| opendevreview | Merged openstack/openstack-ansible master: Bump uWSGI version https://review.opendev.org/c/openstack/openstack-ansible/+/864579 | 00:09 |
|---|---|---|
| *** ysandeep|out is now known as ysandeep | 07:25 | |
| *** akahat|ruck is now known as akahat|ruck|lunch | 08:14 | |
| *** akahat|ruck|lunch is now known as akahat|ruck | 09:22 | |
| *** ysandeep is now known as ysandeep|afk | 10:08 | |
| dok53 | Morning all, think I'm just there with the volume storag on EMC SAN. I'm getting the error shown here: https://paste.openstack.org/show/bYKiF0EmEJZCE7bgweGJ/ I did "pip install storops" and it's installed but cinder doesn't see it. Do any of you know if this is expected or how would I pip install in the cinder venv so that cinder can see it | 10:50 |
| *** dviroel_ is now known as dviroel | 11:09 | |
| opendevreview | Marcus Klein proposed openstack/openstack-ansible-os_neutron master: Allow to set dnsmasq configuration options https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/864872 | 11:17 |
| jrosser | do | 11:28 |
| jrosser | k | 11:28 |
| jrosser | noonedeadpunk: | 11:28 |
| jrosser | noonedeadpunk: | 11:28 |
| jrosser | argh | 11:28 |
| jrosser | dok53: what command did you use to pip install? | 11:29 |
| *** ysandeep|afk is now known as ysandeep | 11:38 | |
| dok53 | noonedead punk just "pip install storops" | 11:39 |
| noonedeadpunk | o/ | 11:40 |
| noonedeadpunk | dok53: cinder runs inside virtualenv | 11:40 |
| noonedeadpunk | so you need to use /openstack/venvs/cinder-${osa_version}/bin/pip | 11:41 |
| noonedeadpunk | Or, define all that in your user_variables plus add storops - https://opendev.org/openstack/openstack-ansible-os_cinder/src/branch/master/defaults/main.yml#L304-L314 | 11:42 |
| noonedeadpunk | after that you can re-run openstack-ansible os-cinder-install.yml -e venv_rebuild=true | 11:42 |
| dok53 | Ah that's what I thought, I was following old instructions for our SAN. Thanks yet again :) | 11:43 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Add zookeeper deployment https://review.opendev.org/c/openstack/openstack-ansible/+/864750 | 11:44 |
| jrosser | i think we have a hook for user defined pip packages https://opendev.org/openstack/openstack-ansible-os_cinder/src/branch/master/defaults/main.yml#L317 | 11:44 |
| jrosser | dok53: ^ that would be what you want | 11:44 |
| jrosser | noonedeadpunk: did you see discussion of if uwsgi role should / should not use the repo server? | 11:45 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Add zookeeper deployment https://review.opendev.org/c/openstack/openstack-ansible/+/864750 | 11:46 |
| noonedeadpunk | oh, didn't know we have that, sweet | 11:46 |
| noonedeadpunk | jrosser: no I haven't | 11:46 |
| noonedeadpunk | seems it was quite late yestarday? | 11:47 |
| jrosser | ah well damiandabrowski showed that we build uwsgi in each place we use it | 11:47 |
| jrosser | yes it was late :/ | 11:47 |
| noonedeadpunk | um, I can totally recall intention to build uwsgi wheels once only and then install everywhere | 11:49 |
| noonedeadpunk | that was one of main points of the role (except we need to install uwsgi only once for metal) | 11:50 |
| noonedeadpunk | And I'm quite sure wheels for it were built back in the days on repo container | 11:50 |
| noonedeadpunk | though on AIO we can result in not building wheels at all | 11:51 |
| noonedeadpunk | as it's single host only kind of | 11:51 |
| jrosser | LXC aio should? | 11:51 |
| jrosser | anyway seems there is no place in the uwsgi role to specify a build host | 11:52 |
| noonedeadpunk | no, not really. | 11:52 |
| noonedeadpunk | well, we do build them in CI though, I think | 11:52 |
| jrosser | regardless it would be good not to have gcc being installed everywhere | 11:52 |
| noonedeadpunk | do we ever specify build host? | 11:54 |
| noonedeadpunk | except relying on python_venv_build default? | 11:54 |
| noonedeadpunk | we have quite complex logic for picking this up tbh there | 11:55 |
| noonedeadpunk | ok, now looking at code I think we actually don't build wheels in CI | 11:55 |
| noonedeadpunk | There was only a patch to enable that https://review.opendev.org/c/openstack/openstack-ansible/+/752311 | 11:56 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Enable venv_wheel_build_enable for CI https://review.opendev.org/c/openstack/openstack-ansible/+/752311 | 11:56 |
| noonedeadpunk | but yes, you're right jrosser, we need to install both on repo and on host I guess. I _think_ we have some "smart" variable in the role that installs packages only when we do not build wheels | 12:01 |
| noonedeadpunk | So basically, if wheels are not build - we install ssl-dev on destination container, if wheels are set - we install on repo | 12:02 |
| noonedeadpunk | (https://opendev.org/openstack/ansible-role-python_venv_build/src/branch/master/tasks/python_venv_install.yml#L31) | 12:02 |
| noonedeadpunk | not sure if we need it everywhere regardless tbh | 12:02 |
| noonedeadpunk | jrosser: btw, have you succeed in having mgmt net != ssh net ? | 12:05 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/ansible-role-zookeeper master: dnm https://review.opendev.org/c/openstack/ansible-role-zookeeper/+/864887 | 12:18 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/ansible-role-zookeeper master: Initial commit to zookeeper role https://review.opendev.org/c/openstack/ansible-role-zookeeper/+/864752 | 13:32 |
| *** ysandeep is now known as ysandeep|afk | 13:41 | |
| jrosser | noonedeadpunk: yes we totally have mgmt net != ssh net on all my deployments | 13:58 |
| noonedeadpunk | oh, ok. But you had to override all this crap with ansible_host? | 14:01 |
| jrosser | no not at all | 14:02 |
| jrosser | maybe this means something terrible is happening :) | 14:02 |
| jrosser | do you have an example? | 14:02 |
| *** ysandeep|afk is now known as ysandeep | 14:06 | |
| noonedeadpunk | jrosser: I was looking at all things like https://opendev.org/openstack/openstack-ansible/src/branch/master/inventory/group_vars/all/infra.yml#L24 or https://opendev.org/openstack/openstack-ansible-galera_server/src/branch/master/defaults/main.yml#L129 which kind of means that clusters would be over ssh network, right? | 14:17 |
| noonedeadpunk | or well. maybe we set ansible_host to management network.... | 14:18 |
| noonedeadpunk | hm, I actually haven't checked if we do :D | 14:18 |
| noonedeadpunk | ah, yes, indeed, we set ansible_host == container_address https://opendev.org/openstack/openstack-ansible/src/branch/master/osa_toolkit/generate.py#L215-L216 | 14:29 |
| jrosser | right makes sense | 14:34 |
| jrosser | but kind of unclear / magical at the same time | 14:34 |
| noonedeadpunk | yeah... But then I will adjust zookeeper role as well to be same magical... | 14:39 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Add zookeeper deployment https://review.opendev.org/c/openstack/openstack-ansible/+/864750 | 14:45 |
| damiandabrowski | so noonedeadpunk was right, when I explicitly set venv_wheel_build_enable=True, then uWSGI wheel was built on repo container. | 14:57 |
| damiandabrowski | but I'm a bit confused now. Maybe I misunderstood a logic here: https://opendev.org/openstack/ansible-role-python_venv_build/src/branch/master/vars/main.yml#L80 | 14:57 |
| damiandabrowski | But does it mean that if I have lets say 3 glance containers but run glance playbook with limit to only one of them, then wheels won't be built on repo container? | 14:58 |
| noonedeadpunk | yes, that's correct | 15:11 |
| noonedeadpunk | but if you run with serial - they will | 15:12 |
| noonedeadpunk | eventually, when you setup/upgrade there's assumption, that you don't use limit, but rely on playbook serial. | 15:12 |
| noonedeadpunk | as limit in couple of roles can just break deployment | 15:13 |
| noonedeadpunk | however I wonder... If wheels are already built, and you run with limit... will it utilize already present wheels? As I guess with current logic it won't? | 15:14 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/ansible-role-zookeeper master: Initial commit to zookeeper role https://review.opendev.org/c/openstack/ansible-role-zookeeper/+/864752 | 15:16 |
| damiandabrowski | yeah it won't: https://opendev.org/openstack/ansible-role-python_venv_build/src/branch/master/defaults/main.yml#L81 | 15:18 |
| damiandabrowski | I guess we can remove `and (venv_wheel_build_enable | bool)` from this condition? | 15:19 |
| damiandabrowski | these args shouldn't harm even if wheels are not present on repo container, right? | 15:20 |
| noonedeadpunk | I think if wheels are not present, then install will fail out, isn't it? | 15:26 |
| noonedeadpunk | tbh, I think we could jsut hardcode venv_wheel_build_enable in user_variables or smth.... | 15:27 |
| noonedeadpunk | or document that somewhere | 15:27 |
| damiandabrowski | bah, i thought there is some sort of fallback if wheels are not present | 15:30 |
| damiandabrowski | but I agree, hardcoding venv_wheel_build_enable=True seems reasonable | 15:30 |
| noonedeadpunk | well it's reasonable for osa context, unless you try to use it elsewhere | 15:32 |
| noonedeadpunk | but yeah, you still likely will do some overrides... | 15:32 |
| damiandabrowski | if it's reasonable from osa context, we can override it in a place like /opt/openstack-ansible/inventory/group_vars/all/all.yml :D | 15:33 |
| noonedeadpunk | or just document | 15:33 |
| noonedeadpunk | tbh I dunno what's better here | 15:34 |
| noonedeadpunk | Hm, I'm catching this on master: https://paste.openstack.org/show/b8dfAKhU3Dffle6Oe2ua/ | 15:46 |
| dok53 | Me again :( so I got cinder talking to the backend SAN by doing what noonedeadpunk suggested with the venv pip installation, I also needed to install an old, old naviseccli but that's because our system is so old which is why I may move to a different backend rather than waste time on it. But before I do has anyone any input on this https://paste.openstack.org/show/bBilT9XJZvIx2ty303UR/ The volumes create & delete no problem and I can see it | 15:47 |
| dok53 | created on our SAN it's just I can't attach it | 15:47 |
| noonedeadpunk | I'm not sure what triggers these warnings though | 15:49 |
| noonedeadpunk | dok53: sorry, no idea. Anything on compute side? | 15:49 |
| noonedeadpunk | for example here: https://zuul.opendev.org/t/openstack/build/a8b1124b603944a3bf5e52c507896521/log/job-output.txt#6704 | 15:53 |
| noonedeadpunk | and couple more times down the line | 15:54 |
| noonedeadpunk | and looks like this warning is even not when ansible runs | 15:56 |
| noonedeadpunk | maybe it's from zuul ansible.... | 15:57 |
| dok53 | Not to worry noonedeadpunk, think I'll call it quits, that system is too old. Thanks, have a good evening I'm off | 16:00 |
| *** dviroel is now known as dviroel|brb|lunch | 16:00 | |
| noonedeadpunk | no it's not zuul as I do see that in aio as well | 16:14 |
| noonedeadpunk | maybe it's some ara thing? | 16:14 |
| noonedeadpunk | because that;'s totally not when our ansible runs https://paste.openstack.org/show/bkhcbXcp7FOxf4R2kRlg/ | 16:15 |
| *** ysandeep is now known as ysandeep|out | 16:15 | |
| opendevreview | Damian Dąbrowski proposed openstack/ansible-role-uwsgi master: Install OpenSSL development headers https://review.opendev.org/c/openstack/ansible-role-uwsgi/+/864783 | 17:09 |
| damiandabrowski | ^ I've fixed this patch to install openssl dev headers on a host where uWSGI is being built | 17:11 |
| *** dviroel|brb|lunch is now known as dviroel | 17:12 | |
| noonedeadpunk | damiandabrowski: any reason not to add these package names to vars? | 17:21 |
| noonedeadpunk | jsut add variable uwsgi_devel_distro_packages in vars/debian.yml vars/redhat.yml | 17:22 |
| damiandabrowski | separate var may be actually a good idea, let me fix it | 17:32 |
| opendevreview | Damian Dąbrowski proposed openstack/ansible-role-uwsgi master: Install OpenSSL development headers https://review.opendev.org/c/openstack/ansible-role-uwsgi/+/864783 | 17:35 |
| noonedeadpunk | fwiw, zookeeper role looks quite fair, except SSL https://review.opendev.org/c/openstack/ansible-role-zookeeper/+/864752 | 17:51 |
| noonedeadpunk | also thanks to fungi redirect is not an issue anymore | 17:52 |
| noonedeadpunk | I've just re-issued re-check for integrated repo patch | 17:52 |
| noonedeadpunk | I will work on SSL in follow-up patch I guess | 17:55 |
| damiandabrowski | great, thanks Dmitriy | 17:57 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Add zookeeper deployment https://review.opendev.org/c/openstack/openstack-ansible/+/864750 | 17:58 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Add zookeeper deployment https://review.opendev.org/c/openstack/openstack-ansible/+/864750 | 17:58 |
| *** dviroel_ is now known as dviroel | 18:16 | |
| prometheanfire | master branch still == zed? | 19:18 |
| *** dviroel is now known as dviroel|out | 21:15 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!