| *** gyee has quit IRC | 00:58 | |
| *** jamesdenton_wfh has joined #openstack-ansible | 01:37 | |
| openstackgerrit | James Denton proposed openstack/openstack-ansible-os_neutron master: Add networking-generic-switch support for Ironic integration https://review.opendev.org/717457 | 01:43 |
|---|---|---|
| openstackgerrit | James Denton proposed openstack/openstack-ansible-os_neutron master: Stop haproxy service on standalone network nodes https://review.opendev.org/708378 | 01:55 |
| *** oligau has quit IRC | 02:06 | |
| *** oligau has joined #openstack-ansible | 02:09 | |
| *** jamesdenton_wfh has quit IRC | 02:37 | |
| *** hamzy has quit IRC | 02:46 | |
| *** hamzy has joined #openstack-ansible | 02:46 | |
| *** prometheanfire has quit IRC | 03:11 | |
| *** prometheanfire has joined #openstack-ansible | 03:13 | |
| *** spatel has joined #openstack-ansible | 03:24 | |
| *** spatel has quit IRC | 03:25 | |
| *** evrardjp has quit IRC | 04:36 | |
| *** evrardjp has joined #openstack-ansible | 04:37 | |
| *** prometheanfire has quit IRC | 04:40 | |
| *** prometheanfire has joined #openstack-ansible | 04:40 | |
| *** kleini has joined #openstack-ansible | 04:46 | |
| *** prometheanfire has quit IRC | 05:01 | |
| *** prometheanfire has joined #openstack-ansible | 05:12 | |
| *** udesale has joined #openstack-ansible | 05:41 | |
| *** miloa has joined #openstack-ansible | 06:02 | |
| *** udesale has quit IRC | 06:09 | |
| *** udesale has joined #openstack-ansible | 06:13 | |
| *** viks____ has joined #openstack-ansible | 06:16 | |
| *** this10nly has joined #openstack-ansible | 06:26 | |
| *** thuydang has joined #openstack-ansible | 06:41 | |
| noonedeadpunk | thuydang: huh, it's for sure shouldn't be on utility.... | 06:50 |
| noonedeadpunk | can you paste lome logs from stdout about how it failed? | 06:51 |
| *** rpittau|afk is now known as rpittau | 06:56 | |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-rabbitmq_server master: Bump rabbitmq versions https://review.opendev.org/717821 | 06:59 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-rabbitmq_server master: Install rabbitmq from external_repo for all distros https://review.opendev.org/717819 | 07:10 |
| *** thuydang has quit IRC | 07:11 | |
| *** thuydang has joined #openstack-ansible | 07:22 | |
| *** tosky has joined #openstack-ansible | 07:24 | |
| thuydang | Dear noonedeadpunk, this is the stdout after rally start to build venv http://paste.openstack.org/show/791715/ | 07:34 |
| *** thuydang has quit IRC | 07:44 | |
| *** thuydang has joined #openstack-ansible | 07:44 | |
| thuydang | Sorry, I didn't the paste is limited in lines. So this is the exact stdout and ls of the folder in repo container: http://paste.openstack.org/show/791718/ | 07:45 |
| *** fghaas has joined #openstack-ansible | 07:54 | |
| *** isakgicu has joined #openstack-ansible | 08:34 | |
| openstackgerrit | amolkahat proposed openstack/openstack-ansible-os_tempest master: Mapping of tempest_extra_config parameter to tempest.conf file was absent. This patch make sure that parameters which are specified under tempest_extra_config should get added to the tempest.conf file. https://review.opendev.org/717982 | 08:39 |
| *** DanyC has joined #openstack-ansible | 08:49 | |
| *** sshnaidm|afk is now known as sshnaidm | 08:57 | |
| noonedeadpunk | thuydang: huh, both pastes missing https://opendev.org/openstack/ansible-role-python_venv_build/src/branch/master/tasks/python_venv_wheel_build.yml#L79-L104 :( | 09:10 |
| noonedeadpunk | so the thing is that all requirements and contraints jobs are delegated to the repo container | 09:11 |
| noonedeadpunk | which we see what actually happens in your poaste... | 09:11 |
| noonedeadpunk | so all these tasks were ran on the infra1_repo_container-0a35e648 (including the last one I think) | 09:12 |
| noonedeadpunk | I can actually recall some bug related to rally in 20.0.0 but not 100% sure about that. | 09:13 |
| noonedeadpunk | you can probably try doing minor upgrade to the latest train at least for the deploy host and try deploying rally again | 09:13 |
| noonedeadpunk | but htings does look pretty the same in terms of the role.. | 09:15 |
| *** thuydang has quit IRC | 09:20 | |
| *** admin0 has joined #openstack-ansible | 09:31 | |
| jrosser | noonedeadpunk: i was trying to look where we "fixed" the rally problems recently but am not finding it :) | 09:43 |
| noonedeadpunk | Me too... But I kinda recall having and fixing them lol | 09:43 |
| jrosser | indeed! | 09:43 |
| jrosser | unless we were just waiting for things in u-c to land then bump our sha | 09:44 |
| noonedeadpunk | oh, here it is | 09:46 |
| noonedeadpunk | https://opendev.org/openstack/openstack-ansible/commit/f899d972784e6ee568e153b4cb0564a3e1eee888 | 09:46 |
| noonedeadpunk | but it's too fresh I think | 09:46 |
| noonedeadpunk | e had one more iirc :p | 09:47 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-rabbitmq_server master: Install rabbitmq from external_repo for all distros https://review.opendev.org/717819 | 09:58 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-rabbitmq_server master: Install rabbitmq from external_repo for all distros https://review.opendev.org/717819 | 09:58 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-rabbitmq_server master: Bump rabbitmq versions https://review.opendev.org/717821 | 10:01 |
| noonedeadpunk | dirk: do you know if there's any possibility to create some meta packages in https://download.opensuse.org/repositories/network:/messaging:/amqp/ so that they do correspond requirements in https://packagecloud.io/rabbitmq/rabbitmq-server/packages/ ? | 10:17 |
| noonedeadpunk | or somehow to coordinate this stuff... as like epmd is a requirement, while in suse repo it's called erlang-epmd | 10:18 |
| noonedeadpunk | or maybe you can suggest somebody we can reach to address this? | 10:19 |
| noonedeadpunk | (oh, talking about requirements to https://packagecloud.io/rabbitmq/rabbitmq-server/packages/opensuse/15.1/rabbitmq-server-3.8.3-1.suse.noarch.rpm ) | 10:19 |
| noonedeadpunk | like 3.8.2 goes well though | 10:20 |
| *** ygk_12345 has joined #openstack-ansible | 10:24 | |
| *** rpittau is now known as rpittau|bbl | 10:25 | |
| *** DanyC has quit IRC | 10:36 | |
| *** DanyC has joined #openstack-ansible | 10:37 | |
| *** DanyC has joined #openstack-ansible | 10:38 | |
| *** jamesdenton_wfh has joined #openstack-ansible | 11:33 | |
| *** DanyC has quit IRC | 11:46 | |
| openstackgerrit | James Denton proposed openstack/openstack-ansible-os_neutron master: Add networking-generic-switch support for Ironic integration https://review.opendev.org/717457 | 11:47 |
| *** rh-jelabarre has joined #openstack-ansible | 11:48 | |
| *** DanyC has joined #openstack-ansible | 11:48 | |
| *** sshnaidm is now known as sshnaidm|afk | 11:58 | |
| *** rpittau|bbl is now known as rpittau | 12:06 | |
| openstackgerrit | Merged openstack/openstack-ansible master: Update used cirros images https://review.opendev.org/717725 | 12:12 |
| *** thuydang has joined #openstack-ansible | 12:19 | |
| thuydang | Thanks noonedeadpunk, that sounds good. | 12:26 |
| noonedeadpunk | thuydang: jsut tine question - do you really was isntalling rally, or was just following our guide? | 12:27 |
| thuydang | I'm following the OSA guide and stuck there | 12:27 |
| noonedeadpunk | As in the last case your cloud is probably ready-to-go, as rally is not essential component and used for testing things like cloud SLA | 12:28 |
| noonedeadpunk | so it's not required and being deployed the last one, after all other services | 12:28 |
| noonedeadpunk | so actually, you can ingore rally if you don't need it at this point | 12:28 |
| *** fghaas has left #openstack-ansible | 12:28 | |
| thuydang | Ah, that's good to know. I didn't remember enabling it, so I thought it's important | 12:29 |
| *** rohit02 has joined #openstack-ansible | 12:36 | |
| rohit02 | hi......can we install openstack-ansible train on centos 8? | 12:37 |
| rohit02 | does it support centos8? | 12:37 |
| *** ygk_12345 has quit IRC | 12:37 | |
| jrosser | rohit02: train branch does not support centos8 | 12:40 |
| jrosser | we have some work in progress on master branch which it is likley to backport to train at some point | 12:40 |
| jrosser | but currently there is no-one working specifically on centos8 support | 12:41 |
| rohit02 | ok..thank u so much jrosser | 12:41 |
| *** redrobot has joined #openstack-ansible | 12:59 | |
| *** udesale_ has joined #openstack-ansible | 13:03 | |
| *** udesale has quit IRC | 13:05 | |
| *** thuydang has quit IRC | 13:09 | |
| *** thuydang has joined #openstack-ansible | 13:09 | |
| *** rohit02 has quit IRC | 13:21 | |
| *** sshnaidm|afk is now known as sshnaidm | 13:27 | |
| *** _kukacz has joined #openstack-ansible | 13:45 | |
| *** thuydang has quit IRC | 13:46 | |
| *** dave-mccowan has joined #openstack-ansible | 13:59 | |
| fridtjof[m] | I'm currently working on migrating all my hosts to Ubuntu 18.04, so I can move to stein. | 14:02 |
| fridtjof[m] | For compute hosts (with shared storage on a separate server), is it enough to mgirate all instances off of it, disable the service, reinstall it with 18.04, and then run the setup playbooks again? | 14:02 |
| fridtjof[m] | Or should I completely remove it with the -ops playbook, then add it back again? | 14:03 |
| *** this10nly has quit IRC | 14:08 | |
| *** dave-mccowan has quit IRC | 14:10 | |
| noonedeadpunk | fridtjof[m]: ou will also have to re-run repo_build role to get wheel for ubuntu 18.04 | 14:18 |
| noonedeadpunk | Also you may ask questions to ebbex - he just completed such upgrade I guess | 14:19 |
| fridtjof[m] | i'm done with that fortunately, already set up a second infra host on 18.04 for replication | 14:19 |
| noonedeadpunk | ah, ok then | 14:19 |
| noonedeadpunk | so if hostname and ip will be the same, it should be just fine I think | 14:20 |
| fridtjof[m] | okay - anything i should pay attention to on the ansible side? any state i should clear on the deployment host? or will that solve itself automatically? | 14:22 |
| noonedeadpunk | I think you'd better drop cached fact for this hosts | 14:22 |
| fridtjof[m] | I hope you don't mind me asking, but how do i do that? | 14:24 |
| jrosser | fridtjof[m]: there are some notes here which you could validate https://review.opendev.org/#/c/710803/ | 14:24 |
| jrosser | noonedeadpunk: do you know why this is necessary? https://github.com/openstack/openstack-ansible-os_ceilometer/blob/master/tasks/ceilometer_pre_install.yml#L32-L41 | 14:25 |
| * noonedeadpunk never used swift... | 14:25 | |
| fridtjof[m] | oh, very nice! thank you | 14:25 |
| *** DanyC_ has joined #openstack-ansible | 14:27 | |
| noonedeadpunk | jrosser: dunno, maybe it mounts smth into home to get stats? | 14:27 |
| jrosser | i'm really not sure, it doesnt seem to be referenced in many other places and there is some quite funky use of variables in group vars to support that | 14:28 |
| noonedeadpunk | but I guess it jsut needs ResellerAdmin role to be assigned | 14:28 |
| jrosser | its a unix user on the host though? | 14:29 |
| noonedeadpunk | yeah, it is | 14:29 |
| noonedeadpunk | which makes even more weird | 14:29 |
| *** DanyC has quit IRC | 14:30 | |
| noonedeadpunk | oh | 14:30 |
| jrosser | i was looking at this https://github.com/openstack/openstack-ansible/blob/master/inventory/group_vars/ceilometer_all.yml#L32 | 14:30 |
| noonedeadpunk | maybe it's for using swift as a storage? | 14:30 |
| jrosser | and trying to figure out what what going on | 14:30 |
| noonedeadpunk | but as swift has exactly the same behaviour.... | 14:32 |
| noonedeadpunk | I guess other ceilometer stuff lives in group_vars | 14:32 |
| noonedeadpunk | I meant in defaults inside role | 14:33 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-os_horizon master: Horizon configuration to support domains dropdown https://review.opendev.org/607474 | 14:35 |
| Adri2000 | hello, anyone knows if there is an existing doc or discussion on the best way to create additional, custom, lxc containers alongside existing ones? | 14:50 |
| jrosser | Adri2000: the ops repo elk stack does just that https://github.com/openstack/openstack-ansible-ops/tree/master/elk_metrics_7x | 14:52 |
| Adri2000 | jrosser: that looks interesting indeed, thank you | 14:56 |
| jrosser | you need the env.d to specify the new container and then reference the host it should live on in openstack_user_config | 14:57 |
| Adri2000 | right, that seems easy enough :) | 14:59 |
| *** thuydang has joined #openstack-ansible | 15:03 | |
| noonedeadpunk | I guess we have some docs regarding managing inventory here https://docs.openstack.org/openstack-ansible/latest/reference/inventory/inventory.html | 15:03 |
| *** sshnaidm is now known as sshnaidm|afk | 15:07 | |
| fridtjof[m] | when deploying an 18.04 machine, does it matter how i configure networking? (e.g. netplan instead of an interfaces file) | 15:16 |
| *** DanyC_ has quit IRC | 15:17 | |
| jrosser | the easiest solution is to install ifupdown with apt, then just copy over the config you had for 16.04 | 15:18 |
| *** krah has joined #openstack-ansible | 15:24 | |
| fridtjof[m] | yeah, im considering that | 15:24 |
| fridtjof[m] | On the other hand, MAAS generated the netplan equivalent of the interfaces file i had before anyway, except for the linuxbridge-agent veth configuration. | 15:24 |
| fridtjof[m] | (which i also had to manually add before, of course) | 15:25 |
| *** sshnaidm|afk is now known as sshnaid | 15:27 | |
| *** sshnaid is now known as sshnaidm | 15:27 | |
| fridtjof[m] | wait. Do i even need the veth pair if my external network is "flat" only, without VLANs? | 15:28 |
| fridtjof[m] | the comments here seem to imply it's only necessary for when you want to use multiple VLANs on that interface | 15:29 |
| *** DanyC has joined #openstack-ansible | 15:30 | |
| fridtjof[m] | hm, the example openstack_user_config also applies host_bind_override on a 'flat' type network. I guess i'm going to apply it then | 15:31 |
| *** gyee has joined #openstack-ansible | 15:31 | |
| jrosser | fridtjof[m]: can you be a bit more specific? do you mean eth12? | 15:33 |
| fridtjof[m] | yes, eth12 | 15:33 |
| fridtjof[m] | If i'm not going to run VLANs on the bridge i'd attach eth12 to (on a compute host where the neutron agent isn't in a container), do i need the veth pair at all? | 15:35 |
| jamesdenton_wfh | you might want to specify it but it should be ignored | 15:36 |
| fridtjof[m] | https://netplan.io/faq#use-pre-up-post-up-etc-hook-scripts | 15:36 |
| fridtjof[m] | okay, in any way i'll want to continue using interfaces anyway, i guess | 15:37 |
| fridtjof[m] | netplan doesn't do hooks properly | 15:38 |
| jamesdenton_wfh | it leaves much to be desired | 15:38 |
| *** miloa has quit IRC | 15:39 | |
| noonedeadpunk | #startmeeting openstack_ansible_meeting | 16:01 |
| openstack | Meeting started Tue Apr 7 16:01:14 2020 UTC and is due to finish in 60 minutes. The chair is noonedeadpunk. Information about MeetBot at http://wiki.debian.org/MeetBot. | 16:01 |
| openstack | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 16:01 |
| *** openstack changes topic to " (Meeting topic: openstack_ansible_meeting)" | 16:01 | |
| openstack | The meeting name has been set to 'openstack_ansible_meeting' | 16:01 |
| noonedeadpunk | #topic office hours | 16:01 |
| *** openstack changes topic to "office hours (Meeting topic: openstack_ansible_meeting)" | 16:01 | |
| *** _kukacz has quit IRC | 16:01 | |
| noonedeadpunk | Ok, so the first thing I'd lova to raise is reitrement of our roles | 16:01 |
| noonedeadpunk | specifically repo_build and pip_install | 16:02 |
| noonedeadpunk | So patches for them are ready for review https://review.opendev.org/#/c/716389/ https://review.opendev.org/#/c/717717/ | 16:02 |
| noonedeadpunk | the second thing is release highlights. | 16:03 |
| noonedeadpunk | We should place them by thursday I guess | 16:03 |
| *** sshnaidm is now known as sshnaidm|afk | 16:03 | |
| noonedeadpunk | However I'm still not sure what we can write down into them | 16:04 |
| noonedeadpunk | mnaser: maybe you have some ideas regarding that? | 16:04 |
| mnaser | noonedeadpunk: i don't know if something sticks to me right now :X | 16:04 |
| noonedeadpunk | yeah, the same thing... I guess we just tried to maintain things mostly | 16:05 |
| noonedeadpunk | We can kinda... announce centos8 support.... | 16:05 |
| mnaser | i would have loved to put centos 8 on that but.. | 16:05 |
| *** thuydang has quit IRC | 16:05 | |
| noonedeadpunk | yeah lol | 16:05 |
| mnaser | but i dont know in terms of resource and time how much we can do :X | 16:06 |
| noonedeadpunk | py3 support is release goal so we can't place that I think | 16:08 |
| *** rpittau is now known as rpittau|afk | 16:08 | |
| *** udesale_ has quit IRC | 16:08 | |
| jrosser | o/ hello | 16:10 |
| noonedeadpunk | so, should be jsut skipp filling in highlights? | 16:13 |
| noonedeadpunk | * should we just skip | 16:16 |
| noonedeadpunk | Also I think we should bump ansible version to 2.9 for ansible and mariadb to 10.4. | 16:17 |
| jrosser | ansible 2.9 looks do-able with the collection for rabbitmq | 16:17 |
| noonedeadpunk | For ansible it may cost us first collection usage | 16:17 |
| noonedeadpunk | yeah | 16:18 |
| jrosser | perhaps we need ansible-collection-requirements file | 16:18 |
| noonedeadpunk | actual mitogen usage might be smth to write down to highlights if we could merge that for U | 16:18 |
| jrosser | becasue there is a new format for galaxy that splits it into roles and collections in the same file | 16:18 |
| jrosser | yes i work all day today on mitogen | 16:18 |
| noonedeadpunk | oh, cool | 16:19 |
| jrosser | trying to reproduce what the gate fails on | 16:19 |
| jrosser | i think we should make a choice on how we are going to list collection requirements | 16:19 |
| jrosser | and where to install them | 16:20 |
| noonedeadpunk | also this will obviously requie modification of our bump tooling but that's fine | 16:20 |
| jrosser | https://docs.ansible.com/ansible/latest/user_guide/collections_using.html#install-multiple-collections-with-a-requirements-file | 16:20 |
| jrosser | choice is if we keep the roles and collections in the same file, or not | 16:21 |
| jrosser | particularly as we don't actually use galaxy at all to clone roles | 16:21 |
| jrosser | so for things to finish, we should certainly try for ansible 2.9 | 16:22 |
| jrosser | mariadb 10.4 will be trickier becasue i think there is changed stuff beyond just bumping the package version | 16:22 |
| jrosser | we ought to be able to land mitogen even if it's opt-in | 16:22 |
| jrosser | and zun feels very close too | 16:22 |
| jrosser | noonedeadpunk: actually ansible 2.9 is important as that then leads to ceph-ansible 5.0 which in turn now ceph octopus is released | 16:27 |
| noonedeadpunk | Yes, ansible is pretty important. | 16:28 |
| noonedeadpunk | So as we don't use galaxy I think it's probably should be another file? | 16:28 |
| noonedeadpunk | Actually I'd love to use the same but not sure we can fit this format at the moment | 16:29 |
| noonedeadpunk | so probably we should start from another one and then think about mergning them... | 16:29 |
| jrosser | i would be happy to start with a second file that just lists collections | 16:30 |
| jrosser | and use the galaxy cli to install them | 16:30 |
| noonedeadpunk | ++ | 16:30 |
| jrosser | we can always change that all later if needed | 16:30 |
| noonedeadpunk | maybe we'll decide to publish our roles one day.... | 16:31 |
| jamesdenton_wfh | o/ | 16:32 |
| jrosser | i think that the tricker bit is dealing with this in openstack-ansible-tests | 16:32 |
| jrosser | the first place we need to do this is in the rabbitmq role which uses the tests repo for functional tests | 16:33 |
| jrosser | and we would have to land a change there to use the collection first | 16:33 |
| noonedeadpunk | yeah, which would be kinda different from what we have in openstack-ansible | 16:36 |
| *** evrardjp has quit IRC | 16:36 | |
| noonedeadpunk | or we can copy there boostrap-ansible script for ansible isntallation inside functional tests | 16:37 |
| *** evrardjp has joined #openstack-ansible | 16:37 | |
| noonedeadpunk | which is probably pretty simmilar to https://opendev.org/openstack/openstack-ansible-tests/src/branch/master/test-ansible-env-prep.sh | 16:38 |
| jrosser | perhaps the simplest thing is to just install all the needed at some early point in openstack-ansible-tests | 16:38 |
| jrosser | not try to involve the specific roles in that | 16:38 |
| jrosser | *needed collections | 16:38 |
| noonedeadpunk | yeah, while isntalling ansible | 16:38 |
| jrosser | ah grr i wish i knew how to make this mitogen job fail | 16:39 |
| noonedeadpunk | maybe build image with dib? | 16:40 |
| noonedeadpunk | in case it's smth specific for ci.... | 16:40 |
| jrosser | maybe its becasue i run the steps manually and not with gate-check-commit...... | 16:40 |
| noonedeadpunk | hm, might be... | 16:41 |
| noonedeadpunk | but... like it runs one by one... except having some extra env set? | 16:41 |
| jrosser | it's all under the same shell i guess from the script | 16:41 |
| jrosser | i have to go afk now - will carry this on tomorrow | 16:42 |
| noonedeadpunk | yeah sure | 16:43 |
| *** DanyC has quit IRC | 16:45 | |
| *** DanyC has joined #openstack-ansible | 16:47 | |
| *** jbadiapa has quit IRC | 16:57 | |
| noonedeadpunk | #endmeeting | 16:59 |
| *** openstack changes topic to "Launchpad: https://launchpad.net/openstack-ansible || Weekly Meetings: https://wiki.openstack.org/wiki/Meetings/openstack-ansible || Review Dashboard: https://bit.ly/2SAcGAn" | 16:59 | |
| openstack | Meeting ended Tue Apr 7 16:59:37 2020 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 16:59 |
| openstack | Minutes: http://eavesdrop.openstack.org/meetings/openstack_ansible_meeting/2020/openstack_ansible_meeting.2020-04-07-16.01.html | 16:59 |
| openstack | Minutes (text): http://eavesdrop.openstack.org/meetings/openstack_ansible_meeting/2020/openstack_ansible_meeting.2020-04-07-16.01.txt | 16:59 |
| openstack | Log: http://eavesdrop.openstack.org/meetings/openstack_ansible_meeting/2020/openstack_ansible_meeting.2020-04-07-16.01.log.html | 16:59 |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible master: [WIP] Use the Mitogen connection method in OSA https://review.opendev.org/591236 | 17:36 |
| * prometheanfire has been looking for general partitioning guidelines for infra hosts but can't find any (yet), anyone know if OSA has any (even if they are generic) | 18:01 | |
| jrosser | noonedeadpunk: so the mitogen patch gets stranger - on the most recent patchset i am making it use my fork of ansible with an added debug message for where it failed before | 18:30 |
| jrosser | but now it looks like its got past that in the CI job without failing | 18:30 |
| jrosser | unless of course i've broken the exception handling here https://github.com/jrosser/ansible/commit/0d455268be8e55bdc42fbab0af6eb614cd889c08 | 18:31 |
| jrosser | prometheanfire: i guess it "depends" but you might want a small root partition and larger /openstack and /var mounted in OSA terms? glance needs a bunch of scratch space for pushing images around | 18:41 |
| prometheanfire | I was thinking of putting swift on the infra nodes (on a separate partition(s) for glance / general objstore) | 18:42 |
| prometheanfire | but ya, that is kinda what I thought | 18:43 |
| prometheanfire | is lvm not the way to go for container backends anymore? | 18:43 |
| jrosser | i mean for glance it needs local scratch space, for example creating image from volume may have to round-trip the entire dataset through the glance container | 18:44 |
| prometheanfire | oh, true | 18:44 |
| jrosser | that caught us out when someone tried to build an image with packer with a large root fs | 18:45 |
| prometheanfire | well, was thinking 250 /openstack; 50 /var to start, we don't 'plan' on having many images | 18:46 |
| prometheanfire | 2 2TB nvme drives, so, the rest is spare, for swift (thinking 500G from each) | 18:47 |
| prometheanfire | hmm, erased the 200 for / | 18:47 |
| prometheanfire | that leaves 1T, 'just in case' | 18:48 |
| noonedeadpunk | jrosser: exception lgtm | 18:50 |
| jrosser | it's passing buster lxc job | 18:50 |
| jrosser | quite quick too | 18:50 |
| jrosser | can't really see whats different between installing ansible from pip vs my fork | 18:51 |
| jrosser | well except its 2.8.8 vs 2.8.10 | 18:51 |
| jrosser | noonedeadpunk: it's going to be ~1hr10mins for the buster lxc job, down from anout 1hr50mins | 18:52 |
| *** DanyC has quit IRC | 18:53 | |
| jrosser | which is better than it looks becasue the job setup time and log storage is big chunks on either end that we can't do anything about | 18:53 |
| noonedeadpunk | that's true and pretty exciting | 18:54 |
| noonedeadpunk | yeah, really neat patch | 18:57 |
| *** fghaas has joined #openstack-ansible | 18:57 | |
| jrosser | no mitogen https://4e33cbff6d7dc83f63af-c0f8a1c31c5446efb34290a311cd3d95.ssl.cf5.rackcdn.com/717728/1/check/openstack-ansible-deploy-aio_lxc-debian-buster/49ca402/logs/ara-report/index.html | 18:57 |
| jrosser | with mitogen https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_b11/591236/29/check/openstack-ansible-deploy-aio_metal-debian-buster/b11fbaa/logs/ara-report/ | 18:57 |
| jrosser | really interesting to see how vastly quicker setup_hosts is | 18:58 |
| jrosser | becasue thats really the only part of the AIO deploy that targets many hosts at once, because a large number of the tasks hit all the containers | 18:59 |
| noonedeadpunk | would be interesting to see lxc jobs difference actually | 18:59 |
| jrosser | where for the service setup playbooks it's pretty much all just against one container per play | 18:59 |
| jrosser | so i don't think we will see the true speedup of mitogen in an AIO | 18:59 |
| jrosser | i expect it to be much better on a real multinode deploy | 18:59 |
| noonedeadpunk | but still connection might be more efficient that our connection plugin | 19:00 |
| jrosser | yes, did some experiments there | 19:00 |
| jrosser | 100 tasks of ping: | 19:00 |
| jrosser | stock ansible took 25 seconds against localhost | 19:01 |
| jrosser | and with mitogen 7.1 seconds | 19:01 |
| noonedeadpunk | wow | 19:01 |
| jrosser | if you enable fact caching with jsonfile + mitogen then it's 11.5 seconds | 19:01 |
| jrosser | so we pay quite a big per-task price for that | 19:01 |
| noonedeadpunk | I think we should cache facts like in memcached | 19:01 |
| noonedeadpunk | or redis | 19:02 |
| jrosser | there are some ansible bugs related to this | 19:02 |
| jrosser | not good ones either :( | 19:02 |
| jrosser | https://github.com/ansible/ansible/issues/48691 | 19:02 |
| jrosser | we should merge that ansible 2.8.10 patch | 19:04 |
| * jrosser back tomorrow | 19:04 | |
| * prometheanfire just realized that he needed to install buster and not stretch | 19:07 | |
| prometheanfire | ah well... | 19:07 |
| noonedeadpunk | that's sad:( | 19:07 |
| noonedeadpunk | regarding ansible bug it's sad s well | 19:07 |
| prometheanfire | only got to the partitioning screen, so not that bad | 19:08 |
| prometheanfire | 200 / 400 /openstack (lvm) and 200 /var (log host stuff, probably | 19:08 |
| noonedeadpunk | prometheanfire: I think we use journald everywhere for logging | 19:16 |
| noonedeadpunk | but it's still /var/log though:) | 19:17 |
| jamesdenton_wfh | noonedeadpunk speaking of journald, i don't think switch supports it | 19:18 |
| jamesdenton_wfh | *swift | 19:18 |
| noonedeadpunk | yeah, that's kinda true... | 19:19 |
| jamesdenton_wfh | I had errors with swift in an AIO, complaining about some config missing related to logging. Might need to put it back. | 19:19 |
| jamesdenton_wfh | https://github.com/openstack/openstack-ansible-os_swift/commit/d0fac1b559ade15cc6d66e39168da45b99be4a9a | 19:19 |
| jamesdenton_wfh | [account-auditor], [container-auditor], [object-auditor] specifically | 19:20 |
| jrosser | noonedeadpunk: on dear! I compare an lxc and metal job..... doh :( | 19:22 |
| noonedeadpunk | jamesdenton_wfh: huh, it seems that swift does write to journald though... however it writes about config error indeed https://e5b3447e32b9e9818c69-5541c36990a7c89f041bed0b72004574.ssl.cf5.rackcdn.com/713262/1/check/openstack-ansible-deploy-aio_metal-ubuntu-bionic/38e5319/logs/host/swift-account-auditor.service.journal-13-44-41.log.txt | 19:23 |
| prometheanfire | noonedeadpunk: sure | 19:23 |
| prometheanfire | I'd rather install ussuri as well, but can't have it all | 19:24 |
| jamesdenton_wfh | so it doesn't need to need use_journal then? | 19:26 |
| jamesdenton_wfh | *seem to need | 19:26 |
| jamesdenton_wfh | And yeah, the services perpetually restart. It was enough for me just to add the sections back, but not configure anything | 19:26 |
| prometheanfire | jamesdenton_wfh: sup yo | 19:27 |
| jamesdenton_wfh | hey prometheanfire. you couldn't stay away | 19:27 |
| jamesdenton_wfh | :D | 19:27 |
| noonedeadpunk | ah, yeah, I've dropped sections | 19:27 |
| noonedeadpunk | yeah.... my bad | 19:27 |
| noonedeadpunk | let/s return them | 19:27 |
| *** idlemind has joined #openstack-ansible | 19:28 | |
| prometheanfire | jamesdenton_wfh: well, finally doing what I was hired to do, a year later... | 19:28 |
| jamesdenton_wfh | i can push a patch if you'd like, unless you already have it handy | 19:28 |
| noonedeadpunk | just get used to consider dropping empty section as a safe operation | 19:28 |
| jamesdenton_wfh | prometheanfire you were hired to slam your face into a brick wall? | 19:28 |
| noonedeadpunk | no I don't have it handy | 19:28 |
| *** DanyC has joined #openstack-ansible | 19:29 | |
| prometheanfire | jamesdenton_wfh: yes, more or less | 19:29 |
| jamesdenton_wfh | noonedeadpunk i will take care of it. | 19:29 |
| noonedeadpunk | nice? thanks | 19:29 |
| prometheanfire | need a small cloud, has to be cheap and open source | 19:29 |
| *** fghaas has left #openstack-ansible | 19:29 | |
| prometheanfire | proxmox could work, but ironic is calling :D | 19:30 |
| jamesdenton_wfh | you should kick the tires on this, too: https://review.opendev.org/#/c/717457/ | 19:32 |
| * prometheanfire should see if he's still core on os_ironic :P | 19:32 | |
| prometheanfire | hmmm, that looks interesting | 19:33 |
| noonedeadpunk | do we need some testing for this scenario? | 19:33 |
| noonedeadpunk | btw, is https://review.opendev.org/#/c/717457/7/templates/plugins/ml2/ml2_conf_genericswitch.ini.j2 really needs to be present in this patch? | 19:35 |
| noonedeadpunk | (in terms if it's supposed to be just empty) | 19:39 |
| noonedeadpunk | ok, I see the idea | 19:40 |
| noonedeadpunk | tbh loved to use content there isntead of overrides but don't see right now how to easily work around this stuff https://opendev.org/openstack/openstack-ansible-os_neutron/src/branch/master/tasks/neutron_post_install.yml#L59 | 19:42 |
| jamesdenton_wfh | hey, sorry. yes, testing can be done w/ OVS, just need to build it out | 19:44 |
| jamesdenton_wfh | and yes, that's supposed to be an empty file that then gets populated with the respective ini overrides defined by the operator. | 19:45 |
| jamesdenton_wfh | neutron_plugins[neutron_plugin_type] is, AFAICT, going to ever only be a single ML2 plugin (ovs or lxb or whatever). I think genericswitch may be the only one brought in so far that is not built-in and doesn't require an agent. so, i worked within the boundaries of what we have there | 19:47 |
| noonedeadpunk | I porbably see some issue there... | 19:47 |
| noonedeadpunk | yeah, that's why I was thinking about | 19:48 |
| prometheanfire | jamesdenton_wfh: would you still say that LB is the way to go (for simple setups) | 19:48 |
| noonedeadpunk | s/why/what/ | 19:49 |
| jamesdenton_wfh | yes. upstream kinda sorta wants to squash it, but i'm hoping not | 19:49 |
| *** DanyC has quit IRC | 19:49 | |
| prometheanfire | squash LB? | 19:49 |
| noonedeadpunk | I think it supposed to be LXB | 19:50 |
| jamesdenton_wfh | yes, there's a thread on ML from a few months back | 19:50 |
| jamesdenton_wfh | linuxbridge, is what you mean, prometheanfire ? | 19:50 |
| jamesdenton_wfh | noonedeadpunk the genericswitch ini is getting done here: https://opendev.org/openstack/openstack-ansible-os_neutron/src/branch/master/tasks/neutron_post_install.yml#L135-L144 | 19:51 |
| prometheanfire | jamesdenton_wfh: that'd suck if so, I must have missed the thread | 19:52 |
| noonedeadpunk | ah, so it can be deployed in several places.. | 19:53 |
| noonedeadpunk | yeah | 19:53 |
| jamesdenton_wfh | Yeah, so the "core" ML2 driver (ovs, lxb, opendaylight, etc) would get done in your link, and then any supplemental ml2 drivers (sriov, genericswitch, etc) get done in the second | 19:54 |
| noonedeadpunk | except genericswitch is going to be the only one? is this scenario is even possible? | 19:55 |
| noonedeadpunk | if genericswitch can be only secondary one, than we can just omit src and add `content` key (which will be also ommited by default) | 19:56 |
| jamesdenton_wfh | prometheanfire http://lists.openstack.org/pipermail/openstack-discuss/2019-November/010702.html | 19:56 |
| prometheanfire | thanks | 19:57 |
| noonedeadpunk | so we don't have an empty teplate in here :p | 19:57 |
| noonedeadpunk | and, deployers would be able to do lookup stuff | 19:57 |
| jamesdenton_wfh | noonedeadpunk It's not likely genericswitch would be the only driver. You would have lxb or ovs, and genericswitch would be there only for programming the switch for ironic, when needed. lxb/ovs handle VMs and DHCP, routers, etc. | 19:57 |
| jamesdenton_wfh | when you say omit src, and add content key.. where exactly to do thart? | 19:58 |
| noonedeadpunk | but if you deploy like only ironic (like a replacement of maas) | 19:58 |
| jamesdenton_wfh | Well, if you're deploying only ironic without neutron, then this is all moot | 19:59 |
| noonedeadpunk | yeah, ok, thanks for explaining.. | 19:59 |
| jamesdenton_wfh | When you deploy ironic w/ neutron integration, you can configure the baremetal port with switchport data (switch name/id and interface #) and the driver allows neutron to SSH to the switch and configure accordingly. | 20:00 |
| noonedeadpunk | so I was thinking about setting https://opendev.org/openstack/openstack-ansible-os_neutron/src/branch/master/tasks/neutron_post_install.yml#L137 to smth like "{{ ('content' not in neutron_plugins[item]) | ternary(neutron_plugins[item].plugin_ini ~ '.j2', omit) }}" | 20:01 |
| noonedeadpunk | and add to the same task content: "{{ neutron_plugins[item].content | default(omit) }}" | 20:02 |
| jamesdenton_wfh | oh, i see | 20:02 |
| prometheanfire | jamesdenton_wfh: well, tbh, we may use sriov anyway | 20:03 |
| jamesdenton_wfh | well, actually, the config file is required. it's just empty from a deployment perspective until someone sets overrides (i.e. switch definitions) | 20:03 |
| jamesdenton_wfh | it was that, or have some goofy jinja loops based on variables | 20:04 |
| noonedeadpunk | but we set it to "" in defaults | 20:04 |
| noonedeadpunk | which eventually should create just empty file | 20:04 |
| jamesdenton_wfh | oh i see | 20:04 |
| noonedeadpunk | I just don't remeber if we can not setting overrides or not | 20:05 |
| noonedeadpunk | but we can default them at the same task to empty dict | 20:05 |
| jamesdenton_wfh | yeah, would have to test that for sure | 20:05 |
| noonedeadpunk | or actually we can just go with current version | 20:08 |
| openstackgerrit | James Denton proposed openstack/openstack-ansible-os_swift master: Revert deleted config section in Swift config files https://review.opendev.org/718208 | 20:12 |
| jamesdenton_wfh | What is master? U? V? | 20:12 |
| noonedeadpunk | U or now | 20:13 |
| noonedeadpunk | (and defenitely for at least one mor month | 20:13 |
| jamesdenton_wfh | do we have any ironic jobs now? | 20:15 |
| noonedeadpunk | I guess only integrated ones only | 20:16 |
| *** DanyC has joined #openstack-ansible | 20:18 | |
| jamesdenton_wfh | one thing leads to another, that leads to another... :D | 20:18 |
| jamesdenton_wfh | prometheanfire you will still need something like lxb or ovs if you want dhcp | 20:31 |
| *** DanyC has quit IRC | 20:33 | |
| *** thuydang has joined #openstack-ansible | 20:33 | |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible master: [WIP] Use the Mitogen connection method in OSA https://review.opendev.org/591236 | 20:55 |
| openstackgerrit | Jonathan Rosser proposed openstack/openstack-ansible master: [WIP] Use the Mitogen connection method in OSA https://review.opendev.org/591236 | 20:55 |
| *** kleini has quit IRC | 20:58 | |
| *** DanyC has joined #openstack-ansible | 21:02 | |
| *** DanyC has quit IRC | 21:06 | |
| jamesdenton_wfh | noonedeadpunk So, this appears to work: src: "{{ (neutron_plugins[item].plugin_conf_touch) | ternary(omit, neutron_plugins[item].plugin_ini ~ '.j2') }}" | 21:21 |
| jamesdenton_wfh | well, i lied. one sec | 21:23 |
| *** thuydang has quit IRC | 21:32 | |
| jamesdenton_wfh | Needed the inverse, like you had it: src: "{{ ('touch' not in neutron_plugins[item]) | ternary(neutron_plugins[item].plugin_ini ~ '.j2', omit) }}" | 21:33 |
| jamesdenton_wfh | I removed all from the /etc/neutron/plugins/ml2/ director, and that built out linuxbridge_agent.ini, ml2_conf.ini and the ml2_conf_genericswitch.ini (w/o blank template) | 21:34 |
| *** DanyC has joined #openstack-ansible | 21:38 | |
| openstackgerrit | James Denton proposed openstack/openstack-ansible-os_neutron master: Add networking-generic-switch support for Ironic integration https://review.opendev.org/717457 | 21:42 |
| openstackgerrit | James Denton proposed openstack/openstack-ansible-os_neutron master: Add networking-generic-switch support for Ironic integration https://review.opendev.org/717457 | 22:13 |
| *** this10nly has joined #openstack-ansible | 22:47 | |
| NobodyCam | Good Afternoon OSA folks, is there a newer version of this: | 22:47 |
| NobodyCam | # The infra hosts that will run the central agents. | 22:47 |
| NobodyCam | # For running Ceilometer in HA for workload partitioning, see | 22:47 |
| NobodyCam | # https://docs.openstack.org/pike/admin/#support-for-ha-deployment | 22:47 |
| NobodyCam | or should I be asking in the ceilometer folks channel | 22:49 |
| NobodyCam | the comment is from conf.d/ceilometer.yml | 22:49 |
| *** jamesdenton_wfh has quit IRC | 23:05 | |
| openstackgerrit | Bjoern Teipel proposed openstack/ansible-role-python_venv_build stable/train: Verify version number for virtualenv>=20.0.0 https://review.opendev.org/718278 | 23:09 |
| *** isakgicu has quit IRC | 23:18 | |
| *** tosky has quit IRC | 23:26 | |
| *** this10nly has quit IRC | 23:46 | |
| prometheanfire | hmm, looks like ovn is experamental https://docs.openstack.org/openstack-ansible-os_neutron/latest/app-ovn.html | 23:57 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!