opendevreview | OpenStack Proposal Bot proposed openstack/openstack-ansible master: Imported Translations from Zanata https://review.opendev.org/c/openstack/openstack-ansible/+/948038 | 02:26 |
---|---|---|
f0o | noonedeadpunk: so I guess all is well for me since I use NFS | 06:20 |
f0o | I need to read into gnocchi how it actually works and where it dumps data... | 06:21 |
noonedeadpunk | well, you can use swift for data | 06:22 |
noonedeadpunk | a nasty thing about NFS are dead locks in case of any network interruptions, which are extremely annoying | 06:23 |
noonedeadpunk | but, you can migrate to ceph anytime more or less (with some effort ofc) | 06:23 |
f0o | the main concern with ceph is the knowledge gap and unknown compute requirements for it | 06:36 |
f0o | but one step at a time; let's get coordination_hosts deployed and just gnocchi with swift. then deal with ceilometer later since i probably need that override for coordinator entries | 06:42 |
noonedeadpunk | f0o: fwiw, you can also push ceilometer data to prometheus or opentelemetry | 07:10 |
noonedeadpunk | no multi-tenancy though | 07:10 |
f0o | debugging why gnocchi fails to install now xD | 08:09 |
f0o | oddly enough it failed at the package installtion step; some large pip errors... just gonna double check that the containers have connectivity and all | 08:09 |
f0o | didnt know gnocchi doesnt support distro installation | 08:10 |
f0o | at least ubuntu does have gnocchi packages tho | 08:14 |
noonedeadpunk | oh, well... | 08:16 |
noonedeadpunk | that can be the case | 08:16 |
noonedeadpunk | we don't have distro path everywhere indeed | 08:17 |
f0o | ZK installed fine tho; so that's neat! | 08:26 |
f0o | :D | 08:26 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible master: wip https://review.opendev.org/c/openstack/openstack-ansible/+/948111 | 08:49 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible master: wip https://review.opendev.org/c/openstack/openstack-ansible/+/948111 | 08:58 |
f0o | noonedeadpunk: https://paste.opendev.org/show/b5c0nV3S50zKWLEyv42B/ | 09:01 |
f0o | Seems like its not installing build-essentials on debian/ubuntu | 09:01 |
f0o | https://github.com/openstack/openstack-ansible-os_gnocchi/blob/master/vars/debian.yml << its listed as gnocchi_devel_distro_packages but when I check dpkg its not installed | 09:05 |
noonedeadpunk | well, this is what is installed on repo container to build wheels for source install | 09:05 |
f0o | huh | 09:06 |
noonedeadpunk | unless wheels build is disabled | 09:06 |
noonedeadpunk | which can be smth you can try | 09:06 |
f0o | well we got distro as installation method; so I guess the repo container wont build wheels then at all | 09:06 |
noonedeadpunk | try `venv_wheel_build_enable: false` | 09:07 |
f0o | and gnocchi container is now trying to install things with pip which fails since build-essentials arent installed | 09:07 |
noonedeadpunk | yeah | 09:07 |
noonedeadpunk | but I guess it's already false.. | 09:07 |
f0o | so I either install those dev packages in all gnocchi containers or I implement distro pacakges into the gnocchi role... | 09:08 |
f0o | it's swift all over again xD | 09:08 |
f0o | can I force the repo container to build wheels? | 09:09 |
noonedeadpunk | if you have it even... | 09:09 |
f0o | I do have a repo container | 09:09 |
noonedeadpunk | but rly, try venv_wheel_build_enable: false | 09:11 |
noonedeadpunk | but I can't tell more without seeing ansible output tbh | 09:11 |
f0o | TASK [python_venv_build : Show venv install failure message] ****************************************************************************************************************************************************** | 09:12 |
f0o | fatal: [h1_2-gnocchi-container-32776452]: FAILED! => {"changed": false, "msg": "The python packages have failed to install, please check the log file located at /var/log/python_venv_build.log for more information.\n"} | 09:12 |
f0o | and the file it refers to is just a big log where the error is that it cannot build python-systemd because pkg-config doesnt exist on gnocchi contianer | 09:12 |
noonedeadpunk | and venv_wheel_build_enable: false was set? | 09:12 |
f0o | not set yet, setting now | 09:13 |
f0o | TASK [python_venv_build : Install distro packages for venv build] << this is new | 09:13 |
f0o | well it's compiling now | 09:14 |
f0o | it passed! | 09:17 |
noonedeadpunk | nice :) | 09:28 |
noonedeadpunk | I think there's some logic which will prevent wheels build for distro setups regardless | 09:28 |
noonedeadpunk | I'm not sure what it is though... | 09:28 |
f0o | do I need any additional config to add ceilometer to horizon? | 11:19 |
f0o | or should it "just work"? | 11:19 |
f0o | reason I ask is because I'm not seeing ceilometer in horizon... but I got os-ceilometer installed correctly | 11:20 |
noonedeadpunk | there's no horizon ui afaik | 11:22 |
f0o | oh | 11:22 |
noonedeadpunk | ceilometer does not have any api or client | 11:22 |
noonedeadpunk | there's an api for gnocchi | 11:22 |
noonedeadpunk | but I don't think there;s a horizon dashboard for it | 11:22 |
f0o | hrm and aodh? | 11:22 |
f0o | I could swear there used to be a UI in mitaka times | 11:23 |
f0o | AKA an eternity ago | 11:23 |
noonedeadpunk | oh, well. back then there was an API part of ceilometer | 11:23 |
noonedeadpunk | it was dropped around rocky | 11:23 |
f0o | haha | 11:23 |
f0o | just gotta find a way to verify ceilometer is working now | 11:24 |
noonedeadpunk | I don't see aodh dashboard with quick googling | 11:24 |
f0o | utility container doesnt have metrics command so probably some client missing like I had with siwft | 11:24 |
noonedeadpunk | well, you can try fetching data from gnocchi | 11:24 |
noonedeadpunk | there should be gnocchiclient | 11:24 |
noonedeadpunk | but again, dunno for distro setup... | 11:25 |
noonedeadpunk | it might be missing indeed | 11:25 |
f0o | yessir! | 11:25 |
noonedeadpunk | yeah: https://opendev.org/openstack/openstack-ansible/src/branch/master/inventory/group_vars/utility_all/defaults.yml#L43-L49 | 11:25 |
f0o | a quick python3-gnocchiclient later and boom metric list shows stuff | 11:26 |
f0o | So I assume it's working then :D | 11:27 |
noonedeadpunk | just for source we grep upper-constraints for clients | 11:27 |
noonedeadpunk | so we don't need t odefine the list explicitly | 11:27 |
f0o | hrm | 12:31 |
f0o | something went fubar | 12:32 |
f0o | not sure if it's cinder after coordination hosts or if glance just never recovered | 12:32 |
f0o | well `openstack image save ...` works so I guess it's ciner | 12:33 |
f0o | yeah cant even create empty volumes | 12:37 |
f0o | cant see any errors | 12:37 |
f0o | any pointers? | 12:39 |
f0o | it's not erroring its just stuck in creating | 12:40 |
f0o | and logs dont show anything useful | 12:40 |
noonedeadpunk | zookeeper effect? | 12:42 |
noonedeadpunk | are cinder-volume service is running and reporting healthy | 12:42 |
noonedeadpunk | as I'd guess it's smth between cinder-scheduler and cinder-volume | 12:43 |
f0o | ah | 12:43 |
f0o | oslo_messaging.exceptions.MessageDeliveryFailure: Unable to connect to AMQP server on 10.20.1.114:5671 after inf tries: Exchange.declare: (406) PRECONDITION_FAILED - inequivalent arg 'durable' for exchange 'cinder-backup_fanout' in vhost 'cinder': received 'false' but current is 'true' | 12:43 |
f0o | so Ceilometer nuked cinder amqp it seems | 12:43 |
f0o | or at least something nuked it | 12:44 |
f0o | I had this shit before | 12:44 |
f0o | stupid backups | 12:44 |
f0o | xD | 12:44 |
noonedeadpunk | you had indeed | 12:47 |
noonedeadpunk | but I wonder why this happens | 12:47 |
f0o | no clue but deleting the exchange and restarting the scheduler always fixes it | 12:49 |
noonedeadpunk | My guess is that you might be getting older rabbitmq then expected by some services... | 12:50 |
f0o | all rabbit nodes are identical tho | 12:50 |
noonedeadpunk | as with distro install from some releases before, we use rabbitmq/mariadb from native repos | 12:51 |
noonedeadpunk | what version of rabbit/osa? | 12:51 |
f0o | rabbitmq-server 4.0.3-1 | 12:51 |
f0o | osa is stable/2024.2 afaik | 12:51 |
noonedeadpunk | yeah, it's modern enough for sure | 12:51 |
f0o | I think ti's actually cinder-backup that creates it wrong | 12:51 |
noonedeadpunk | well, I'm using cinder-backup with 2024.2 and it's fine | 12:52 |
f0o | but that is same version as all other cinders | 12:52 |
f0o | hrm | 12:52 |
f0o | I also noticed that ceilometer notification agent is hammering the DB like hell | 12:54 |
noonedeadpunk | DB? Does ceilometer even has an access to DB? | 12:54 |
f0o | mariadb went from ~3-10% to 200+ | 12:54 |
f0o | when I turn off the ceilometer-central lxc it stops | 12:55 |
f0o | something might be crashlooping there | 12:55 |
f0o | possibly also amqp related since the fanout was borken | 12:55 |
noonedeadpunk | it's probably gnocchi | 12:55 |
f0o | unfortunately I'm not at the office anymore so i cant easily check | 12:55 |
noonedeadpunk | as I think it uses mysql for coordination by default | 12:56 |
noonedeadpunk | but it makes total sense to re-configure gnocchi to use zookeeper instead | 12:56 |
f0o | well I got ZK installed and running | 12:57 |
f0o | ah it doesnt configure it by defalt | 12:57 |
f0o | I see | 12:57 |
noonedeadpunk | I think it was missed toghether with ceilometer... | 12:59 |
f0o | hehe | 12:59 |
f0o | also | 12:59 |
f0o | h1_2-swift-proxy-container-a9384c07 | 12:59 |
noonedeadpunk | Will try to check on that for 2025.1.... | 12:59 |
f0o | eh | 12:59 |
f0o | swift-proxy-server[177020]: ModuleNotFoundError: No module named 'ceilometermiddleware' | 12:59 |
f0o | it also nuked siwft xD | 12:59 |
noonedeadpunk | right... doh | 12:59 |
noonedeadpunk | I'd expect it being in swift requirements rather then needed to be defined explicilty | 13:00 |
noonedeadpunk | but this one is an easy patch | 13:00 |
f0o | :D | 13:00 |
f0o | just a fun cat and mouse again | 13:00 |
f0o | need ceilo for swift accounting; ceilo breaks swift | 13:00 |
f0o | need swift for gnocchi | 13:01 |
f0o | need gnocchi for ceilo | 13:01 |
f0o | so many catch.22 | 13:01 |
f0o | :D | 13:01 |
noonedeadpunk | oh, wait: https://opendev.org/openstack/openstack-ansible-os_swift/src/branch/master/defaults/main.yml#L327 | 13:01 |
noonedeadpunk | but you;'re having distro setup... | 13:01 |
noonedeadpunk | ah damn it | 13:01 |
f0o | yep | 13:01 |
noonedeadpunk | so apt install python3-ceilometermiddleware ? | 13:02 |
f0o | that's what I'm doing now | 13:02 |
noonedeadpunk | will you be able to push the patch to add it here? https://opendev.org/openstack/openstack-ansible-os_swift/src/branch/master/vars/debian.yml#L32 (and also for redhat) | 13:03 |
f0o | that seems to work | 13:04 |
f0o | will do when in the office again | 13:04 |
f0o | on phone its tedious | 13:04 |
f0o | so that explains gnocchi hammering the DB too since it failed continously to fetch some gnocci-config from swift and restarted | 13:04 |
noonedeadpunk | yeah... | 13:06 |
f0o | fun :D | 13:06 |
NeilHanlon | noonedeadpunk: btw i added some docs review for https://review.opendev.org/c/openstack/openstack-ansible-ops/+/943866/11 | 13:08 |
NeilHanlon | also good morning.. i hope you're feeling better! | 13:08 |
f0o | noonedeadpunk: for later; how can I "detect" if ceilometer is being pulled here? or should it just install the middleware no matter what? | 13:11 |
f0o | also how do I set gnocchi_coordination_url to use zookeeper correctly? I see https://opendev.org/openstack/openstack-ansible-os_gnocchi/src/branch/master/defaults/main.yml#L94-L98 and wouldnt even know how to replicate this for ZK | 13:14 |
f0o | https://opendev.org/openstack/openstack-ansible-os_cinder/src/branch/master/vars/main.yml#L151-L158 << even worse but maybe I can just copy paste that? :D | 13:16 |
noonedeadpunk | f0o: yeah, I guess you can just install it | 13:25 |
noonedeadpunk | NeilHanlon: quite some edits | 13:25 |
noonedeadpunk | and that time I even used a spell checker | 13:26 |
NeilHanlon | 😅 sorry! | 13:28 |
NeilHanlon | my docs team has been beating me about using active voice in technical writing so I can't leave well enough alone | 13:28 |
noonedeadpunk | yeah, right... | 13:29 |
noonedeadpunk | I actually can recall someone from my team discouraging me to use passive voice | 13:30 |
noonedeadpunk | but my english is bad, so I did not get nuances of it I think | 13:30 |
noonedeadpunk | Like Rule4: https://xahteiwi.eu/resources/presentations/devopsdays-berlin-2022/ | 13:31 |
opendevreview | Daniel Preussker proposed openstack/openstack-ansible-os_swift master: Add Ceilometer middleware to distro packages https://review.opendev.org/c/openstack/openstack-ansible-os_swift/+/948194 | 13:31 |
f0o | regarding https://review.opendev.org/c/openstack/openstack-ansible-os_swift/+/948194 ; I have no idea if the package name is correct for redhat; I tried looking on pkgs.org but I realized that it doesnt have the OpenStack repos | 13:32 |
NeilHanlon | English is an awful, awful language. :P | 13:34 |
f0o | better than suomi 💀 | 13:34 |
f0o | luckily I only need to remember that once a year 😅 | 13:35 |
noonedeadpunk | f0o: yeah, it's fine https://mirror.stream.centos.org/SIGs/9-stream/cloud/x86_64/openstack-caracal/Packages/p/ | 13:39 |
f0o | <3 | 13:44 |
opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-ops master: Add a collection for managing encryption of secret data https://review.opendev.org/c/openstack/openstack-ansible-ops/+/943866 | 13:45 |
opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-ops master: Add a collection for managing encryption of secret data https://review.opendev.org/c/openstack/openstack-ansible-ops/+/943866 | 13:46 |
opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-ops master: Fix conditional on ansible_vault_in_place_copy https://review.opendev.org/c/openstack/openstack-ansible-ops/+/947834 | 13:46 |
opendevreview | Merged openstack/openstack-ansible master: Imported Translations from Zanata https://review.opendev.org/c/openstack/openstack-ansible/+/948038 | 14:16 |
opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Add CI jobs for Ceph on Ubuntu 24.04 https://review.opendev.org/c/openstack/openstack-ansible/+/948255 | 16:47 |
opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Add CI jobs for Ceph on Ubuntu 24.04 https://review.opendev.org/c/openstack/openstack-ansible/+/948255 | 16:54 |
noonedeadpunk | potentially, we'd need to do partial backport to 2024.2 | 16:56 |
f0o | noonedeadpunk: ceilometer-agent-notification is absolutely insane; it's not only spamming gnocchi with Kpps but also erroring like hell about all sorts of things | 17:32 |
f0o | are you sure this is stable? | 17:32 |
f0o | 2025-04-25 17:31:56.559 5403 ERROR cotyledon._utils oslo_messaging.exceptions.MessageDeliveryFailure: Unable to connect to AMQP server on 10.20.1.114:5671 after inf tries: Exchange.declare: (406) PRECONDITION_FAILED - inequivalent arg 'durable' for exchange 'heat' in vhost 'swift': received 'false' but current is 'true' | 17:33 |
f0o | I dont even have Heat | 17:33 |
f0o | and why would Heat be on vhost Swift? | 17:33 |
f0o | it's really spamming it too | 17:34 |
f0o | on gnocchi I get the error | 17:34 |
f0o | Apr 25 17:34:27 h1-2-gnocchi-container-32776452 uwsgi[205]: Fri Apr 25 17:34:27 2025 - *** uWSGI listen queue of socket "10.20.3.251:8041" (fd: 3) full !!! (100/100) *** | 17:34 |
f0o | but all of this is the ceilometer-agent-notification going berserker | 17:34 |
f0o | and yes these logs are on all controllers, not just that singular one | 17:40 |
f0o | 2025-04-25 17:42:39.841 92 ERROR cotyledon._utils oslo_messaging.exceptions.MessageDeliveryFailure: Unable to connect to AMQP server on 10.20.2.175:5671 after inf tries: Exchange.declare: (406) PRECONDITION_FAILED - inequivalent arg 'durable' for exchange 'nova' in vhost 'swift': received 'false' but current is 'true' << also this... | 17:42 |
opendevreview | Ivan Anfimov proposed openstack/ansible-role-httpd master: tox: Remove basepython https://review.opendev.org/c/openstack/ansible-role-httpd/+/948259 | 17:44 |
opendevreview | Ivan Anfimov proposed openstack/ansible-role-httpd master: tox: Bump minimum version and remove basepython https://review.opendev.org/c/openstack/ansible-role-httpd/+/948259 | 17:45 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible master: tox: Bump minimum version and remove basepython https://review.opendev.org/c/openstack/openstack-ansible/+/948261 | 17:47 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible master: tox: Bump minimum version and remove basepython https://review.opendev.org/c/openstack/openstack-ansible/+/948261 | 17:48 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible-os_horizon master: Loop around module list in horizon_translations_update https://review.opendev.org/c/openstack/openstack-ansible-os_horizon/+/656045 | 18:09 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible-os_horizon master: Loop around module list in horizon_translations_update https://review.opendev.org/c/openstack/openstack-ansible-os_horizon/+/656045 | 18:09 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible-os_horizon master: Loop around module list in horizon_translations_update https://review.opendev.org/c/openstack/openstack-ansible-os_horizon/+/656045 | 18:11 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible-os_horizon master: Loop around module list in horizon_translations_update https://review.opendev.org/c/openstack/openstack-ansible-os_horizon/+/656045 | 18:13 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible-os_horizon master: Loop around module list in horizon_translations_update https://review.opendev.org/c/openstack/openstack-ansible-os_horizon/+/656045 | 18:15 |
-opendevstatus- NOTICE: Gerrit is getting restarted to pick up container image updates. It should only be gone for a moment. | 18:15 | |
f0o | I've rerun setup-everything again and I still need to hard kill ceilometer-agent-notification.service on all ceilo containers because it's just absolutely going insane and throwing the oddest errors like above | 18:41 |
f0o | any openstack service on the vhost swift... sahara? sure thing, nova? ofc neutron? probably | 18:41 |
f0o | and while it does it, it's also ddos'ing gnocchi to a point where it's just crashing. all of 5*16 instances of it | 18:42 |
f0o | also after using ZK for coordinator on ceilometer the ceilometer-polling service fails with ModuleNotFoundError: No module named 'kazoo' | 18:44 |
f0o | so os-ceilometer needs python3-kazoo in distro packages at least | 18:47 |
f0o | no clue what to do about the ceilometer-agent-notification.service going berserker | 18:48 |
f0o | hrm i now have all those exchanges in all my vhosts... i never installed sahara or trove or similar and yet I got those exchanges on all vhosts | 19:06 |
f0o | guess ill need to manually clean those up and hope not to delete some exchange thats needed | 19:06 |
opendevreview | Merged openstack/openstack-ansible master: Maximize amount of ANSIBLE_FORKS for CI https://review.opendev.org/c/openstack/openstack-ansible/+/947128 | 20:07 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible-os_horizon master: Loop around module list in horizon_translations_update https://review.opendev.org/c/openstack/openstack-ansible-os_horizon/+/656045 | 20:40 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible master: wip https://review.opendev.org/c/openstack/openstack-ansible/+/948111 | 21:14 |
opendevreview | Ivan Anfimov proposed openstack/openstack-ansible master: wip https://review.opendev.org/c/openstack/openstack-ansible/+/948111 | 21:29 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!