| *** lbragstad has quit IRC | 00:12 | |
| *** lbragstad has joined #openstack-ansible | 00:14 | |
| *** dxiri has quit IRC | 00:16 | |
| *** schwicht has joined #openstack-ansible | 00:43 | |
| *** schwicht has quit IRC | 00:47 | |
| *** aicilamd has joined #openstack-ansible | 00:54 | |
| *** aicilamd has quit IRC | 00:58 | |
| *** Emine has quit IRC | 01:55 | |
| *** waynr has joined #openstack-ansible | 01:58 | |
| waynr | howdy folks | 01:58 |
|---|---|---|
| waynr | so i'm using openstack-ansible which uses ansible 2.1 and i want to use an action plugin that was introduced in ansible 2.4 | 01:59 |
| waynr | i've copied that plugin to /etc/ansible/roles/plugins/actions but openstack-ansible doesn't seem to be picking it up | 01:59 |
| waynr | anyone have advice for this situation? | 01:59 |
| waynr | i've backported plugins from newer versions of ansible before with relatively minor problems | 02:00 |
| waynr | (like having to resolve missing imports) | 02:00 |
| waynr | the module i'm attempting to backport for openstack-ansible is wait_for_connection because for some reason some of the ansible lxc containers aren't available by ssh immediately after the openstack-ansible playbooks finish running | 02:02 |
| kaiokmo | waynr: which tag are you using? | 02:08 |
| waynr | newton-eol | 02:11 |
| waynr | since i'm stringing together a bunch of different playbook runs in my current effort i think i've found a workable alternative to the wait_for_connection module | 02:12 |
| waynr | i've isolated a single-task playbook that just echos 'hello' on each host in a while loop that will try running the playbook 60 times at intervales of 60 seconds each | 02:13 |
| waynr | not ideal | 02:13 |
| waynr | the problem i'm seeing in CI is hard to reproduce locally though | 02:14 |
| * waynr throws in the towel for the night | 02:20 | |
| *** mmercer has quit IRC | 02:29 | |
| *** mmercer has joined #openstack-ansible | 02:30 | |
| cloudnull | waynr which action plugin are you backporting? | 02:34 |
| cloudnull | you might need to add it to a path like /etc/ansible/roles/plugin_name | 02:35 |
| cloudnull | and then export the ANSIBLE_ACTION_PLUGIN variable to that action path | 02:35 |
| cloudnull | like so | 02:37 |
| cloudnull | https://github.com/openstack/openstack-ansible/blob/master/scripts/openstack-ansible.rc#L27 | 02:37 |
| cloudnull | if you already have the rc file on disk you could then just extend it | 02:37 |
| cloudnull | path /usr/local/bin/openstack-ansible.rc | 02:38 |
| cloudnull | the action plugin variable extends like a bash PATH | 02:38 |
| *** mmercer has quit IRC | 02:41 | |
| *** vnogin has joined #openstack-ansible | 02:42 | |
| *** vnogin has quit IRC | 02:47 | |
| *** nicolasbock has quit IRC | 03:33 | |
| *** udesale has joined #openstack-ansible | 03:59 | |
| *** dave-mccowan has quit IRC | 04:05 | |
| *** poopcat has quit IRC | 04:07 | |
| *** hamzy has joined #openstack-ansible | 04:15 | |
| *** Bhujay has joined #openstack-ansible | 04:18 | |
| *** gkadam has joined #openstack-ansible | 05:04 | |
| openstackgerrit | Kaio Kassiano Moura Oliveira proposed openstack/openstack-ansible-os_monasca master: Add support for openSUSE https://review.openstack.org/588584 | 05:05 |
| *** shyamb has joined #openstack-ansible | 05:12 | |
| ivve | morning | 05:41 |
| ivve | the recently changed ceph modification for gnocchi isn't working | 05:43 |
| ivve | https://docs.openstack.org/openstack-ansible-ceph_client/queens/configure-ceph.html#configure-os-gnocchi-with-ceph-client | 05:44 |
| ivve | was recently changed and and "s" was added to package(s) | 05:44 |
| ivve | to the var | 05:44 |
| ivve | now it exits as following | 05:45 |
| *** joabdearaujo has quit IRC | 05:45 | |
| *** kaiokmo has quit IRC | 05:45 | |
| ivve | fatal: [os-ctrl0402_gnocchi_container-364fc1bc]: FAILED! => {"attempts": 5, "changed": false, "failed": true, "module_stderr": "mesg: ttyname failed: Inappropriate ioctl for device\nTraceback (most recent call last):\n File \"/tmp/ansibl | 05:45 |
| ivve | e_MDIPVA/ansible_module_apt.py\", line 1059, in <module>\n main()\n File \"/tmp/ansible_MDIPVA/ansible_module_apt.py\", line 1037, in main\n allow_unauthenticated=allow_unauthenticated\n File \"/tmp/ansible_MDIPVA/ansible_module_a | 05:45 |
| ivve | pt.py\", line 485, in install\n pkgspec = expand_pkgspec_from_fnmatches(m, pkgspec, cache)\n File \"/tmp/ansible_MDIPVA/ansible_module_apt.py\", line 425, in expand_pkgspec_from_fnmatches\n pkgname_pattern, version = package_split( | 05:45 |
| ivve | pkgspec_pattern)\n File \"/tmp/ansible_MDIPVA/ansible_module_apt.py\", line 300, in package_split\n parts = pkgspec.split('=', 1)\nAttributeError: 'list' object has no attribute 'split'\n", "module_stdout": "", "msg": "MODULE FAILURE" | 05:45 |
| ivve | , "rc": 1} | 05:45 |
| *** joabdearaujo has joined #openstack-ansible | 05:47 | |
| *** rgogunskiy has joined #openstack-ansible | 06:01 | |
| *** jbadiapa has joined #openstack-ansible | 06:01 | |
| *** ivve has quit IRC | 06:02 | |
| *** shyamb has quit IRC | 06:03 | |
| *** shyamb has joined #openstack-ansible | 06:04 | |
| openstackgerrit | ShangXiao proposed openstack/ansible-role-python_venv_build master: Add release notes link to README https://review.openstack.org/574572 | 06:12 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_gnocchi master: Fix usage of "|" for tests https://review.openstack.org/598427 | 06:14 |
| *** ivve has joined #openstack-ansible | 06:14 | |
| *** Bhujay has quit IRC | 06:20 | |
| *** DanyC has joined #openstack-ansible | 06:23 | |
| *** shyamb has quit IRC | 06:27 | |
| *** shyamb has joined #openstack-ansible | 06:27 | |
| *** shyamb has quit IRC | 06:36 | |
| *** shyamb has joined #openstack-ansible | 06:41 | |
| *** vnogin has joined #openstack-ansible | 06:43 | |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_nova master: Fix usage of "|" for tests https://review.openstack.org/598434 | 06:44 |
| *** vnogin has quit IRC | 06:47 | |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_ironic master: Fix usage of https://review.openstack.org/598435 | 06:47 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_monasca master: Fix usage of https://review.openstack.org/598436 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_monasca-agent master: Fix usage of https://review.openstack.org/598437 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_cinder master: Fix usage of https://review.openstack.org/598438 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_glance master: Fix usage of https://review.openstack.org/598439 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_searchlight master: Fix usage of https://review.openstack.org/598440 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_zaqar master: Fix usage of https://review.openstack.org/598441 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_magnum master: Fix usage of https://review.openstack.org/598442 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_trove master: Fix usage of https://review.openstack.org/598443 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_designate master: Fix usage of https://review.openstack.org/598444 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_aodh master: Fix usage of https://review.openstack.org/598445 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_rally master: Fix usage of https://review.openstack.org/598446 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_masakari master: Fix usage of https://review.openstack.org/598447 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_heat master: Fix usage of https://review.openstack.org/598448 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_neutron master: Fix usage of https://review.openstack.org/598449 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_watcher master: Fix usage of https://review.openstack.org/598450 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_tempest master: Fix usage of https://review.openstack.org/598451 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_panko master: Fix usage of https://review.openstack.org/598452 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_zun master: Fix usage of https://review.openstack.org/598453 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_cloudkitty master: Fix usage of https://review.openstack.org/598454 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_keystone master: Fix usage of https://review.openstack.org/598455 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_sahara master: Fix usage of https://review.openstack.org/598456 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_tacker master: Fix usage of https://review.openstack.org/598457 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_ceilometer master: Fix usage of https://review.openstack.org/598458 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_congress master: Fix usage of https://review.openstack.org/598459 | 06:48 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_swift master: Fix usage of https://review.openstack.org/598460 | 06:49 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_barbican master: Fix usage of https://review.openstack.org/598461 | 06:49 |
| *** DanyC has quit IRC | 06:51 | |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_monasca master: Fix usage of "|" for tests https://review.openstack.org/598436 | 06:53 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_zun master: Fix usage of "|" for tests https://review.openstack.org/598453 | 06:53 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_cloudkitty master: Fix usage of "|" for tests https://review.openstack.org/598454 | 06:54 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_keystone master: Fix usage of "|" for tests https://review.openstack.org/598455 | 06:54 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_sahara master: Fix usage of "|" for tests https://review.openstack.org/598456 | 06:55 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_tacker master: Fix usage of "|" for tests https://review.openstack.org/598457 | 06:55 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_ceilometer master: Fix usage of "|" for tests https://review.openstack.org/598458 | 06:55 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_congress master: Fix usage of "|" for tests https://review.openstack.org/598459 | 06:55 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_swift master: Fix usage of "|" for tests https://review.openstack.org/598460 | 06:55 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_barbican master: Fix usage of "|" for tests https://review.openstack.org/598461 | 06:56 |
| openstackgerrit | zhulingjie proposed openstack/openstack-ansible-os_ironic master: Fix usage of "|" for tests https://review.openstack.org/598435 | 06:56 |
| *** dxiri has joined #openstack-ansible | 06:56 | |
| *** olivierb has joined #openstack-ansible | 07:00 | |
| *** luksky has joined #openstack-ansible | 07:04 | |
| *** shyamb has quit IRC | 07:12 | |
| *** shardy has joined #openstack-ansible | 07:16 | |
| *** shyamb has joined #openstack-ansible | 07:19 | |
| *** pcaruana has joined #openstack-ansible | 07:19 | |
| *** bjolo has joined #openstack-ansible | 07:28 | |
| *** shardy_ has joined #openstack-ansible | 07:31 | |
| *** shardy has quit IRC | 07:34 | |
| hwoarang | good morning. Could I get some reviews on https://review.openstack.org/#/c/592529/ please? | 07:38 |
| *** DanyC has joined #openstack-ansible | 07:39 | |
| *** shardy_ is now known as shardy | 07:39 | |
| *** Bhujay has joined #openstack-ansible | 07:41 | |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_sahara master: Add project source link to README https://review.openstack.org/598478 | 07:48 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_sahara master: Add project source link to README https://review.openstack.org/598478 | 07:54 |
| *** shyamb has quit IRC | 07:55 | |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_swift master: Add project source link to README https://review.openstack.org/598483 | 07:56 |
| *** luksky has quit IRC | 07:58 | |
| *** ostackz has quit IRC | 07:59 | |
| *** DanyC has quit IRC | 08:01 | |
| *** DanyC has joined #openstack-ansible | 08:01 | |
| *** DanyC has quit IRC | 08:06 | |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_tempest master: Remove duplicate link in README https://review.openstack.org/598485 | 08:07 |
| *** electrofelix has joined #openstack-ansible | 08:08 | |
| *** Bhujay has quit IRC | 08:17 | |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_masakari master: Add project source link to README https://review.openstack.org/598510 | 08:23 |
| *** luksky has joined #openstack-ansible | 08:28 | |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_barbican master: Add bugs url link to README https://review.openstack.org/598522 | 08:29 |
| *** fghaas has joined #openstack-ansible | 08:29 | |
| fghaas | Hi, I just looked into scripts/run-upgrade.sh and it doesn't seem to have seen an update since April. I have a trivial patch lined up that merely sets SOURCE_SERIES and TARGET_SERIES to queens/rocky, but I am guessing there would be other things that need fixing as well. I see no pending changes to run-upgrade.sh in https://review.openstack.org/#/q/status:open+project:openstack/openstack-ansible, so I thought I'd holler here t | 08:32 |
| fghaas | o see if anyone has already been working on something lest we needlessly duplicate efforts. | 08:32 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_ironic master: Add bugs url link to README https://review.openstack.org/598534 | 08:34 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_sahara master: Add bugs url link to README https://review.openstack.org/598546 | 08:38 |
| fghaas | In the event that no-one is actually working on the script, but the likes of odyssey4me or evrardjp are already aware of things they *know* need fixing, I'm obviously grateful for any input. | 08:39 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_masakari master: Add bugs url link to README https://review.openstack.org/598553 | 08:41 |
| *** olivierb_ has joined #openstack-ansible | 08:42 | |
| evrardjp | fghaas: well I didn't get the chance to work on that for a while... :( | 08:42 |
| fghaas | No worries at all. I'll go ahead and take a swing at it. :) | 08:43 |
| evrardjp | fghaas: we try to add the migration of the variables when they are happening, so I think we are good on that point | 08:43 |
| *** olivierb has quit IRC | 08:43 | |
| evrardjp | my concern would be the re-order of plays, but we haven't got a big change between queens and rocky that I remember would cause a problem in the re-ordering | 08:43 |
| evrardjp | all we did were respectful and not problematic for upgrades as far as I can tell. So right now we have to iron out bugs | 08:44 |
| evrardjp | but I don't think a big refactor is needed, per se, in the current state | 08:44 |
| evrardjp | ofc I have ideas of large refactor, but no urgency :) | 08:44 |
| fghaas | Right, and the Queens container cleanup only ever went into stable/queens, not master. Meaning, nothing to remove there. | 08:44 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_swift master: Add bugs url link to README https://review.openstack.org/598558 | 08:45 |
| fghaas | (assuming, as I hope is accurate, run-upgrade.sh shouldn't support a direct Pike→Rocky upgrade, which would make things much more complicated) | 08:45 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_swift master: Add bugs url link to README https://review.openstack.org/598558 | 08:46 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_ceilometer master: Add bugs url link to README https://review.openstack.org/598561 | 08:51 |
| *** shyamb has joined #openstack-ansible | 08:51 | |
| openstackgerrit | Florian Haas proposed openstack/openstack-ansible master: WIP: scripts: Fix run-upgrade.sh SOURCE_SERIES and TARGET_SERIES https://review.openstack.org/598562 | 08:52 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_tacker master: Add bugs url link to README https://review.openstack.org/598563 | 08:57 |
| odyssey4me | fghaas: thanks for taking a peek at that! | 09:00 |
| odyssey4me | hwoarang: I expect https://review.openstack.org/591672 needs a rebase? | 09:00 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_cloudkitty master: Add bugs url link to README https://review.openstack.org/598564 | 09:00 |
| odyssey4me | I could do with some reviews folks if you have a moment: https://goo.gl/k9cCfF | 09:01 |
| odyssey4me | Some of them unblock some of the roles which are failing patches. | 09:02 |
| hwoarang | odyssey4me: it does not let me rebase but i think that by abandoning the other patch will make it move forward. otherwse i can restore it and get the other one in | 09:03 |
| odyssey4me | hwoarang: you can rebase via CLI instead of gerrit | 09:03 |
| odyssey4me | or should be able to | 09:04 |
| hwoarang | let me see if 'recheck' will do the trick. i dont want to lose the precious votes :( | 09:04 |
| odyssey4me | rebase won't lose the votes | 09:04 |
| odyssey4me | it looks like it was based on the abandoned patch, which is why it's not moving ahead | 09:04 |
| odyssey4me | I can rebase for you if you like? | 09:04 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_watcher master: Add bugs url link to README https://review.openstack.org/598566 | 09:05 |
| hwoarang | i am trying to find out how to rebase from cli | 09:05 |
| odyssey4me | git rebase -i HEAD~3; then comment out the patch you don't want; then git rebase origin/master | 09:06 |
| hwoarang | i guess i need a new branch with only this particular patch there, and then review again | 09:06 |
| hwoarang | yeah ok | 09:06 |
| odyssey4me | well, you can just git review -d <review id> first | 09:06 |
| noonedeadpunk | evrardjp: hi. have a minute? I found it impossible to use find like {{ horizon_lib_dir }}/*/enabled/ here https://review.openstack.org/#/c/593092/ | 09:06 |
| openstackgerrit | Markos Chandras (hwoarang) proposed openstack/openstack-ansible-plugins master: strategy: linear: Inspect the delegated host on delegated tasks. https://review.openstack.org/591672 | 09:06 |
| odyssey4me | hwoarang: I guess that patch is important for rocky too? any earlier branches? | 09:07 |
| hwoarang | rocky for sure | 09:07 |
| hwoarang | other branches are on older ansible and we did not have the problem there | 09:07 |
| odyssey4me | ah ok | 09:07 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_panko master: Add bugs url link to README https://review.openstack.org/598567 | 09:08 |
| odyssey4me | would it make sense to add https://review.openstack.org/#/c/592529/2 on top of that patch, so that there's an extended test for it? | 09:08 |
| noonedeadpunk | evrardjp: so it seems, that my first try was almost the best effort. But I've added ternary to defaults as odyssey4me offered | 09:08 |
| evrardjp | noonedeadpunk: yes the ternary could be part of a different patch, so that it doesn't mix | 09:08 |
| evrardjp | and it is less confusing | 09:09 |
| evrardjp | noonedeadpunk: but what is the problem? | 09:09 |
| evrardjp | You just have to extend the regexp | 09:09 |
| evrardjp | doesn't that work? | 09:09 |
| noonedeadpunk | evrardjp: the problem is, that find doesn't accept * inside of the path | 09:09 |
| evrardjp | oh it's just the basenames | 09:09 |
| evrardjp | I see | 09:09 |
| evrardjp | but paths is a list | 09:10 |
| noonedeadpunk | yep. so then I need to specify a list of paths, which isn't really better, than setting them in vars | 09:10 |
| evrardjp | well, one is dynamically found | 09:11 |
| evrardjp | wait, it's also recursive, so does this matter? | 09:11 |
| odyssey4me | hmm, where does horizon_panel_enable_state come from noonedeadpunk ? | 09:11 |
| evrardjp | that's old odyssey4me | 09:11 |
| odyssey4me | https://review.openstack.org/#/c/593092/4/vars/main.yml | 09:12 |
| odyssey4me | oh I see you mean it's an old var | 09:12 |
| hwoarang | we also need https://review.openstack.org/#/c/598195/ for the opnfv releases if anyone has the time :) | 09:12 |
| evrardjp | odyssey4me: yes | 09:12 |
| odyssey4me | so, as it stands, that patch will break the repo build - so we can't merge this until I reform the repo build for horizon at least | 09:13 |
| noonedeadpunk | recursion works only in the provided path, but I don't think that it's a good idea to set path as just {{ horizon_lib_dir }} and descend up to enable with recursion... | 09:13 |
| evrardjp | noonedeadpunk: no I mean to get full automation and reduce variable, we will have to do multiple finds | 09:14 |
| odyssey4me | hwoarang: done | 09:14 |
| evrardjp | 1) find all the "enabled", "conf", "local settings" folders | 09:14 |
| evrardjp | 2) use that as a list for the find files to copy by listing all their files, excepted __init__.py | 09:14 |
| evrardjp | 3) profit! | 09:15 |
| odyssey4me | evrardjp noonedeadpunk I wonder, does horizon not just auto-enable anything installed? | 09:15 |
| evrardjp | odyssey4me: the docs say otherwise | 09:15 |
| evrardjp | the docs say you must link | 09:15 |
| odyssey4me | if not, perhaps we should engage with the teams to make that happen - because having to implement stuff like this in deployment tooling is quite silly | 09:15 |
| evrardjp | that's why we also had ppl coming around saying "this thing doesn't work" | 09:15 |
| noonedeadpunk | but still setting a path for every panel is needed | 09:15 |
| odyssey4me | ok, happy to go for it right now, but perhaps we should PTG discussion this | 09:15 |
| evrardjp | odyssey4me: well, when you use distro package, nothing to do, because distro put the things at the right place when installing the package | 09:16 |
| * odyssey4me adds it to the etherpad | 09:16 | |
| noonedeadpunk | agree - distro package enables everything needed | 09:16 |
| evrardjp | noonedeadpunk: what do you mean? | 09:16 |
| evrardjp | I mean this is for only source installs, and it is okay to me to do two finds... I am not sure what the problem is | 09:17 |
| evrardjp | they don't have all the same structure? | 09:17 |
| evrardjp | I mean this could potentially remove ALL the custom things in horizon -- it's a big positive change if we wire everything automatically | 09:18 |
| odyssey4me | ok I now see what horizon_panel_enable_state is - it specifies 'link' or 'file' depending on whether it's a source-based install or a distro-based install | 09:18 |
| noonedeadpunk | evrardjp: I mean, that I still need _horizon_panels, as without list of panels I can't get find working, as asterisk is not an option | 09:18 |
| evrardjp | ppl wanting their own dashboard can just package on pip and install it | 09:18 |
| *** vnogin has joined #openstack-ansible | 09:19 | |
| evrardjp | noonedeadpunk: that's the part I don't get | 09:19 |
| evrardjp | if they are not installed, they will not be found by the find module | 09:19 |
| noonedeadpunk | so you mean just to specify all path for all available dashboards and get rid of _horizon_active_panels? | 09:21 |
| fghaas | odyssey4me: I ain't done much yet :) | 09:21 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_trove master: Add bugs url link to README https://review.openstack.org/598570 | 09:21 |
| noonedeadpunk | yep, it's an option, and will simplify a bit. but isn't explicit providing path of active dashboards better? | 09:22 |
| noonedeadpunk | and yes, it can't be merged unless venv won't contain all dashboards | 09:22 |
| odyssey4me | fghaas: soo... I did have another topic I thought might interest you and will likely have quite a big impact on the training material you have | 09:23 |
| * fghaas is all ears | 09:23 | |
| *** vnogin has quit IRC | 09:23 | |
| fghaas | nspawn? | 09:23 |
| odyssey4me | fghaas: yes, exactly | 09:23 |
| fghaas | Thought so. | 09:23 |
| odyssey4me | fghaas: I'm interested to know what sort of lead time you guys would need to transition to nspawn being the focal point of the training. One cycle, or two? | 09:24 |
| odyssey4me | We'd like to deprecate the usage of LXC at some point, and set a date for its removal - but we know that it has a broad impact and want to make sure we make the transition as smooth as possible. | 09:24 |
| fghaas | Well I read https://docs.openstack.org/releasenotes/openstack-ansible/rocky.html#b2 in such a way that nspawn wasn't the default for Rocky, yet. Did I misunderstand? | 09:25 |
| odyssey4me | For rocky, consider it experimental at best. | 09:25 |
| fghaas | Exactly. | 09:25 |
| noonedeadpunk | odyssey4me: actually I don't know if horizon_panel_enable_state really needed... As state - file is pretty strage for panels... | 09:25 |
| odyssey4me | But rocky gives the community a chance to try it out and improve it. | 09:25 |
| fghaas | odyssey4me: "cycle" being an OpenStack release cycle in this context? | 09:26 |
| odyssey4me | noonedeadpunk: I kinda agree - I would think that for distro packages, we should just skip the auto-linking altogether. The linking tasks should be source-based builds only IMO. | 09:26 |
| odyssey4me | fghaas: yes | 09:26 |
| fghaas | Kay. We refresh our courses on a monthly basis, so the nspawn transition for the course material would need to happen within one month. So, definitely one OpenStack release cycle, not two. | 09:27 |
| odyssey4me | fghaas: so rocky is experimental, and we'll likely be porting patches back to improve it from stein... stein we'd like to aim for it to consistently work in gates... and maybe also change all the docs to make it the primary use-case | 09:27 |
| odyssey4me | fghaas: wow, that's fast-moving :) | 09:27 |
| odyssey4me | fghaas: ok, so in terms of what we can do to help - could you add some notes to https://etherpad.openstack.org/p/osa-stein-ptg on the topic 'The path to deprecating LXC and making nspawn THE container back-end we use' | 09:28 |
| fghaas | Actually it's not, you folks have made our jobs very easy. We basically change the training material to point to the latest OSA point release, do a test run, and that's the whole refresh (in most months). | 09:28 |
| odyssey4me | Well, that's good to know. :) | 09:29 |
| fghaas | For September we add a Rocky upgrade as a bonus lab, in October we plan to start on Queens and then have the usual upgrade run into Rocky. | 09:30 |
| fghaas | Is the lxc→nspawn tradition supported in-place? | 09:30 |
| odyssey4me | My hope is that we can, in stein, switch the primary/default back-end but leave LXC in place. Then in T we remove LXC. That effectively gives deployers 3 major releases to transition. | 09:30 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_designate master: Add bugs url link to README https://review.openstack.org/598575 | 09:30 |
| fghaas | I mean, should I be able to take an LXC-based deployment and move it to nspawn? I'd be happy to add that as our new bonus lab. | 09:31 |
| odyssey4me | oh no, there's still a lot of work to do just to get a greenfield working reliably - upgrades are not even yet looked at | 09:31 |
| noonedeadpunk | oh, just offtopic question - are bionic now fully supported by OSA in rocky? As I see tests are passing and everything is supposed to work. As I was thinking about ubuntu upgrade to bionic and then upgrade from queens to rocky | 09:31 |
| *** DanyC has joined #openstack-ansible | 09:31 | |
| odyssey4me | well, technically, it is possible to transition - there's just no real automation or documentation for it | 09:31 |
| odyssey4me | this is exactly what we need to get done this cycle | 09:32 |
| fghaas | odyssey4me: well wouldn't it be much the same thing as what we collaborated on for the Queens container cleanup? | 09:32 |
| odyssey4me | noonedeadpunk: bionic's not quite there yet - and it's only in master/stein right now | 09:32 |
| fghaas | Deploy new nspawn containers, have a playbook that nukes the lxc ones? | 09:32 |
| noonedeadpunk | odyssey4me: got it. thanks | 09:33 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_searchlight master: Add bugs url link to README https://review.openstack.org/598576 | 09:33 |
| odyssey4me | fghaas: yep, it'll go something like that - a little different though... it's nuke the lxc container, change the container back-end for it, then recreate it and it's now nspawn | 09:33 |
| fghaas | Hmm. That's fine for an HA setup where you have 3 instances of each container, but would disrupt single-instance configurations in that the API would be unavailable for some time. | 09:34 |
| odyssey4me | fghaas: I figured that you may have a bunch of words in slides and materials about LXC, and may need to prep some new ones for nspawn | 09:34 |
| odyssey4me | fghaas: oh yeah, in that instance it might be better to add a new nspawn and remove the old lxc... but then IP address and agent names change - so each has their price | 09:35 |
| fghaas | I'm actually thinking about adding API service HA to the course; makes for a more realistic environment that way. | 09:35 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_zaqar master: Add bugs url link to README https://review.openstack.org/598577 | 09:35 |
| odyssey4me | fghaas: well, even on an AIO you can increase affinity to make each service have multiple containers | 09:36 |
| odyssey4me | so perhaps that'd be a useful thing to abuse for the transition | 09:36 |
| fghaas | Not worried about the slides much. It's all Markdown, so it's not like we have to manually edit Google Slides or Impress or (shudder) PowerPoint. | 09:36 |
| odyssey4me | oh sure, but nspawn means research | 09:36 |
| *** hamzy_ has joined #openstack-ansible | 09:37 | |
| *** hamzy has quit IRC | 09:37 | |
| *** shyamb has quit IRC | 09:37 | |
| odyssey4me | perhaps it'd be handy for us to express in the docs why we're transitioning | 09:37 |
| *** vnogin has joined #openstack-ansible | 09:37 | |
| fghaas | I know about the affinity bit and we use it for rabbitmq and Galera to have _some_ HAish things, but I'm thinking more about actually deploying 3 control nodes because really, no person in their right mind should deploy fewer | 09:37 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_gnocchi master: Add bugs url link to README https://review.openstack.org/598579 | 09:38 |
| openstackgerrit | ShangXiao proposed openstack/ansible-role-python_venv_build master: Add bugs url link to README https://review.openstack.org/598580 | 09:41 |
| openstackgerrit | ShangXiao proposed openstack/ansible-role-python_venv_build master: Add source code and bugs url link to README https://review.openstack.org/598580 | 09:46 |
| openstackgerrit | Merged openstack/openstack-ansible-plugins master: strategy: linear: Inspect the delegated host on delegated tasks. https://review.openstack.org/591672 | 09:51 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-lxc_hosts master: Add bugs url link to README https://review.openstack.org/598582 | 09:51 |
| *** priteau has joined #openstack-ansible | 09:52 | |
| odyssey4me | hwoarang: https://review.openstack.org/591672 is ready for porting back when you are ;) | 09:54 |
| openstackgerrit | Markos Chandras (hwoarang) proposed openstack/openstack-ansible-plugins stable/rocky: strategy: linear: Inspect the delegated host on delegated tasks. https://review.openstack.org/598584 | 09:55 |
| hwoarang | awesome | 09:55 |
| -openstackstatus- NOTICE: Jobs using devstack-gate (legacy devstack jobs) have been failing due to an ara update. We use now a newer ansible version, it's safe to recheck if you see "ImportError: No module named manager" in the logs. | 09:56 | |
| hwoarang | i am curious about https://review.openstack.org/#/c/598249/ - do we need to do anything special when bumping ansible? | 09:56 |
| hwoarang | this seems to fail due to config_template | 09:56 |
| *** hamzaachi has joined #openstack-ansible | 09:57 | |
| odyssey4me | hwoarang: Yeah, I did https://review.openstack.org/592317 and that works fine - so either we've got some sort of conflict in the config_template loading, or it can't find config_template because we're not settings the paths to help it find the libraries. | 09:57 |
| odyssey4me | For role tests, we put things in default paths. For the integrated build, config_template is in the roles/ path and not a standard location... which is why I suspect that to be the problem here - perhaps something changed upstream to make what used to work, not work any more. | 09:58 |
| hwoarang | strange that it breaks in a bugfix release | 09:58 |
| odyssey4me | Strange? This is Ansible, you know. Bugfixes often 'correct' behaviours which were unintended, but were relied on. | 09:59 |
| hwoarang | hehe | 09:59 |
| odyssey4me | I haven't had much chance to try and figure it out - but that's my suspicioin so far. | 10:00 |
| hwoarang | ok i will look into that | 10:00 |
| ivve | gnocchi seems to have trouble installing with ceph as backend.. anyone had any tests around that? | 10:02 |
| openstackgerrit | ShangXiao proposed openstack/openstack-ansible-os_horizon master: Add bugs url link to README https://review.openstack.org/598586 | 10:02 |
| ivve | env is ubuntu 16.04 on deploy & hosts with queens | 10:02 |
| ivve | with the recommended user_vars it doesn't work, without them it fails importing rados module in python | 10:03 |
| ivve | looks like an ansible issue | 10:03 |
| ivve | ill get a pastebin | 10:03 |
| *** shyamb has joined #openstack-ansible | 10:05 | |
| odyssey4me | fghaas: perhaps you have a moment to test and feedback on this patch? https://review.openstack.org/#/c/597977/1/scripts/run-upgrade.sh | 10:17 |
| *** Bhujay has joined #openstack-ansible | 10:21 | |
| *** DanyC has quit IRC | 10:22 | |
| *** shyamb has quit IRC | 10:23 | |
| *** shyamb has joined #openstack-ansible | 10:23 | |
| ivve | http://paste.openstack.org/show/729219/ | 10:23 |
| *** Bhujay has quit IRC | 10:26 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-os_monasca master: Add Ubuntu Bionic 18.04 support https://review.openstack.org/586932 | 10:30 |
| *** nicolasbock has joined #openstack-ansible | 10:33 | |
| *** dave-mccowan has joined #openstack-ansible | 10:33 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-os_glance master: Add ability to configure cross-origin requests https://review.openstack.org/579224 | 10:41 |
| openstackgerrit | Markos Chandras (hwoarang) proposed openstack/openstack-ansible master: Update ansible to 2.5.8 https://review.openstack.org/592315 | 10:57 |
| ivve | bug is here: https://bugs.launchpad.net/openstack-ansible/+bug/1775873 | 10:57 |
| openstack | Launchpad bug 1775873 in openstack-ansible "gnocchi missing configuration to use ceph backend" [Undecided,Confirmed] | 10:57 |
| *** shyamb has quit IRC | 10:57 | |
| ivve | it worked in pike | 10:58 |
| *** DanyC has joined #openstack-ansible | 10:58 | |
| noonedeadpunk | hm, it seems, that it works in queens as well... As I don't remeber hitting this bug | 11:00 |
| ivve | noonedeadpunk: can you paste your user_variable settings for gnocchi | 11:01 |
| ivve | i'd like to compare as there seems to be changes on how it should be configured | 11:01 |
| ivve | (i used the same as for pike) but added an "s" to the packages variable as stated in the docs | 11:01 |
| ivve | and that might not suffice.. | 11:01 |
| openstackgerrit | Merged openstack/openstack-ansible-plugins master: Fix config_template to work with Ansible 2.6 https://review.openstack.org/588322 | 11:02 |
| noonedeadpunk | ivve http://paste.openstack.org/show/729223/ | 11:04 |
| ivve | thanks noonedeadpunk | 11:05 |
| openstackgerrit | Merged openstack/openstack-ansible-rabbitmq_server master: Cleanup generation of rabbitmq server data https://review.openstack.org/582150 | 11:06 |
| noonedeadpunk | ivve: But I'm now on 17.0.5 | 11:06 |
| ivve | ah this is 17.1.? | 11:07 |
| ivve | lemme check | 11:07 |
| ivve | latest in stable branch anyways | 11:07 |
| noonedeadpunk | ivve: As for me using tag may be better solution | 11:08 |
| ivve | aye | 11:10 |
| ivve | i think i will head that way | 11:11 |
| *** udesale has quit IRC | 11:12 | |
| *** shyamb has joined #openstack-ansible | 11:33 | |
| noonedeadpunk | ivve: but it seems, that this bug is really valid, probably I've just forgotten or fixed it just for myself... | 11:40 |
| *** johnny101 has quit IRC | 11:45 | |
| *** Taseer has quit IRC | 11:47 | |
| hwoarang | odyssey4me: what do you think of https://review.openstack.org/#/c/592315/ ? | 11:53 |
| hwoarang | odyssey4me: tldr; ansible.cfg can't be in an insecure location anymore (ie tests/ansible.cfg) | 11:53 |
| hwoarang | the reason it works on openstack-ansible-tests is because we put it in $HOME which is not world writable | 11:54 |
| hwoarang | fun stuff | 11:54 |
| odyssey4me | hmm, ok - what's "insecure" about the current location? | 11:56 |
| odyssey4me | oh, that folder is world writable? I wonder why that is? | 11:57 |
| odyssey4me | It doesn't look world writable to me on my laptop... | 11:58 |
| *** kaiokmo has joined #openstack-ansible | 11:58 | |
| hwoarang | maybe it is in the ci | 11:59 |
| odyssey4me | well, do we even need ansible.cfg? | 12:00 |
| hwoarang | [WARNING] Ansible is in a world writable directory (/home/zuul/src/git.openstack.org/openstack/openstack-ansible/scripts), ignoring it as an ansible.cfg source. | 12:00 |
| *** suggestable has joined #openstack-ansible | 12:00 | |
| odyssey4me | we bootstrap ansible before that, set a bunch of env vars to tell ansible where things are | 12:00 |
| *** gkadam has quit IRC | 12:00 | |
| hwoarang | true but we also define where our plugins are | 12:00 |
| odyssey4me | ah, yeah - for zuul it likely is | 12:00 |
| odyssey4me | we're using /opt/ansible-runtime/bin/ansible-playbook to avoid using the wrapper | 12:01 |
| odyssey4me | I wonder why. | 12:01 |
| odyssey4me | probably only to avoid using the dynamic inventory, which I think we can solve another way | 12:02 |
| odyssey4me | if we, for bootstrap-aio, set ANSIBLE_INVENTORY to something nonsense, then it won't use the dynamicinventory | 12:03 |
| odyssey4me | but we get all the other env vars we set | 12:03 |
| odyssey4me | then we can unset the inventory after bootstrap-aio so that it uses the defaults later | 12:04 |
| odyssey4me | mind if I do the adjustments to the patch? | 12:04 |
| hwoarang | do whatever you want. at least we know the problem right now | 12:04 |
| *** shyamb has quit IRC | 12:08 | |
| *** shyamb has joined #openstack-ansible | 12:08 | |
| *** ansmith has joined #openstack-ansible | 12:15 | |
| *** schwicht has joined #openstack-ansible | 12:17 | |
| *** schwicht has quit IRC | 12:21 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible master: Update ansible to 2.5.8 https://review.openstack.org/592315 | 12:28 |
| odyssey4me | hwoarang: yeah, thanks for figuring that out - good find! | 12:28 |
| *** schwicht has joined #openstack-ansible | 12:29 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible master: Update ansible to 2.5.8 https://review.openstack.org/592315 | 12:30 |
| *** shardy has quit IRC | 12:35 | |
| noonedeadpunk | odyssey4me: stupid question - does vars/source_install.yml includes only for installation from source, or always? | 12:40 |
| hwoarang | it should only be included for source installs | 12:41 |
| noonedeadpunk | hwoarang: ok, thanks | 12:41 |
| *** shyamb has quit IRC | 12:41 | |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_horizon master: Re-written dashboards linking. https://review.openstack.org/593092 | 12:44 |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_horizon master: Re-written dashboards linking. https://review.openstack.org/593092 | 12:47 |
| *** schwicht has quit IRC | 12:48 | |
| noonedeadpunk | odyssey4me hwoarang evrardjp what do you think about this ^ ? | 12:49 |
| evrardjp | noonedeadpunk: I like where you are heading | 12:54 |
| evrardjp | far cleaner already | 12:54 |
| openstackgerrit | Merged openstack/openstack-ansible-plugins stable/rocky: strategy: linear: Inspect the delegated host on delegated tasks. https://review.openstack.org/598584 | 12:54 |
| evrardjp | noonedeadpunk: what's the reason why _horizon_panels cannot be found? | 12:54 |
| evrardjp | my concern is that users with new panels will have to override _horizon_panels or add stuff to extra panels. Which is fine, but it's an extra step that can probably be skipped | 12:56 |
| evrardjp | noonedeadpunk: we now have to wait for odyssey4me 's patches to make this work with the repo build | 12:57 |
| odyssey4me | yep, working on it as we speak | 12:57 |
| *** EmilienM is now known as EvilienM | 12:57 | |
| odyssey4me | I've made the python_venv_build role changes, now just need to make tests work with them. | 12:57 |
| odyssey4me | I'm removing all the venv packaging - it makes the role so much simpler. | 12:58 |
| evrardjp | folks I'd like a few reviews, https://review.openstack.org/#/c/587544/ https://review.openstack.org/#/c/593032/ https://review.openstack.org/#/c/593051/ https://review.openstack.org/#/c/596761/ | 12:59 |
| evrardjp | https://review.openstack.org/#/c/597012/ | 12:59 |
| evrardjp | odyssey4me: already did that, so it's for others. | 12:59 |
| evrardjp | I will issue RC3 this week-end | 12:59 |
| odyssey4me | evrardjp: I'll trade you for https://goo.gl/k9cCfF :) | 12:59 |
| evrardjp | when those are in | 12:59 |
| *** ansmith has quit IRC | 13:01 | |
| *** elmaciej has joined #openstack-ansible | 13:01 | |
| *** elmaciej has quit IRC | 13:01 | |
| evrardjp | I can't vote on MNAIO I never touched it | 13:01 |
| *** tosky has joined #openstack-ansible | 13:02 | |
| waynr | cloudnull: thanks for the advice; before running the playbook I actually copied the plugin directly to /etc/ansible/roles/plugins/action and it still wasn't detected | 13:02 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible master: Update ansible to 2.5.8 https://review.openstack.org/592315 | 13:03 |
| waynr | the action module i wanted to use is called "wait_for_connection" | 13:03 |
| *** aicilamd has joined #openstack-ansible | 13:04 | |
| odyssey4me | evrardjp: yep, that's fair enough | 13:04 |
| odyssey4me | tyvm | 13:04 |
| hwoarang | me neither. i am trying to make some sense based on commit msg and patch | 13:05 |
| hwoarang | odyssey4me: i am getting the mnaio only works with VM right? | 13:05 |
| hwoarang | i mean pxe-ing VMs and not real nodes | 13:05 |
| hwoarang | that's what i gather from the readme file | 13:06 |
| *** ansmith has joined #openstack-ansible | 13:06 | |
| odyssey4me | hwoarang: MNAIO installs on a physical host, then boots VM's setup with all their networking to simulate multiple VM's on a single physical node | 13:06 |
| odyssey4me | it's quite handy for multi-node testing | 13:06 |
| hwoarang | ok i see | 13:07 |
| odyssey4me | we're using the file-based back-end for the VM's to save the VM images off the host, so that we can re-use them in later tests | 13:07 |
| hwoarang | you meant 'multiple VM nodes on a single physical node right' | 13:07 |
| odyssey4me | that way in CI we can to deployment validation in one stage, and subsequent tests in a totally different stage - and in parallel | 13:07 |
| hwoarang | like the VMs are pretending to be openstack nodes | 13:07 |
| hwoarang | where you build openstack on top of them | 13:07 |
| odyssey4me | yes | 13:07 |
| hwoarang | ok | 13:07 |
| hwoarang | good | 13:08 |
| odyssey4me | if you have a host with enough ram/disk, then it's quite nifty | 13:08 |
| *** aicilamd has quit IRC | 13:08 | |
| hwoarang | is exactly what we do in opnfv, but we use bifrost (ironic) to drive the pxe boot | 13:08 |
| odyssey4me | fair enough, that may be interesting to look at some time in the future when I have some time | 13:10 |
| odyssey4me | in fact, at the ptg I'd like to compare notes around how we do things - maybe we can converge some things | 13:10 |
| odyssey4me | I suspect we may be solving similar problems. :) | 13:10 |
| *** schwicht has joined #openstack-ansible | 13:11 | |
| odyssey4me | hwoarang: oh, you're not at the ptg are you? | 13:11 |
| noonedeadpunk | sorry, had a meeting so asked for a review and went out. | 13:13 |
| hwoarang | odyssey4me: no i am not but others from opnfv may be | 13:14 |
| odyssey4me | hwoarang: alright, if I don't manage to chat to someone there then I'll see what I can do to come and visit you up north :p | 13:15 |
| hwoarang | ;p | 13:16 |
| noonedeadpunk | evrardjp: didn't understand your concern about extra panels. I'm not sure that _horizon_panels are found, as they are placed inside vars/source_install.yml, which is not supposed for include in case of package installation | 13:17 |
| noonedeadpunk | oh, probably I've found some solution | 13:19 |
| evrardjp | noonedeadpunk: ok :) | 13:20 |
| evrardjp | noonedeadpunk: the horizon_panels IMO doesn't _need_ to exist | 13:20 |
| evrardjp | noonedeadpunk: the idea is , for a deployer,you want to install a pip package containing your dashboard, and it automatically gets wired. Like if you were installing from packages | 13:21 |
| evrardjp | so we need to be smarter instead of force any wiring on the deployers | 13:21 |
| noonedeadpunk | evrardjp: yes, but as you see I've moved list here https://review.openstack.org/#/c/593092/6/vars/source_install.yml | 13:22 |
| evrardjp | so for me finding the folder "enabled" "conf" "local settings" inside "horizon_lib_dir" should do the trick | 13:22 |
| guilhermesp | mornings team :) | 13:22 |
| adcpdk | Hi Everyone! I have just destroyed all the containers and wanted to proceed with the new installation of openstack ansible AppendixA testing environment | 13:23 |
| evrardjp | noonedeadpunk: that should be in main IMO | 13:23 |
| evrardjp | to avoid having undefined | 13:23 |
| evrardjp | until we remove it completely :) | 13:23 |
| adcpdk | For some reason while running setup-hosts playbook I get the following messages from my controller node | 13:23 |
| adcpdk | Failed to load config for infra1_cinder_api_container-7c46a521 | 13:23 |
| adcpdk | any suggestions? | 13:23 |
| noonedeadpunk | evrardjp: but ansible find module currently doesn't allow to provide such find, so I have to specify this list of dirs somewhere.... | 13:24 |
| evrardjp | noonedeadpunk: are you sure? Because for me the docs say otherwise | 13:25 |
| evrardjp | noonedeadpunk: file_type: directory, path: horizon_lib_dir, recurse: yes, patterns: ['conf', 'enabled', 'local_settings'] | 13:26 |
| noonedeadpunk | evrardjp: so you mean not to check files inside this directories for some mess and link everything, which is contained there? | 13:29 |
| adcpdk | cloudnull, can you assist please? http://paste.openstack.org/show/729234/ | 13:29 |
| noonedeadpunk | I really think, that basic filenames check inside this directories should be held... | 13:29 |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_horizon master: Re-written dashboards linking. https://review.openstack.org/593092 | 13:31 |
| openstackgerrit | Merged openstack/openstack-ansible-nspawn_hosts master: Simplify inserting apt keys into nspawn image https://review.openstack.org/596555 | 13:32 |
| openstackgerrit | Dmitriy Rabotjagov (noonedeadpunk) proposed openstack/openstack-ansible-os_horizon master: Re-written dashboards linking. https://review.openstack.org/593092 | 13:32 |
| evrardjp | noonedeadpunk: nope I mean two tasks | 13:32 |
| evrardjp | :) | 13:32 |
| noonedeadpunk | oh, finally I understood. How slow I am... | 13:32 |
| evrardjp | first one finds the folders that contain said files, second find files inside previously found folders :) | 13:32 |
| evrardjp | noonedeadpunk: you don't have the odyssey4me 's translator | 13:33 |
| evrardjp | I just don't express myself well in english, as I have a tendency to write the strict minimum :p | 13:33 |
| evrardjp | morning guilhermesp | 13:34 |
| noonedeadpunk | my english is extremely far from even a good one, so probably we just use different english) | 13:35 |
| noonedeadpunk | based on your experience with openstack, when it will be safe to try upgrade to the new release? | 13:37 |
| guilhermesp | what do we need besides fix os_nova and os_cinder to get this merged to the rest of the roles so we can backport it? https://review.openstack.org/#/q/topic:smart-sources+(status:open+OR+status:merged) | 13:37 |
| *** lbragstad is now known as elbragstad | 13:37 | |
| *** aicilamd has joined #openstack-ansible | 13:39 | |
| *** luksky has quit IRC | 13:40 | |
| noonedeadpunk | btw, I really very dissapointed with ansible's find module. It' doesn't find anything in case of the following regexp: "^_[0-9]{2,4}_.*.py$" But it feels okey with ^(_[0-9]{4}_.*.py|_[0-9]{2}_.*.py)$ So it looks like a bug, which I probably should report... | 13:41 |
| openstackgerrit | Merged openstack/openstack-ansible-lxc_hosts master: Make apt key import for Ubuntu a uniform process https://review.openstack.org/596557 | 13:44 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/ansible-role-python_venv_build master: Simplify builds/installs by eliminating venv packaging https://review.openstack.org/598957 | 13:44 |
| odyssey4me | noonedeadpunk: something that may be handy is to try the regex out with sample data on https://regex101.com/ - make sure to switch to the python flavor | 13:45 |
| odyssey4me | once you have a regex with data setup, you can also save it and share a url for others to look at | 13:46 |
| *** aicilamd has quit IRC | 13:46 | |
| openstackgerrit | Merged openstack/openstack-ansible-os_gnocchi master: Upgrade pip/setuptools/wheel in role test venv https://review.openstack.org/596885 | 13:47 |
| noonedeadpunk | odyssey4me: yeah, I've checked it with this resource) | 13:47 |
| noonedeadpunk | things I tried to catch are "_1610_project_orchestration_panel.py" and "_12_toggle_data_upload_max_number_fields.py". And both of them were catched by second condition and nothing by first | 13:49 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/ansible-role-python_venv_build master: Simplify builds/installs by eliminating venv packaging https://review.openstack.org/598957 | 13:54 |
| *** aicilamd has joined #openstack-ansible | 13:55 | |
| odyssey4me | evrardjp: got a min to review https://review.openstack.org/592502 & https://review.openstack.org/598210 & https://review.openstack.org/598402 & https://review.openstack.org/598401 | 13:57 |
| evrardjp | all those small things | 13:59 |
| *** hamzaachi has quit IRC | 14:00 | |
| *** aicilamd has quit IRC | 14:00 | |
| noonedeadpunk | odyssey4me: btw, I still need help with masakari, so if you once will have some time for it... :) | 14:00 |
| odyssey4me | noonedeadpunk: yep, it's next on my hit list - I haven't forgotten :) | 14:02 |
| noonedeadpunk | odyssey4me: ok, just ping me then - I'm almost always here at business time | 14:03 |
| openstackgerrit | Merged openstack/openstack-ansible-os_zun master: Remove deprecated Disk/RamFilter usage https://review.openstack.org/597201 | 14:05 |
| openstackgerrit | Merged openstack/openstack-ansible-os_gnocchi master: Use the TESTING_BRANCH env var for constraints https://review.openstack.org/592588 | 14:06 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-lxc_hosts stable/rocky: Make apt key import for Ubuntu a uniform process https://review.openstack.org/598964 | 14:07 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-nspawn_hosts stable/rocky: Simplify inserting apt keys into nspawn image https://review.openstack.org/598965 | 14:07 |
| adcpdk | did anyone experience issues with creating the containers with the following error: Unknown configuration key "lxc.pts" | 14:10 |
| odyssey4me | adcpdk: are you trying to install on bionic perchance? | 14:10 |
| adcpdk | only the following ones: http://paste.openstack.org/show/729237/ | 14:11 |
| adcpdk | I have destroyed my enviromnent completely and trying to bootstrap from clean install | 14:12 |
| odyssey4me | adcpdk: bionic support is not finished yet for master, and will not work for any stable branches yet | 14:12 |
| *** aicilamd has joined #openstack-ansible | 14:13 | |
| adcpdk | odyssey4me it's not bionic, it's 16.04 | 14:14 |
| odyssey4me | adcpdk: oh, then I haven't seen that :/ sorry | 14:14 |
| adcpdk | Linux node01 4.4.0-131-generic #157-Ubuntu SMP Thu Jul 12 15:51:36 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux | 14:14 |
| adcpdk | root@node01:~# cat /etc/lsb-release | 14:14 |
| adcpdk | DISTRIB_ID=Ubuntu | 14:14 |
| adcpdk | DISTRIB_RELEASE=16.04 | 14:14 |
| adcpdk | DISTRIB_CODENAME=xenial | 14:14 |
| adcpdk | DISTRIB_DESCRIPTION="Ubuntu 16.04.5 LTS" | 14:14 |
| adcpdk | I think it's related to the new lxc3.0 | 14:14 |
| odyssey4me | if Ubuntu has ported lxc3 back to Xenial, then we're going to have to port a bunch of things back into stable branches... that's *not* going to be fun | 14:16 |
| adcpdk | ii lxc 3.0.1-0ubuntu1~16.04.2 all Transitional package - lxc -> lxc-utils | 14:17 |
| adcpdk | so the solution is to downgrade to the previous lxc version | 14:17 |
| adcpdk | ? | 14:17 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/ansible-role-python_venv_build master: Simplify builds/installs by eliminating venv packaging https://review.openstack.org/598957 | 14:18 |
| noonedeadpunk | Just launched dist-upgrade for my aio environment and I still have 2.0.8 | 14:19 |
| odyssey4me | adcpdk: not sure at this point, it'd be good to verify whether someone else is seeing the same | 14:20 |
| fghaas | evrardjp, odyssey4me: sadly, that didn't take long to break. :( http://paste.openstack.org/show/729239/ | 14:20 |
| odyssey4me | going afk for a bit | 14:20 |
| fghaas | (this is current master. I'll try stable/rocky next) | 14:20 |
| adcpdk | I think I am going to remove it and re-run setup-hosts | 14:21 |
| noonedeadpunk | adcpdk have you added some non-standart repos? | 14:21 |
| noonedeadpunk | or backport repos? | 14:22 |
| fghaas | Same thing. | 14:22 |
| adcpdk | http://paste.openstack.org/show/729240/ | 14:22 |
| adcpdk | here are my repos | 14:23 |
| adcpdk | please let me know how it shuld look like | 14:23 |
| noonedeadpunk | Mine is installed from "deb http://nova.clouds.archive.ubuntu.com/ubuntu/ xenial-backports main restricted universe multiverse" | 14:24 |
| noonedeadpunk | http://paste.openstack.org/show/729242/ | 14:24 |
| *** hamzy_ is now known as hamzy | 14:26 | |
| openstackgerrit | Markos Chandras (hwoarang) proposed openstack/openstack-ansible-openstack_openrc master: templates: openrc: Also add OS_INSECURE to openrc file https://review.openstack.org/598978 | 14:27 |
| adcpdk | noonedeadpunk I see, can you specify your /etc/apt/sources.list | 14:28 |
| adcpdk | ? | 14:28 |
| noonedeadpunk | http://paste.openstack.org/show/729243/ | 14:29 |
| hwoarang | another opnfv fix please: https://review.openstack.org/598978 which needs backporting | 14:31 |
| openstackgerrit | Merged openstack/openstack-ansible-lxc_hosts stable/queens: Add always tag to systemd version task https://review.openstack.org/598402 | 14:32 |
| openstackgerrit | Merged openstack/openstack-ansible-lxc_hosts stable/rocky: Add always tag to systemd version task https://review.openstack.org/598401 | 14:32 |
| *** spatel has joined #openstack-ansible | 14:34 | |
| spatel | morning | 14:34 |
| spatel | I have noticed this morning my compute nodes using swap memory a lot specially qemu-kvm | 14:34 |
| spatel | PID=47329 swapped 2799592 KB (qemu-kvm) | 14:35 |
| openstackgerrit | Merged openstack/openstack-ansible-os_barbican stable/rocky: Add missing endpoint_type https://review.openstack.org/598210 | 14:48 |
| openstackgerrit | Merged openstack/openstack-ansible-os_aodh master: Use the TESTING_BRANCH env var for constraints https://review.openstack.org/592502 | 14:57 |
| openstackgerrit | Merged openstack/openstack-ansible-os_aodh master: Remove the unnecessary space https://review.openstack.org/584643 | 14:57 |
| openstackgerrit | Merged openstack/openstack-ansible-os_aodh master: Updated from OpenStack Ansible Tests https://review.openstack.org/593573 | 14:57 |
| *** dxiri has quit IRC | 15:12 | |
| openstackgerrit | Merged openstack/openstack-ansible master: Introduce OpenStack Healthchecks https://review.openstack.org/587544 | 15:19 |
| *** macza has joined #openstack-ansible | 15:26 | |
| *** pcaruana has quit IRC | 15:26 | |
| d34dh0r53 | odyssey4me: have you seen this in your blacklist work http://paste.openstack.org/show/729251/? | 15:32 |
| d34dh0r53 | that's stable/pike btw | 15:34 |
| *** rgogunskiy has quit IRC | 15:35 | |
| openstackgerrit | Merged openstack/openstack-ansible master: Add aio host conf for barbican role https://review.openstack.org/579284 | 15:42 |
| openstackgerrit | Merged openstack/openstack-ansible master: Add missing congress_oslomsg_rpc_password https://review.openstack.org/596551 | 15:42 |
| openstackgerrit | Merged openstack/openstack-ansible master: Bind mount glance cache directory https://review.openstack.org/595230 | 15:42 |
| *** vnogin has quit IRC | 15:42 | |
| *** electrofelix has quit IRC | 15:47 | |
| odyssey4me | d34dh0r53: I've not, it looks like is a mutually exclusive list, so the best approach may be to replace the blacklist with an override of the whitelist? | 15:55 |
| d34dh0r53 | odyssey4me: I'm curious as to why we haven't seen this in other versions, playing around with tempest SHAs in pike ATM. I was just wondering if you'd see that | 15:56 |
| odyssey4me | ie override https://github.com/openstack/openstack-ansible-os_tempest/blob/stable/pike/defaults/main.yml#L65-L72 but remove the test you want to blacklist | 15:56 |
| d34dh0r53 | yeah, that's definitely a possibility | 15:56 |
| odyssey4me | well, we install tempest from pypi and don't use git for it now | 15:56 |
| d34dh0r53 | do you happen to know where the version of stestr is defined? | 15:56 |
| odyssey4me | only master uses git | 15:56 |
| d34dh0r53 | ohh, that make sense | 15:57 |
| d34dh0r53 | ok | 15:57 |
| odyssey4me | so tempest for pike is https://github.com/openstack/requirements/blob/stable/pike/upper-constraints.txt#L358 | 15:57 |
| d34dh0r53 | thanks | 15:58 |
| odyssey4me | queens is https://github.com/openstack/requirements/blob/stable/queens/upper-constraints.txt#L364 - so if there's a new feature or change in deps, it'll be between those two tags | 15:58 |
| odyssey4me | I'd suggest just overriding the default whitelist though, because changing deps introduces more headaches. | 15:58 |
| d34dh0r53 | ok | 16:00 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible stable/rocky: Add missing congress_oslomsg_rpc_password https://review.openstack.org/599019 | 16:01 |
| *** fghaas has quit IRC | 16:04 | |
| odyssey4me | d34dh0r53: basically, if you change a dep then they may be in conflict with what's in upper constraints - so you end up going through dependency hell... that's why we switch to installing from pypi in the first place | 16:04 |
| odyssey4me | we know that everything in u-c can install together, and given that tempest is there it makes sense to use it | 16:04 |
| d34dh0r53 | yeah, that makes sense, I'll just modify the whitelist on pike | 16:04 |
| *** suggestable has quit IRC | 16:05 | |
| *** openstackgerrit has quit IRC | 16:06 | |
| cloudnull | spatel re instances swapping, have you checked the swappyness on the host? maybe something you can tune down ? | 16:13 |
| cloudnull | waynr maybe something within the plugin that is causing it not to load? | 16:13 |
| cloudnull | odyssey4me fghaas regarding nspawn migrations. we just did this in the lab. my process was to stop the lxc container, and then run nspawn create for the same container. nspawn picked up the existing rootfs and just like that i was up and running with different container runtime | 16:16 |
| -spatel- [root@ostack-compute-15 ~]# cat /proc/sys/vm/swappiness | 16:17 | |
| -spatel- 5 | 16:17 | |
| spatel | cloudnull: ^^ | 16:17 |
| odyssey4me | cloudnull: oh, that's actually very nice :) | 16:17 |
| cloudnull | that said for lxc, my backend was machinectl so to do the same with other backends we'd need to do a bit more | 16:17 |
| spatel | cloudnull: i am thinking to disable swap filesystem, why do we need that ? | 16:17 |
| odyssey4me | cloudnull: ok, I've got a topic at the PTG to discuss more details so that we can try and nail it down this coming cycle | 16:17 |
| cloudnull | once my containers were running lxc I deleted the corresponding path in /var/lib/lxc/$name which gets the container out of the lxc backend | 16:18 |
| cloudnull | odyssey4me sweet! | 16:18 |
| *** openstackgerrit has joined #openstack-ansible | 16:19 | |
| openstackgerrit | Bjoern Teipel proposed openstack/openstack-ansible-ops master: Reducing memory footprint on mnaio https://review.openstack.org/599022 | 16:19 |
| cloudnull | spatel swap is super useful from a sysem perspective | 16:19 |
| cloudnull | however if you have the available mem and wont overcommit it, then by all means disable it | 16:20 |
| spatel | i have set 4G swap | 16:20 |
| spatel | should i increase in that case.. if its important | 16:20 |
| cloudnull | or just set the swappiness to <10 | 16:20 |
| spatel | currently swappiness is 5 | 16:20 |
| cloudnull | no you dont need much | 16:20 |
| spatel | will try 10 and see | 16:20 |
| odyssey4me | d34dh0r53 cloudnull antonym could I get some eyes on https://review.openstack.org/#/q/topic:file-backed-vms+status:open please - with that, so far, I can bring up the MNAIO from images and have Galera starting properly in a cluster on boot | 16:21 |
| odyssey4me | I'm working on rabbitmq next. | 16:21 |
| *** dxiri has joined #openstack-ansible | 16:24 | |
| cloudnull | 10 will increase the chance for the host to put things into swap | 16:26 |
| cloudnull | if you have 5 and your seeing issues with it you ma y want to set it to 1 | 16:26 |
| odyssey4me | and cloudnull - before you ask, yes, the lessons learned out of getting galera up and running is something I have on my TODO to put into the galera_server role to make the rebootstrap far simpler | 16:27 |
| cloudnull | ha! | 16:27 |
| cloudnull | was going to ask that and if we need to recreate the grstate.dat? | 16:28 |
| odyssey4me | I tried every method we have in our docs, and they all failed. This method reliably works. | 16:28 |
| openstackgerrit | Markos Chandras (hwoarang) proposed openstack/openstack-ansible-openstack_openrc master: templates: openrc: Also add OS_INSECURE and OS_VERIFY to openrc file https://review.openstack.org/598978 | 16:28 |
| cloudnull | cool! | 16:28 |
| cloudnull | ok i gotta run again, flying in an hour and a half | 16:29 |
| cloudnull | take care all | 16:29 |
| odyssey4me | cloudnull: well, interestingly, as per http://galeracluster.com/documentation-webpages/restartingcluster.html if grastate.dat has 'safe_to_bootstrap: 1' then all is well and things will just work. | 16:29 |
| odyssey4me | but if it doesn't, then gvwstate.dat with http://galeracluster.com/documentation-webpages/pcrecovery.html is the best fallback you have | 16:30 |
| odyssey4me | lemme push up an ops docs patch for this, as it's useful knowledge | 16:30 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible master: docs: Add links to very useful Galera recovery docs https://review.openstack.org/599027 | 16:38 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible-tests master: Update ansible to 2.5.8 https://review.openstack.org/592317 | 16:41 |
| *** aicilamd has quit IRC | 16:41 | |
| *** olivierb_ has quit IRC | 16:41 | |
| *** hamzaachi has joined #openstack-ansible | 16:43 | |
| odyssey4me | um, whoa... https://github.com/openstack/networking-ansible cc jamesdenton | 16:46 |
| odyssey4me | aha: https://github.com/openstack/networking-ansible#use-cases | 16:47 |
| odyssey4me | looks like that's new in rocky - prometheanfire have you seen that? | 16:47 |
| *** aicilamd has joined #openstack-ansible | 16:49 | |
| odyssey4me | logan-: that may be of interest to you too | 16:49 |
| *** aicilamd has quit IRC | 16:52 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/ansible-role-python_venv_build master: Simplify builds/installs by eliminating venv packaging https://review.openstack.org/598957 | 17:05 |
| odyssey4me | hwoarang: looks like we're good to go for https://review.openstack.org/#/q/topic:osa-update-ansible+status:open | 17:06 |
| *** hamzaachi has quit IRC | 17:08 | |
| noonedeadpunk | methods for recovering galera in docs are not really the best ones, and for more cases they'll fail. But still it's pretty usefull for diagnostics of the cluster state... | 17:18 |
| noonedeadpunk | I didn't work with nspawn a lot, so is it possible to pass a bare network interface inside of the namespace? | 17:19 |
| noonedeadpunk | like with lxc and phys network type? | 17:20 |
| odyssey4me | noonedeadpunk: I don't know. You'll have to try it out and see. I think it may be possible, given what I've read in the docs. | 17:30 |
| odyssey4me | noonedeadpunk: And yeah, recovering a failed galera cluster is a PITA. There are just so many failure conditions to take into account. | 17:31 |
| odyssey4me | I find myself wondering whether https://www.cockroachlabs.com/ would work as a drop-in replacement... if I had the time. ;) | 17:32 |
| noonedeadpunk | galera is pretty nice and wide-spread, and lots of people already know how to manage it... | 17:33 |
| odyssey4me | Yep, that's true. | 17:33 |
| noonedeadpunk | While cockroach is too young as for me | 17:34 |
| noonedeadpunk | And this needs migration from MySQL to Postgres and I'm not really sure, that every openstack project is ready for postgres | 17:37 |
| noonedeadpunk | Actually, I've never tried it | 17:37 |
| *** DanyC has quit IRC | 17:38 | |
| noonedeadpunk | ok, have a good weekend everyone! | 17:38 |
| *** mmercer has joined #openstack-ansible | 17:40 | |
| *** noonedeadpunk has quit IRC | 17:41 | |
| *** aicilamd has joined #openstack-ansible | 17:46 | |
| *** fghaas has joined #openstack-ansible | 18:04 | |
| *** luksky has joined #openstack-ansible | 18:08 | |
| *** hamzaachi has joined #openstack-ansible | 18:09 | |
| xdfil | mhayden I've been working with SR-IOV virtual interfaces, and I'm trying to have them use a consistent device name. I saw your blog post https://major.io/2015/08/21/understanding-systemds-predictable-network-device-names/ | 18:10 |
| mhayden | yessir | 18:10 |
| xdfil | I haven't found a way to select which "policy" udev should use | 18:11 |
| xdfil | I want to skip the kernel names and such and go strait to the enp* names | 18:11 |
| xdfil | I thought /etc/systemd/network/default.link would allow me to control that but it doesn;t seem to have any effect | 18:12 |
| xdfil | on Ubuntu 16.04 I still get a mix of eno1 eno2 type names and enp* type names | 18:13 |
| mhayden | oh i know what you mean there | 18:14 |
| mhayden | my supermicro xeon D box has that fun problem | 18:14 |
| mhayden | https://major.io/2015/09/28/first-thoughts-linux-on-the-supermicro-5028d-t4nt/ | 18:14 |
| xdfil | So let me ask this: is udev /supposed/ to allow me to select a naming policy? Or is it just telling me what the order of the policies is and there is nothing I can really do about it, and that I have to solve it in the hardware. | 18:16 |
| mhayden | xdfil: look at https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/ and skip to the "I don't like this" section at the end | 18:16 |
| mhayden | xdfil: the goal of the post is to explain how systemd decides on those names | 18:17 |
| mhayden | you can override those names | 18:17 |
| mhayden | hell, you can make it sriov0 and sriov1 if you're feeling really excited | 18:17 |
| xdfil | mhayden: but "i don't like this" will just give me ethX names right? | 18:17 |
| mhayden | that's one option, or you can provide custom names | 18:18 |
| *** priteau has quit IRC | 18:18 | |
| xdfil | custom names seem to require rules based on MAC | 18:19 |
| xdfil | hmm ok, well at least you verified my assumptions about what was actually possible | 18:19 |
| xdfil | I noticed that the PCIe location numbers of the SR-IOV devices don't actually represent the parent physical device | 18:22 |
| xdfil | the only mapping I've found is inside the /sys filesystem of the parent device there is a listing of the virtual devices | 18:22 |
| xdfil | perhaps, none of this really matters | 18:23 |
| xdfil | and I'm just getting hung up on it | 18:23 |
| xdfil | mhayden: another inconsistency I see is not all of the virtual nics have the same fields available when viewing udevadm info -e | 18:27 |
| mhayden | oof | 18:27 |
| xdfil | I'm going to go read your supermicro article now | 18:27 |
| mhayden | is there something in openstack that allows you to set something predictable from below? | 18:27 |
| xdfil | sounds like hardware is the key | 18:27 |
| xdfil | well, as far as openstack goes | 18:27 |
| *** aicilamd has quit IRC | 18:27 | |
| xdfil | it seems like it only needs the PCIe mapping of the physical/parent interface and it will know which interfaces are its children | 18:28 |
| xdfil | so that's comforting. I should probably just ignore this issue in hopes that it doesn't really matter | 18:29 |
| *** DanyC has joined #openstack-ansible | 18:32 | |
| *** mmercer has quit IRC | 18:32 | |
| xdfil | mhayden: yeah your supermicro article seems very similar to my issue. I'm suprised, going into this I thought forcing /everything/ to use ID_NET_NAME_PATH would be a simple thing. | 18:37 |
| *** fghaas has quit IRC | 18:37 | |
| xdfil | I'm still unsure if I should be tinkering with systemd/network or with udev rules. | 18:39 |
| xdfil | udev rules look so ugly, I'd like to avoid touching them | 18:40 |
| odyssey4me | They're not ugly. They're just a kinda pretty you're not used to. | 18:42 |
| xdfil | hahaha | 18:42 |
| xdfil | I was reading this https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/networking_guide/sec-understanding_the_device_renaming_procedure | 18:42 |
| xdfil | even though in my case i'm on 16.04 | 18:43 |
| xdfil | haven't been able to determine how to override them exactly | 18:44 |
| odyssey4me | Oh yeah. I spent some time reading through udev rules not so long ago - it tooks me an hour of confusion to decide that this was not the road I should travel. | 18:44 |
| odyssey4me | While all those tools are powerful, sometimes you just gotta step back and think about taking another approach which is easier to implement and maintain. | 18:45 |
| xdfil | amen | 18:46 |
| xdfil | I put "NamePolicy=path" in /etc/systemd/network/99-default.link | 18:46 |
| xdfil | it does nothing | 18:47 |
| xdfil | who has the final say in net device names? systemd or udev ? | 18:47 |
| prometheanfire | odyssey4me: wha? something about networking-ansible? | 18:47 |
| prometheanfire | cjloader: https://review.openstack.org/#/c/598979/ | 18:49 |
| prometheanfire | might be the problem | 18:49 |
| odyssey4me | prometheanfire: yeah, I just noticed it and thought it looked like something interesting - then saw it was ironic focused, so thought you and cjloader might find it interesting given you've done work in that space | 18:51 |
| prometheanfire | ya, I'll take a look | 18:51 |
| prometheanfire | not sure a new backend would be good, but may be nice for test infra | 18:52 |
| *** tosky has quit IRC | 18:52 | |
| *** openstackstatus has quit IRC | 18:58 | |
| *** DanyC has quit IRC | 19:01 | |
| *** schwicht has quit IRC | 19:08 | |
| *** adcpdk has quit IRC | 19:10 | |
| odyssey4me | prometheanfire: well, I think the idea may be that it can be used to have switches reconfigured after the initial provisioning of the nodes | 19:13 |
| odyssey4me | so, for example, you ask for a node - then instead of going onto a flat network, the final provisioning of the node can change the switch config to put your node onto its own isolated vlan/whatever when it reboots | 19:14 |
| odyssey4me | as far as I know, that's the (very) short version of what is done in RAX OnMetal | 19:14 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/openstack-ansible master: docs: Add links to very useful Galera recovery docs https://review.openstack.org/599027 | 19:20 |
| *** aicilamd has joined #openstack-ansible | 19:24 | |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/ansible-role-python_venv_build master: Simplify builds/installs by eliminating venv packaging https://review.openstack.org/598957 | 19:26 |
| prometheanfire | odyssey4me: ah, that's cool | 19:26 |
| odyssey4me | prometheanfire: yeah, I think it is - basically it's more of a generic tool that enables you to use ansible playbooks to do whatever needs to be done, so if there's no neutron driver for it, but ansible can do it, then you're in the game | 19:27 |
| odyssey4me | beyond that, even if there is a neutron driver for it, but you want to do some more complex orchestration that the driver doesn't handle (or the driver requires an expensive license to do that you need), then you can still do it | 19:28 |
| *** fghaas has joined #openstack-ansible | 19:32 | |
| xdfil | mhayden odyssey4me logan- : Problem Solved! All I had to do was install the biosdevname package. My interfaces changed from "enoX" to "emX", BUT now my SR-IOV VFs are named after the PF they live on, like "em1_1", "em1_2". | 19:34 |
| odyssey4me | xdfil: aha! I'm sure mhayden might be interested, even if he does use CentOS. | 19:35 |
| cloudnull | o/ all | 19:36 |
| xdfil | I think it's universal/systemd | 19:36 |
| cloudnull | hows it this afternoon ? | 19:36 |
| xdfil | o/ hey mr. null | 19:36 |
| *** openstackstatus has joined #openstack-ansible | 19:36 | |
| *** ChanServ sets mode: +v openstackstatus | 19:36 | |
| xdfil | cloudnull: I actually had a bunch of stuff I wanted to run by you in regards to nspawn networking. But it all really boils down to one thing | 19:37 |
| xdfil | Please, make your ansible support macvtap, as well as passing a "physical" nic | 19:39 |
| odyssey4me | xdfil: well, funny you should mention that - I think that's already there, if you want it | 19:39 |
| odyssey4me | https://review.openstack.org/593230 | 19:39 |
| cloudnull | xdfil yup nspawn uses macvaln | 19:40 |
| cloudnull | **macvlan | 19:40 |
| cloudnull | no more bridges, no more veths | 19:40 |
| xdfil | as long as it can do that, it should be compatible with just about any network environment. (like OVS DPDK SR-IOV etc) | 19:40 |
| cloudnull | ++ thats the hope | 19:40 |
| *** schwicht has joined #openstack-ansible | 19:40 | |
| xdfil | right, you mentioned that before, and I have yet to confirm that macvlan would play nice on a OVS+DPDK bridge | 19:41 |
| xdfil | But device passthrough will work for sure | 19:42 |
| odyssey4me | xdfil: well, I think the container networking is seperate from the tenant networking in openstac | 19:42 |
| odyssey4me | nspawn has to do with the OSA back-end, nothing to do with the openstack projects/cloud | 19:42 |
| xdfil | and by play nice, I mean that it will work but also have the performance of DPDK | 19:42 |
| odyssey4me | the openstack cloud really depends on how you configure it, so if openstack can do it then you can already do it today - especially with the config override mechanism | 19:43 |
| xdfil | odyssey4me: right I understand the its 2 seperate topics. underlay/overlay | 19:43 |
| cloudnull | in terms of device performance it shouldn't have any impact given that the agents are on the host and would be free to bind the physical interface. | 19:43 |
| xdfil | well thats not accurate either | 19:43 |
| cloudnull | there's no assumption of bridges any longer. | 19:44 |
| xdfil | cloudnull: but if the "physical" interface is actually a port on an OVS bridge is my concern | 19:44 |
| xdfil | I want the freedom to do that, if I choose to | 19:44 |
| xdfil | so that goes back to just making sure OSA+nspawn will work with macvlan AND device pasthrough | 19:45 |
| cloudnull | thatd be fine too you could bid the containers to the ovs bridge in that case. | 19:45 |
| xdfil | right I think so | 19:46 |
| xdfil | it should work the same | 19:46 |
| xdfil | I was really excited about DPDK, but man, configuring the host to use it is pretty involved. HugePages, CPU pinning, NUMA planning... | 19:48 |
| cloudnull | xdfil http://paste.openstack.org/show/727250/ - thats an example of the nspawn networking on my lab deployment | 19:50 |
| xdfil | The nice side effect of DPDK is that if you get it working, you also end up with very-optimized NUMA compute-node | 19:51 |
| cloudnull | my kick created the bond, the mv-* interfaces were auto generated | 19:51 |
| cloudnull | which is all based on the openstack_user_config | 19:51 |
| *** ansmith has quit IRC | 19:52 | |
| cloudnull | before we assumed br-* was the deployers problem, with nspawn we just say there needs to be an interface to bind to. the role does not care what the interface is | 19:52 |
| cloudnull | that could very well be eth0, bond0, br-mgmt, sriov0, etc | 19:53 |
| spatel | cloudnull: what is the best method to add bulk compute nodes? | 19:53 |
| odyssey4me | hmm, so on ubuntu and centos today I've tried using two different roles and ended up with the same results for a functional test - it fails when trying to do package things in the containers just after they're done being created | 19:53 |
| odyssey4me | in both cases, the host is able to reach the package repo, but not the containers | 19:53 |
| spatel | I am adding them one by one and its hell of a time to run playbook | 19:54 |
| odyssey4me | I'm trying another region now to see, but perhaps something merged that broke things... although I don't see it. | 19:54 |
| spatel | i have 50 compute node waiting and each compute node talking 15 min or more :( | 19:54 |
| odyssey4me | spatel: why do them one at a time - why not all at once? | 19:54 |
| *** aicilamd has quit IRC | 19:54 | |
| cloudnull | spatel just add the compute nodes to your openstack_user_config.yml (or a file in conf.d), run setup-hosts.yml then setup-openstack.yml with a limit of nova_compute | 19:54 |
| *** openstackstatus has quit IRC | 19:55 | |
| cloudnull | you could also just run setup-everything.yml with a limit of "nova_compute" | 19:55 |
| cloudnull | `openstack-ansible setup-everything.yml --limit nova_compute` | 19:55 |
| spatel | oh!! i am doing --limit one each compute node... | 19:55 |
| odyssey4me | spatel: in general, running the three playbooks is always ok - or setup-everything, although that'll chew up ram | 19:56 |
| cloudnull | ^ yea, no need to do them in serial, just spray them out there. | 19:56 |
| odyssey4me | the ony reason the docs do limits is to save time | 19:56 |
| spatel | playbook always run one by one right it won't to parallel | 19:56 |
| odyssey4me | and by save time, I mean save time when you happen to be doing one at a time | 19:56 |
| spatel | s/to/do | 19:56 |
| cloudnull | its all parallel | 19:57 |
| spatel | hmmm! | 19:57 |
| cloudnull | it does some things in serial | 19:57 |
| odyssey4me | spatel: all computes are done at once - I think it might be one first, just to validate that all is good, then all the rest | 19:57 |
| *** openstackstatus has joined #openstack-ansible | 19:57 | |
| *** ChanServ sets mode: +v openstackstatus | 19:57 | |
| xdfil | cloudnull yeah that should be compatible with just about everything. as in any interface, bond, bridge-interface... yeah | 19:57 |
| odyssey4me | yep, compute is all at once: https://github.com/openstack/openstack-ansible/blob/stable/queens/playbooks/os-nova-install.yml#L117 | 19:57 |
| spatel | odyssey4me: & cloudnull thanks... i was thinking its going to be serial | 19:57 |
| odyssey4me | you can adjust that if you prefer | 19:58 |
| cloudnull | xdfil that's the hope, in our tests its been great however it'd be great to get more eyes on all that | 19:58 |
| cloudnull | and work out the rough spots | 19:58 |
| odyssey4me | cloudnull: how do you like them apples? https://review.openstack.org/#/q/topic:osa-update-ansible+(status:open+OR+status:merged) | 19:58 |
| cloudnull | tasty! | 19:59 |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible master: Removed conditional in test setup https://review.openstack.org/575748 | 20:01 |
| openstackgerrit | Kevin Carter (cloudnull) proposed openstack/openstack-ansible master: Removed conditional in test setup https://review.openstack.org/575748 | 20:02 |
| odyssey4me | ok, good news - in a way, it seems that the failure is local to a single region in the cloud I was doing tests in | 20:04 |
| openstackgerrit | David Wilde proposed openstack/openstack-ansible-os_tempest stable/pike: Remove blacklisted entries from whitelist file https://review.openstack.org/599060 | 20:05 |
| cloudnull | crisis averted ! | 20:06 |
| *** schwicht has quit IRC | 20:06 | |
| *** schwicht has joined #openstack-ansible | 20:08 | |
| spatel | cloudnull: i got this error on one of compute node which i am trying to add | 20:10 |
| spatel | http://paste.openstack.org/show/729267/ | 20:10 |
| spatel | any clue | 20:10 |
| *** openstackstatus has quit IRC | 20:11 | |
| cloudnull | looks like /usr/lib/python2.7/dist-packages/libguestfsmod.so is not found ? | 20:11 |
| *** openstackstatus has joined #openstack-ansible | 20:11 | |
| *** ChanServ sets mode: +v openstackstatus | 20:11 | |
| cloudnull | maybe a package not installed? | 20:11 |
| cloudnull | my guess would be libguestfs | 20:11 |
| * cloudnull is not sure what that is called on cent | 20:12 | |
| *** hamzaachi has quit IRC | 20:12 | |
| spatel | i have added 20 node today and one failed with that error | 20:12 |
| cloudnull | see if that guestfs package is not found | 20:12 |
| cloudnull | compare two nodes and see what the difference is | 20:13 |
| odyssey4me | interesting: https://github.com/openstack/openstack-ansible-os_nova/blob/master/vars/redhat-7.yml#L85 | 20:13 |
| odyssey4me | maybe centos has a dep there, but it's not expressed | 20:13 |
| cloudnull | I gotta run, time to fly again . | 20:13 |
| cloudnull | take care all, see you all middle of next week | 20:13 |
| *** schwicht has quit IRC | 20:13 | |
| odyssey4me | ciao cloudnull - happy days off! | 20:14 |
| odyssey4me | spatel: what series is that? | 20:14 |
| spatel | 7.5 | 20:14 |
| spatel | Centos | 20:14 |
| odyssey4me | spatel: I mean which OSA series (pike/queens/etc) | 20:15 |
| spatel | queens | 20:15 |
| spatel | i have kickstarted all 20 node with same kickstart image and this one only having issue.. very strange | 20:15 |
| odyssey4me | so https://github.com/openstack/openstack-ansible-os_nova/blob/stable/queens/tasks/drivers/kvm/nova_compute_kvm_install.yml#L65-L77 just seems to magically expect those files there | 20:15 |
| *** nicolasbock has quit IRC | 20:16 | |
| spatel | comparing with other box | 20:17 |
| odyssey4me | are those files there? if not - what provides those files (from a package point of view) | 20:17 |
| odyssey4me | if not those files - are the same file names present, but in a different folder? | 20:18 |
| *** hamzaachi has joined #openstack-ansible | 20:19 | |
| spatel | /usr/lib/python2.7/dist-packages/libguestfsmod.so isn't available on any compute node | 20:20 |
| spatel | i have on each node and file not found /usr/lib/python2.7/dist-packages/libguestfsmod.so | 20:20 |
| -spatel- [root@ostack-compute-24 ~]# locate libguestfsmod.so | 20:20 | |
| -spatel- /openstack/venvs/nova-17.0.8/lib/python2.7/libguestfsmod.so | 20:20 | |
| -spatel- /usr/lib64/python2.7/site-packages/libguestfsmod.so | 20:20 | |
| spatel | its on this path /usr/lib64/python2.7/site-packages/libguestfsmod.so | 20:20 |
| -spatel- [root@ostack-compute-24 ~]# rpm -qf /usr/lib64/python2.7/site-packages/libguestfsmod.so | 20:21 | |
| -spatel- python-libguestfs-1.36.10-6.el7_5.2.x86_64 | 20:21 | |
| spatel | odyssey4me: i found on failed compute nodes doesn't have /openstack/venv dir | 20:22 |
| spatel | look like that is the issue | 20:23 |
| *** DanyC has joined #openstack-ansible | 20:25 | |
| *** dxiri has quit IRC | 20:26 | |
| spatel | i have removed /var/cache/nova and running playbook | 20:28 |
| *** niceplace has joined #openstack-ansible | 20:32 | |
| *** openstackstatus has quit IRC | 20:36 | |
| *** openstackstatus has joined #openstack-ansible | 20:36 | |
| *** ChanServ sets mode: +v openstackstatus | 20:36 | |
| *** schwicht has joined #openstack-ansible | 20:39 | |
| spatel | odyssey4me: still no luck.. i have remove fill /var/cache/nova* | 20:42 |
| spatel | which playbook deploy nova in /openstack/venv/ ? | 20:42 |
| spatel | look like some kind of bug | 20:43 |
| odyssey4me | spatel: it's the nova one | 20:44 |
| odyssey4me | without more details, it's hard to decide - that said, I'm not in a great place to do so right now | 20:44 |
| odyssey4me | if it's only hit one node - then something special is happening there - comparing nodes is a good place to start | 20:45 |
| spatel | sure no problem i am going to look why its not installing nova in /openstack | 20:45 |
| odyssey4me | knowing those differences would be good | 20:45 |
| odyssey4me | well, perhaps look more carefully at the play results to see if there's a hidden failure earlier | 20:45 |
| odyssey4me | ok folks, I'm out for the night - have a great w/end! | 20:51 |
| openstackgerrit | Jesse Pretorius (odyssey4me) proposed openstack/ansible-role-python_venv_build master: Simplify builds/installs by eliminating venv packaging https://review.openstack.org/598957 | 20:52 |
| evrardjp | odyssey4me: have a good night | 20:53 |
| evrardjp | mnaser: the etherpad of the upgrade SIG is well empty, it might be good to find someone to join the sessions at the PTG: https://etherpad.openstack.org/p/upgrade-sig-ptg-stein | 20:53 |
| *** rgogunskiy has joined #openstack-ansible | 20:55 | |
| *** luksky has quit IRC | 21:00 | |
| *** rgogunskiy has quit IRC | 21:03 | |
| *** rgogunskiy has joined #openstack-ansible | 21:04 | |
| *** Leo_m has joined #openstack-ansible | 21:09 | |
| *** aicilamd has joined #openstack-ansible | 21:09 | |
| *** DanyC has quit IRC | 21:13 | |
| *** fghaas has quit IRC | 21:22 | |
| jrosser | i have some backports been sitting round for a week if someone fancies some easy reviews https://review.openstack.org/#/c/594839/ https://review.openstack.org/#/c/594912/ https://review.openstack.org/#/c/594913/ which will then allow this to merge for rocky https://review.openstack.org/#/c/594927/ | 21:28 |
| *** rgogunskiy has quit IRC | 21:29 | |
| *** aicilamd has quit IRC | 21:33 | |
| *** elbragstad has quit IRC | 22:02 | |
| *** EvilienM is now known as EmilienM | 22:04 | |
| openstackgerrit | Merged openstack/openstack-ansible master: Add networkd prefix when creating networks https://review.openstack.org/571898 | 22:06 |
| openstackgerrit | Merged openstack/openstack-ansible master: docs: Add links to very useful Galera recovery docs https://review.openstack.org/599027 | 22:06 |
| *** hamzaachi has quit IRC | 22:07 | |
| *** hamzaachi has joined #openstack-ansible | 22:08 | |
| *** tosky has joined #openstack-ansible | 22:09 | |
| evrardjp | could we get this in? https://review.openstack.org/#/c/593051/ else it will be harder for me to release rc3 | 22:27 |
| evrardjp | and final ofc | 22:27 |
| mnaser | One second | 22:28 |
| evrardjp | likewise https://review.openstack.org/#/c/596761/ would help | 22:28 |
| evrardjp | after that I will tackle the other branches. | 22:28 |
| *** markvoelker has joined #openstack-ansible | 22:40 | |
| *** hamzaachi has quit IRC | 22:52 | |
| *** spatel has quit IRC | 22:55 | |
| *** macza has quit IRC | 23:07 | |
| *** tosky has quit IRC | 23:14 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!