig out of ansible really | 23:14 |
| inc0 | I mean, you look at config.yaml file and templates | 23:14 |
| kfox1111 | yeah. it will. for anyone but you I think. :/ | 23:14 |
| inc0 | and ignore all the rest | 23:14 |
| kfox1111 | and as soon as we do, then we are back in the same boat. | 23:15 |
| inc0 | you can keep them logically separate within same file | 23:15 |
| inc0 | role* | 23:15 |
| inc0 | no, I don't see why | 23:15 |
| kfox1111 | we're talking forking like 6 default vars? thats not a huge maintenence burdon. | 23:15 |
| inc0 | o say role "nova" config.yaml templates nova.conf | 23:15 |
| inc0 | and role "nova" deploy.yml runs helm install | 23:16 |
| kfox1111 | inc0: we're just phyilisophically in a very different place with reguards to repo layout. | 23:16 |
| kfox1111 | I want totally seperate repos for tasks like config and orchestration. | 23:16 |
| kfox1111 | you want to throw it all into /ansible. | 23:16 |
| inc0 | you want to move it out of ansible? remove conf.yaml from nova, add it to helm chart, deploy.yaml stays the same | 23:16 |
| kfox1111 | we agreed to agree earily to share 1 repo, but 2 different dirs. | 23:16 |
| kfox1111 | dont' weltch on the agreement now. | 23:16 |
| inc0 | right | 23:16 |
| inc0 | 1 repo different dirs | 23:16 |
| inc0 | ansible and non-ansible | 23:17 |
| inc0 | I don't want ansible in 2 places | 23:17 |
| kfox1111 | no, that is not what we agreed to. | 23:17 |
| kfox1111 | split by task type. | 23:17 |
| jascott1 | i can try the CRD/operator approach | 23:18 |
| jascott1 | if so I need to create operator that reacts to CRD, passes those vars to playbook invocation | 23:18 |
| kfox1111 | those of us that don't want to run orchestration want to make very sure when we use some parts we arn't accidently getting more then what we bargened for. | 23:18 |
| kfox1111 | jascott1: yeay. that. :) | 23:18 |
| kfox1111 | yeah. | 23:18 |
| inc0 | so reason I think it's going to be duplicate work is taht Justin will need to examine every variable from defaults.yml and see if it belongs only in config, only in deploy or both | 23:18 |
| inc0 | otherwise we have chaps | 23:18 |
| inc0 | chaos | 23:18 |
| jascott1 | do we automate the the creation of the CRD from ansible vars file? | 23:19 |
| jascott1 | that would keep one origin | 23:19 |
| kfox1111 | inc0: again, hopefully ansible genconfig is very short lived. gotta deal with this problem. | 23:19 |
| kfox1111 | user provides crd. | 23:19 |
| jascott1 | yes but we could gen it for them | 23:19 |
| kfox1111 | we can provide some exmple ones. | 23:19 |
| *** jistr has joined #openstack-kolla | 23:19 | |
| jascott1 | a script to create it, they still add it | 23:20 |
| kfox1111 | then they need tooling to gen it, which has deps, | 23:20 |
| kfox1111 | which should probably be in a container, | 23:20 |
| kfox1111 | and we go around the loop again. turtles all the way down. | 23:20 |
| jascott1 | so provide a template? | 23:20 |
| kfox1111 | we provide some example crd's on a website for users to pull? | 23:20 |
| kfox1111 | templates requrie template engine. same problem. :/ | 23:21 |
| jascott1 | ok lemme see if I got this | 23:21 |
| kfox1111 | though could use helm for that. | 23:21 |
| kfox1111 | for example, see: | 23:22 |
| kfox1111 | https://github.com/upmc-enterprises/elasticsearch-operator/tree/master/charts | 23:22 |
| jascott1 | CRD is just what we need for deploy or we moving genconfig vars too? | 23:22 |
| kfox1111 | under the elasticsearch chart is a chart that just does come generic config of a single CRD resource. | 23:22 |
| kfox1111 | both I think. | 23:22 |
| kfox1111 | it is a specific request for a cluster. | 23:22 |
| jascott1 | that makes it simpler | 23:23 |
| kfox1111 | "give me an openstack taht looks like this" | 23:23 |
| jascott1 | so 2 CRDs? | 23:23 |
| kfox1111 | 1 crd. | 23:23 |
| jascott1 | ok | 23:23 |
| kfox1111 | there should be a program that watches for those types of cruds, converts the settings in them to ansible and runs the orchestration ansible code on it. | 23:24 |
| kfox1111 | that ansible code should do stuff like: look to see if config is launched. if not, do genconfig. look to see if nova is launched. if not, call helm to launch each missing bit, etc. | 23:24 |
| jascott1 | based on <service>_enabled: in CRD | 23:25 |
| kfox1111 | yeah. | 23:25 |
| kfox1111 | no idea what exactly that schema should look like. | 23:25 |
| kfox1111 | the elasticsearch example again is a good starting point. | 23:25 |
| kfox1111 | https://github.com/upmc-enterprises/elasticsearch-operator/blob/master/example/example-es-cluster-minikube.yaml | 23:26 |
| kfox1111 | well.. | 23:26 |
| kfox1111 | they do it like this in the chart: | 23:27 |
| kfox1111 | https://github.com/upmc-enterprises/elasticsearch-operator/blob/master/charts/elasticsearch/templates/cluster.yaml#L10 | 23:27 |
| kfox1111 | you can enable/disable kibana from being in the cluster in the crd. | 23:27 |
| kfox1111 | their example I guess doesn't enable it unless you specifiy an image. | 23:27 |
| kfox1111 | but I like your idea much better of having an enable flag. more like what the helm chart is doing. | 23:28 |
| kfox1111 | one use case to consider is the one I posted to the helm issue tracker. | 23:28 |
| kfox1111 | say you have 3 types of compute nodes for nova. | 23:28 |
| kfox1111 | maybe nova.compute.xxxxx should be a list of settings rather then an enable/disable, | 23:29 |
| kfox1111 | so the user can specify 3 instances of nova-compute helm chart, with the particular settings for each. | 23:29 |
| kfox1111 | cinder-volume is probably in the same boat. | 23:29 |
| jascott1 | hmmm | 23:29 |
| kfox1111 | nova.compute[{nodeLabel: type1, other:...}, {nodeLabel: type2, other:...}] | 23:30 |
| *** goldyfruit has quit IRC | 23:31 | |
| kfox1111 | (we have at least 3 unique types of hardware in our one cloud, and at least 2 in another) | 23:31 |
| kfox1111 | nics are all different for sure. some have gpu's, etc. | 23:31 |
| jascott1 | inc0 how is that scenario handled in kolla-ansible? | 23:31 |
| *** dave-mccowan has quit IRC | 23:31 | |
| kfox1111 | node overrides I think. but relies on ansible pushing out settings directly to specific nodes during the deployment. | 23:32 |
| kfox1111 | becuase k8s is in the middle, the same trick wont work. | 23:32 |
| kfox1111 | we need individual instances of the microservice charts targeted to the hardware classes via node selectors to do the same. | 23:33 |
| inc0 | jascott1: you can do either per-node config overrides or override any ansible var in inventory | 23:33 |
| inc0 | node selectors would work, but I wonder if you can specify node label and pass it to ansible inventory for orch | 23:34 |
| *** dardelean has joined #openstack-kolla | 23:34 | |
| inc0 | or however you want to say "nodes with these labels, use vlan tags 200:300" | 23:34 |
| kfox1111 | inc0: thats what I'm talkign about I think. specing out the CRD in such a way as the operator knows it will have 3 node selectors and needs 3 helm instances. | 23:34 |
| *** dardelean has quit IRC | 23:34 | |
| inc0 | we | 23:34 |
| inc0 | we'd need to have configmap per node class | 23:35 |
| *** dardelean has joined #openstack-kolla | 23:35 | |
| kfox1111 | hmm.. ultimately, yeah, I guess thats true. | 23:35 |
| inc0 | or base configmap and override per node class | 23:35 |
| kfox1111 | yeah. | 23:35 |
| kfox1111 | I vaguely recall we decided something like that during a summit. | 23:35 |
| kfox1111 | but taht was like a year ago? | 23:35 |
| inc0 | this is what we discussed in ATL I think, nova-compute --config-file base.conf --config-file nodelabel.conf | 23:36 |
| kfox1111 | yeah. that. | 23:36 |
| kfox1111 | which kind of matches up with the way the helm charts are layed out. they can share more generic config and each instance has overrides if you want to override the more generic things. | 23:36 |
| jascott1 | where nova-compute is "helm install nova-compute -f <file>"? | 23:36 |
| kfox1111 | jascott1: no, thats the command inside the container. | 23:37 |
| inc0 | no, nova-compute is acutal nova-compute command | 23:37 |
| kfox1111 | there could be an overarching helm config package for base.conf. and then an optional one per instance of the nova-compute helm chart | 23:38 |
| kfox1111 | same happens I guess with the neutron packages too. | 23:38 |
| inc0 | or we can just assume user would create them manually | 23:38 |
| kfox1111 | you need nutron-l3-agent configured differently for netwokr nodes and compute nodes for exmaple. | 23:39 |
| kfox1111 | inc0: for now, thats probably just fine for the configmaps. | 23:39 |
| inc0 | and orch would check "this node has label foo, is there nova-compute-foo configmap available? if yes, pass it" | 23:39 |
| kfox1111 | just the helm instance settinsg are unavoidable. | 23:39 |
| *** dardelean has quit IRC | 23:39 | |
| kfox1111 | inc0: yeah. there is an 'optional' configmap flag in the pod that could handle that. | 23:39 |
| inc0 | right | 23:39 |
| inc0 | and that'd also have to be pass to config.json for command | 23:40 |
| kfox1111 | but regardless, we need as part of the crd, some iterated sets of config that loop over and launch multiple instances of the helm chart. | 23:40 |
| inc0 | so here's my very personal request | 23:40 |
| inc0 | let's merge base ansible | 23:40 |
| kfox1111 | with overrides to the helm config for that particular chart. | 23:40 |
| inc0 | and worry about crd later | 23:40 |
| kfox1111 | crd's the api. ansibles the implementation. | 23:41 |
| kfox1111 | the api is the real critical part. | 23:41 |
| inc0 | good idea, let's not make it requirement for base patch because we can do it later | 23:41 |
| inc0 | I agree | 23:41 |
| inc0 | but I just want quickstart asap | 23:41 |
| inc0 | better than we have today | 23:41 |
| inc0 | then it will incresingly be better, with crd and all | 23:41 |
| inc0 | but we need quickstart asap | 23:41 |
| kfox1111 | fair enough. move it to /orchestration, and I'm good with it. | 23:41 |
| inc0 | ok, I guess we can do that | 23:42 |
| jascott1 | aright will do | 23:42 |
| inc0 | jascott1: let's just copy defaults.yaml | 23:42 |
| kfox1111 | k. thx. | 23:42 |
| inc0 | we'll clean it up later | 23:42 |
| kfox1111 | +1 | 23:42 |
| jascott1 | ansible seems to have var file flag | 23:43 |
| kfox1111 | that works. | 23:43 |
| kfox1111 | its a temporary thing for a user to ever call it directly. | 23:43 |
| jascott1 | cool I will give it a shot | 23:43 |
| inc0 | jascott1: roles will not | 23:43 |
| inc0 | but you can check how much of role is new | 23:43 |
| kfox1111 | the operator will be calling it. so the extra little bit of cli isn't too bad. | 23:43 |
| kfox1111 | inc0: going to the helm conference? | 23:45 |
| inc0 | yeah | 23:45 |
| kfox1111 | sweet. I got permission today. | 23:45 |
| inc0 | too close to pass;) | 23:45 |
| kfox1111 | so will see you there. :) | 23:45 |
| kfox1111 | indeed. | 23:46 |
| inc0 | we can ride share with jascott1 ;) | 23:46 |
| jascott1 | yeah I volunteered to do misc grunt work | 23:46 |
| jascott1 | like shuttle organizers etc if need be | 23:46 |
| inc0 | is there more than 2 of us driving there? | 23:47 |
| jascott1 | idk | 23:48 |
| kfox1111 | not sure. | 23:53 |
| kfox1111 | I offered to drive but mgmt is leaning towards flying. they panic when ice could be involved. | 23:53 |
| SamYaple | mmm flying through ice sounds safe | 23:53 |
| * kfox1111 chuckles | 23:53 | |
| *** jistr has quit IRC | 23:54 | |
| *** jistr has joined #openstack-kolla | 23:55 | |
| *** mdorman has quit IRC | 23:55 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!