*** dviroel|rover is now known as dviroel|out | 01:09 | |
*** ysandeep|out is now known as ysandeep | 05:17 | |
*** ysandeep is now known as ysandeep|brb | 07:06 | |
*** ysandeep|brb is now known as ysandeep | 07:59 | |
anskiy | Hello! I'm getting this: https://paste.opendev.org/show/b8GsGA2vZv5A9bTCvwov/ on Yoga (same during upgrade from Xena, except it doesn't fail only on control-plane node (which is repo_server)). This is `no_containers: true` deployment. I'm unsure of what would be the proper way to fix this, that is: to which role should I add that missing variable. | 08:04 |
---|---|---|
jrosser_ | anskiy: which OS is it? | 08:07 |
jrosser_ | and you are trying to use journald-remote or not? | 08:07 |
anskiy | ubuntu 20.04. Yeah, I'm using it | 08:07 |
jrosser_ | here is where that var is mentioned https://codesearch.opendev.org/?q=systemd_utils_prefix | 08:11 |
jrosser_ | i suspect that system_utils_prefix should really be defined somewhere in the vars/ of openstack-ansible-plugins/roles/journald_remote | 08:17 |
anskiy | I think that intention for not defining it there was that it's the prefix for the source side, and that it might be better to get it from somewhere else, preferably, where it is already defined. That is, lxc_host looks like the proper place for it, reducing the amount of duplicate entries, but there is no alternative for this role for `no_containers: true` deployment. | 08:25 |
anskiy | But this thing would anyway break for the case, when running the `infra-journal-remote.yml` role alone. | 08:25 |
jrosser_ | vars defined in the lxc_hosts role are only in scope during the running of that role | 08:26 |
jrosser_ | it also looks a little redundant as its /lib/systemd everywhere? | 08:27 |
jrosser_ | i am not sure there is any issue with source or target side either | 08:29 |
anskiy | but when you run setup-infrastructure, that variable get defined for the nodes, for which repo-install (part of setup-infrastructure) has to run. | 08:30 |
jrosser_ | i'm not really following | 08:31 |
jrosser_ | variables are usually evaluated on a per host basis in ansible | 08:31 |
anskiy | give me a couple minutes, gonna show you the failed run for setup-infrastructure | 08:31 |
anskiy | you think it looks redundant because of it is "/lib/systemd" for redhat and debian anyways, or just that it is defined in multiple roles? | 08:34 |
anskiy | technically there could be different prefixes for source and target hosts, but that could be just some exotic thing... | 08:35 |
jrosser_ | there isnt really a split in the variable between source and target | 08:41 |
anskiy | jrosser_: https://paste.opendev.org/show/bvA0csWsi8NHRwPiAzPK/ here is the failed run for setup-infrastructure playbook. As you can see, `systemd_service : Create TEMP run dir` task doesn't fail on oscp99 node, on which repo-install playbook runs. | 08:41 |
jrosser_ | all the tasks in journald_remote run against all `hosts` https://github.com/openstack/openstack-ansible/blob/master/playbooks/infra-journal-remote.yml#L17 | 08:41 |
anskiy | https://opendev.org/openstack/openstack-ansible-plugins/src/branch/master/roles/journald_remote/defaults/main.yml#L23 there is this thing, but yes, contents of both variables are the same | 08:42 |
jrosser_ | then this line is wrong https://opendev.org/openstack/openstack-ansible-plugins/src/branch/master/roles/journald_remote/vars/main.yml#L23 | 08:44 |
jrosser_ | it should also be journald_remote_systemd_prefix | 08:45 |
jrosser_ | as i say the same role runs against all the hosts regardless of if they are source or target | 08:45 |
jrosser_ | the values defined in other roles are not relevant (should be removed also) as they are not in scope for the journald_remote role | 08:46 |
anskiy | they could be in scope, if you run more higher level role, such as setup-infrastructure, but I've got your point: roles need to be self-contained or, cluster-wide variables should be defined somewhere down the /opt/openstack-ansible. | 08:50 |
anskiy | gonna file a change with substituting https://opendev.org/openstack/openstack-ansible-plugins/src/branch/master/roles/journald_remote/vars/main.yml#L23 this for `journald_remote_systemd_prefix`. Thank you! | 08:51 |
jrosser_ | no, thats not a role | 08:51 |
anskiy | oh, yeah, sorry, playbook | 08:51 |
jrosser_ | is oscp99 your journald target? | 08:52 |
anskiy | no, it's the osa | 08:52 |
anskiy | https://paste.opendev.org/show/bDBheOt76lD3TNZzwpjQ/ relevant part of openstack_user_config, if you want to continue this discussion :) | 08:55 |
jrosser_ | i'm just trying to wrap my head around the oscp99 that did not fail :) | 08:55 |
anskiy | `system_utils_prefix` gets defined for it here: https://opendev.org/openstack/openstack-ansible-repo_server/src/branch/master/vars/debian.yml#L16, only for the case, when you run setup-infrastructure playbook, which contains BOTH `repo-install.yml` and `infra-journal-remote.yml` playbooks. So, it looks, that for this specific case, scope for the higher level playbook contains state from all lower level playbooks. | 09:00 |
jrosser_ | vars defined in a role are only in scope when that role is being run, as far is i understand | 09:06 |
jrosser_ | anskiy: what happens if you just run `openstack-ansible playbooks/infra-journal-remote.yml` | 09:08 |
opendevreview | Jean-Philippe Evrard proposed openstack/openstack-ansible-os_neutron master: Allow overriding computed maximums https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/854591 | 09:10 |
anskiy | jrosser_: I've checked that with minimal example: it works as you say | 09:15 |
anskiy | jrosser_: that's the first paste, I've sent :https://paste.opendev.org/show/b8GsGA2vZv5A9bTCvwov/ | 09:15 |
jrosser_ | ah ok, and in that case it failed to template all 3 hosts | 09:16 |
jrosser_ | anskiy: i think that what is happening here is probably a bit accidental, that in the lxc_hosts role vars/debian-host.yml is included and that defines systemd_utils_prefix | 09:25 |
jrosser_ | i am a little surprised that the definition of that variable persists across plays (setup-infrastructure.yml is many plays each calling many roles) | 09:26 |
jrosser_ | and as playbooks/infra-journal-remote.yml should be able to be run independantly what we have at the moment is really not correct | 09:26 |
anskiy | jrosser_: I don't think I'm too curios to try to reproduce it further from the minimal example :) Thank you! I'm gonna submit a fix for this today. | 09:34 |
*** ysandeep is now known as ysandeep|lunch | 09:47 | |
jrosser_ | evrardjp: looks like there are still centos-8-stream jobs running for https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/854591/ | 09:50 |
jrosser_ | those need getting rid of | 09:51 |
*** ysandeep|lunch is now known as ysandeep | 10:27 | |
*** dviroel|out is now known as dviroel|rover | 11:23 | |
opendevreview | Danila Balagansky proposed openstack/openstack-ansible-plugins master: Use `journald_remote_systemd_prefix` for systemd prefix https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/854611 | 11:31 |
opendevreview | Danila Balagansky proposed openstack/openstack-ansible-plugins master: Bind logging target via systemd-journal-remote.socket https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/854613 | 11:56 |
*** tosky_ is now known as tosky | 12:25 | |
*** ysandeep is now known as ysandeep|brb | 12:47 | |
*** ysandeep|brb is now known as ysandeep | 14:28 | |
*** dviroel is now known as dviroel|rover | 15:02 | |
*** tosky is now known as Guest1129 | 15:12 | |
*** tosky_ is now known as tosky | 15:12 | |
*** dviroel|rover is now known as dviroel|rover|lunch | 15:45 | |
*** ysandeep is now known as ysandeep|out | 15:58 | |
*** dviroel|rover|lunch is now known as dviroel|rover | 17:10 | |
*** ianw_ is now known as ianw | 20:37 | |
*** tosky_ is now known as tosky | 22:06 | |
*** thelounge1753 is now known as thelounge175 | 22:28 | |
*** anskiy1 is now known as anskiy | 22:28 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!