Tuesday, 2022-05-24

*** ysandeep|out is now known as ysandeep|rover03:24
*** ysandeep|rover is now known as ysandeep|afk04:24
*** ysandeep|afk is now known as ysandeep|rover04:45
noonedeadpunkjrosser: let's just hardcode https://opendev.org/openstack/openstack-ansible-tests/src/branch/master/test-vars.yml#L21 for now (not relying on env var) and will revert once branch06:36
jrosserok06:37
noonedeadpunkah, but I see now how it's different06:37
jrosserwe don't have an easy variable that we can use to set the branch of "all services" in the os_<blah> roles06:38
jrosserlike a openstack_git_install_branch here would be just what we need https://github.com/openstack/openstack-ansible-os_keystone/blob/master/defaults/main.yml#L4106:39
jrosserbut thats for another time, this is kind of a big mess06:39
jrosserand unfortunatley i have made a mistake with the gluster/rockylinux stuff and we're broken06:39
noonedeadpunkbut we set them all06:40
jrossernot in the functional tests do we?06:40
jrosseroh sorry you are right, we do06:40
noonedeadpunkwe do, yeah. But it's based on TESTING_BRANCH at the first place https://opendev.org/openstack/openstack-ansible-tests/src/branch/master/test-vars.yml#L14706:40
noonedeadpunknow I'm thinking, that we should check if TESTING_BRANCH is already set06:41
jrosserwe should make SERVICES_BRANCH or something for that06:41
jrosseranyway, this is pretty important to get fixed whichever way https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/84299906:42
jrosserthough of course theres no actual test coverage there with that patch06:42
jrosserwhich is how this got broken in the first place :(06:42
noonedeadpunkgive me several mins, will make some patch to show what I mean :)06:43
jrossersure06:43
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Provide opportunity to override TESTING_BRANCH  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84310806:58
noonedeadpunkjrosser: not sure if it will work though :) ^06:58
noonedeadpunkat least it was the idea07:00
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Allow lxc container backing store to be defined as part of the scenario  https://review.opendev.org/c/openstack/openstack-ansible/+/84301107:06
jrossernoonedeadpunk: i think it will break on u-c still being master07:09
jrosserfrom the cached zuul repo07:09
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Add zuul jobs and template for testing different LXC backing stores  https://review.opendev.org/c/openstack/openstack-ansible/+/84302707:13
noonedeadpunkah yes, as we're using local zuul one...07:15
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Provide opportunity to override TESTING_BRANCH  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84310807:18
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Provide opportunity to override TESTING_BRANCH  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84310807:32
*** ysandeep|rover is now known as ysandeep|rover|lunch08:02
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Provide opportunity to override TESTING_BRANCH  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84310808:17
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Allow lxc container backing store to be defined as part of the scenario  https://review.opendev.org/c/openstack/openstack-ansible/+/84301108:18
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Add zuul jobs and template for testing different LXC backing stores  https://review.opendev.org/c/openstack/openstack-ansible/+/84302708:19
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Provide opportunity to override TESTING_BRANCH  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84310808:19
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Provide opportunity to override TESTING_BRANCH  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84310808:25
noonedeadpunkok, I dunno... environment passed to the run_test.sh obviosly not picked up on runtime...08:25
noonedeadpunkor test-ansible-env-prep.sh even not executed during test...08:25
noonedeadpunkSo centos stil fails...08:26
jrosseri was confused by that08:33
jrosseri had a patch yesterday which added set -x to test-ansible-env-prep.sh becasue it was really confusing what was happening08:34
*** ysandeep|rover|lunch is now known as ysandeep|rover08:38
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Provide opportunity to override TESTING_BRANCH  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84310808:48
* noonedeadpunk trying to make "projects update" presentation08:51
opendevreviewAndrew Bonney proposed openstack/openstack-ansible-ops master: Update beat templates from 7.11 reference  https://review.opendev.org/c/openstack/openstack-ansible-ops/+/77884910:20
*** dviroel|out is now known as dviroel11:21
noonedeadpunkso Rocky now fails with gluster installation: Failed to download metadata for repo 'centos-gluster8': Cannot prepare internal mirrorlist: No URLs in mirrorlist11:35
mbahnHi, we're planning to add multiple federated service providers. Both openidc: one for keycloak (don't mistake that with keystone) authentication and one for another external openidc provider) but I just saw that only a single trusted IdP in keystone_sp.trusted_idp_list is supported. Why is that so? Is it because of the keystone httpd config?11:37
noonedeadpunkI think only jrosser / andrewbonney can answer that :)11:43
jrossernoonedeadpunk: i have a patch for glusterfs12:26
noonedeadpunkoh?12:26
jrosserhttps://review.opendev.org/c/openstack/openstack-ansible-plugins/+/84299912:26
noonedeadpunkah, lol, indeed.12:26
jrosserbut of course that fails because of test repo stuff12:28
noonedeadpunkyeah-yeah...12:30
noonedeadpunkI really wonder why testing_branch is not being seen inside run_tests.sh12:38
noonedeadpunkI kind of have a guss why....12:39
noonedeadpunkAnd your approach was right with adding new variable12:41
noonedeadpunkAnyway can be fixed with follow-up12:41
jrossermbahn: it is true that trusted_idp_list in the keystone ansible role, so you can potentially define several. However, there are several places in the current code where the list is assumed to have only one entry. You can see those here https://codesearch.opendev.org/?q=trusted_idp_list.0.&i=nope&literal=nope&files=&excludeFiles=&repos=12:42
jrossermbahn: if you want to use more two OIDC IdP then it would probably be only small work to iterate over those lists where there is currently use of the '.0' element12:43
jrosserthat would be a really nice thing to contribute to OSA if you were able to12:44
jrossermy team have a keystone<>keycloak OIDC integration already so can maybe give you some pointers if needed12:44
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests master: Add SERVICES_BRANCH variable for better control  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84314012:51
opendevreviewMerged openstack/openstack-ansible-tests master: Provide opportunity to override TESTING_BRANCH  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84310813:12
mbahn@jrosser: okay thanks for the info :) I'll have a look at it and see if I can contribute. Just btw, I already have one OIDC integration running and working (EGI Check-In if you're interested) and tried to add the keycloak integration for our own institute13:18
mbahnJust to get the info right: if I add the second IdP in the user_variables.yml it would be translated as '.1' in Ansible?13:19
jrossermbahn: ok cool, we had really quite good success building an openstack-ansible all-in-one test setup for working on this stuff13:19
jrossermbahn: well, yes it would be .1, but the better thing to do would be to replace the explicit indexing into the list with a 'for' loop over all the list elements, however many there might be13:20
jrosserthis whole section here would be a for loop https://github.com/openstack/openstack-ansible-os_keystone/blob/master/templates/keystone-httpd.conf.j2#L48-L10113:22
mbahncool thanks. Let's see if I can manage to do something in the next days13:22
jrosseri don't know if this would also need seperate data for each idp https://github.com/openstack/openstack-ansible-os_keystone/blob/master/tasks/keystone_federation_sp_idp_setup.yml13:24
jrosseroh, well thats already iterating over the trusted_idp_list here https://github.com/openstack/openstack-ansible-os_keystone/blob/master/tasks/main_keystone_federation_sp_idp_setup.yml#L213:25
mgariepymariadb 10.6.8 is avalaible now :D 13:27
* jrosser wonders what new bugs there will be13:28
noonedeadpunkand 10.8.3 :D13:28
jrosseri would rather they make centos-9 packages13:28
noonedeadpunkit's not LTS though13:28
mgariepywell it does fix a couple cve..13:28
noonedeadpunkdoh. and no ubuntu 22.04? Rly?13:29
mbahnwell, good thing that I already worked with the keystone-httpd.conf.j2 (see last commit) :D understanding what ansible does with j2 files is easier now13:29
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Allow lxc container backing store to be defined as part of the scenario  https://review.opendev.org/c/openstack/openstack-ansible/+/84301113:37
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Add zuul jobs and template for testing different LXC backing stores  https://review.opendev.org/c/openstack/openstack-ansible/+/84302713:38
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Add zuul jobs and template for testing different LXC backing stores  https://review.opendev.org/c/openstack/openstack-ansible/+/84302713:38
jrosseri wonder why SERVICES_BRANCH ends up as master still13:47
opendevreviewMerged openstack/openstack-ansible stable/xena: Allow fast SSH cipher for upgrade jobs  https://review.opendev.org/c/openstack/openstack-ansible/+/82925813:58
noonedeadpunkhm14:00
noonedeadpunkrun_tests.sh doesn't pass it?14:01
jrossermaybe we should print the values of these as they are passed in as well for debugging14:15
jrosserbefore assigning a default14:16
jrosserhopefully this is going to fix the rocky/glusterfs trouble https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/84299914:25
jrosserrocky jobs can be seen passing with that patch here https://review.opendev.org/c/openstack/openstack-ansible/+/84301114:26
noonedeadpunknice14:38
jrosseroh did you see i think now lxc_hosts/container_create don't need openstack-ansible-tests any more14:42
noonedeadpunkI saw you pushed several changed to move that to integrated and loved that :)14:53
opendevreviewMerged openstack/openstack-ansible-plugins master: Fix installation of glusterfs repo on rocky linux  https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/84299914:57
noonedeadpunk#startmeeting openstack_ansible_meeting15:00
opendevmeetMeeting started Tue May 24 15:00:24 2022 UTC and is due to finish in 60 minutes.  The chair is noonedeadpunk. Information about MeetBot at http://wiki.debian.org/MeetBot.15:00
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.15:00
opendevmeetThe meeting name has been set to 'openstack_ansible_meeting'15:00
noonedeadpunk#topic roll call15:00
noonedeadpunko/15:00
damiandabrowski[m]hey!15:00
mgariepyhalf there o/15:01
jrossero/ hello15:02
noonedeadpunk#topic office hours15:05
noonedeadpunkI think we're super close to merging repo, centos and ubuntu15:07
noonedeadpunkI suggest creating milestone release after that15:08
noonedeadpunkand would be awesome to make it this week actually... maybe next...15:08
jrosserrepo server gluster stuff just has a few cleanup patches to merge which i think are now ok15:09
jrossercentos is very close, just the pam_env thing to verify and merge15:09
damiandabrowski[m]jrosser: have You managed to solve this issue with mounting cephfs after container restart or You need some help?15:09
jrosserhmm?15:10
jrosseri've not been looking at cephfs15:10
damiandabrowski[m]glusterfs* sorry ;)15:10
jrosseroooh - do we still have trouble there15:10
jrosseroh it was a race condition with systemd wasnt it15:11
damiandabrowski[m]yeah, exactly15:11
jrossertbh i have not spent any time on that15:11
jrosseralso i don't really have any great idea after reading the systemd bug report15:11
damiandabrowski[m]i think there might be some workaround with putting systemd-mount behind systemd service15:13
damiandabrowski[m]I'll try to spend some time on this during this week but can't promise anything :/15:13
noonedeadpunkOk, next to that I've started creating presentation for project update15:15
noonedeadpunkI think deadline for it is end of the week for it to be shown/promoted during summit15:16
jrossersomething is also wrong with repo server upgrades here https://zuul.opendev.org/t/openstack/build/7e33eaf8cfb5445c92bcbe38647b792d/log/job-output.txt#19556-1955815:16
jrosseri've not seen that before15:16
*** ysandeep|rover is now known as ysandeep|out15:17
noonedeadpunkunarchive issue? interesting...15:17
noonedeadpunkHere's the link for presentation https://docs.google.com/presentation/d/1INXXvAY9MuwASYW4eXLye7YvFaosv5K-7XS7jX6AlpI/edit?usp=sharing15:17
noonedeadpunkfeedback is welcome. And also if somebody wants to participate in writing down sound or add some slides - let me know:)15:17
damiandabrowski[m]good job so far!15:19
jrosseri wonder if the unarchive starts before the mount is made completely15:19
NeilHanlonslides look great so far (:15:20
noonedeadpunkjrosser: that would be weird given it's systemd managing mount15:20
jrosseryeah, though i don't know if it's asynchronous15:21
noonedeadpunkI'd say ansible module should wait until service/mount started15:21
noonedeadpunkBut it look like some permissions issue tbh...15:22
noonedeadpunk`Cannot change ownership to uid 998, gid 33`15:22
jrosseroh now thats interesting actually15:23
jrosseri don't think that we ensure that the repo / nginx uid and gid are consistent between the repo servers15:23
jrosserthat could go wrong15:23
jrosserthough in this upgrade case it should be only one of then makes the archive and unpacks it into the glusterfs15:24
jrosserbut i had been thinking we need to do something about the uid/gid anyway15:25
noonedeadpunkoh yes, actually that's good point15:25
noonedeadpunkbut are you sure it's about same server?15:25
noonedeadpunkas it's infra job, which means there're 3 of them?15:26
jrosserthats right15:26
noonedeadpunkMaybe we archived on one, and unpackking on another one? But unlikely...15:26
jrosserhah well look at the failing tasks - its on two different ones15:27
noonedeadpunkwe actually need to run_once this https://opendev.org/openstack/openstack-ansible-repo_server/src/branch/master/tasks/repo_pre_install.yml#L80-L88 as well :)15:28
noonedeadpunkBut good that we didn't15:28
jrossersomething very odd is happening there15:28
jrosserit's running a bunch of tasks against only two of the 315:29
noonedeadpunkI think we have serial: 1, 100%15:29
noonedeadpunkso "as designed"?15:29
noonedeadpunkhttps://opendev.org/openstack/openstack-ansible/src/branch/master/playbooks/repo-install.yml#L2515:29
jrosseraah right yes15:30
jrosserso we need is_first_play_host type of thing like the other roles15:31
jrosseri can look at that15:31
jrosserit could be at the moment we are lucky that uid/gid are assigned with the same algorithm in 3 identical containers15:32
jrosserthats not going to be the case for long lived deployments or cross-OS in the future15:33
noonedeadpunkyes, totally15:34
noonedeadpunkI haven't actually looked much in Octavia PKI since I pushed path. I hope to get to it tomorrow15:35
noonedeadpunkAs it;s smth we need to land sooner better15:35
jrosserwe could look at bugs? i think a bunch have been addressed15:39
noonedeadpunk#topic bug triage15:41
noonedeadpunkyeah. I can't recall anyting new, but we closed a lot lately15:41
noonedeadpunk#link https://bugs.launchpad.net/openstack-ansible/+bug/197376215:42
noonedeadpunkjrosser: I think you was working on this one?15:42
jrosseri didnt make a patch, but basically we template out a file that is already provided by the distro15:43
jrosserthen i think our changes to make it ext4 get set back to btrfs15:43
noonedeadpunkuh15:44
jrosseri thought we could make a systemd drop-in to override the fs type= which would then be persistent15:44
jrosserbut i didnt try it yet15:44
noonedeadpunkSo we basically should be using systemd overrides for that?15:44
jrossertheres some vauge comment in the original file about it being a temporary placeholder until some later version of systemd15:45
noonedeadpunkI think https://opendev.org/openstack/openstack-ansible-lxc_hosts/commit/337ddf87802c7597b7c51a071eb9829a4b87defa is related15:45
jrosserso it might be that we need to template it out, and also to make the drop in as well15:45
noonedeadpunkBut I'm not sure it makes things better or worse15:45
jrosserwell, i see this on my Xena deploys today15:46
noonedeadpunkCommit is for master only. And I think it makes things even worse15:46
jrosserdo they both make the same unit though?15:47
jrosserhow is it worse15:47
noonedeadpunkIt creates another mount? for same endpoints15:47
noonedeadpunkAnd kind of not ensuring that existing one is stopped/disabled15:48
noonedeadpunkI can look at that15:49
noonedeadpunkAnd smth backportable should be made as well15:49
noonedeadpunkfor stable branches15:49
jrosseri have an AIO here and i think that use of systemd_service overwrites the original unit15:49
jrosserbut iirc the old code does that too15:50
jrosseranyway15:50
noonedeadpunkHuh, I thought we're placing our units only under /etc/system15:50
noonedeadpunkoh, well, if they have same names at the end - it's just an override then15:51
noonedeadpunkhttps://opendev.org/openstack/ansible-role-systemd_mount/src/branch/master/tasks/systemd_mounts.yml#L5615:51
jrosserright, in my AIO i have /etc/systemd/system/var-lib-machines.mount15:52
jrosserand its our one with ext415:52
noonedeadpunkThen I guess it should be fine :)15:54
noonedeadpunkthe only part left - backport smth to stable branches15:54
noonedeadpunkI'm thinking to just put simple override of fs type to /etc/systemd/system and be done with it15:55
jrosserwe should look back though eavesdrop to what spatel shared15:55
spatelReading thread 15:57
jrosserthis is what we tread on https://packages.ubuntu.com/search?suite=focal&arch=any&mode=exactfilename&searchon=contents&keywords=var-lib-machines.mount15:58
*** dviroel is now known as dviroel|lunch15:58
spatelGot it :) 15:58
noonedeadpunkI think it's quite clear what the issue is...16:00
noonedeadpunk#endmeeting16:00
opendevmeetMeeting ended Tue May 24 16:00:21 2022 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)16:00
opendevmeetMinutes:        https://meetings.opendev.org/meetings/openstack_ansible_meeting/2022/openstack_ansible_meeting.2022-05-24-15.00.html16:00
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/openstack_ansible_meeting/2022/openstack_ansible_meeting.2022-05-24-15.00.txt16:00
opendevmeetLog:            https://meetings.opendev.org/meetings/openstack_ansible_meeting/2022/openstack_ansible_meeting.2022-05-24-15.00.log.html16:00
jrosseri am a little surprised that `systemctl status var-lib-machines.mount` doesnt show the path to the unit file on my Xena deploy16:01
jrosserbut the title is `Virtual Machine and Container Storage (Compatibility)`16:02
noonedeadpunkoh, yes, it's weird indeed...16:04
noonedeadpunkI have some feeling, that overrides does not work for mounts16:07
noonedeadpunkor well, it fully overrides whole unit rather then just specific options16:08
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-lxc_hosts stable/xena: Do not mess up with distro-provided machines mount  https://review.opendev.org/c/openstack/openstack-ansible-lxc_hosts/+/84316916:12
noonedeadpunkI guess it can be as simple as that ^16:17
noonedeadpunkspatel: would be great if you could check that...16:18
spatel+1  looks good so far 16:20
spatelosa provide file will override this one correct? /lib/systemd/system/var-lib-machines.mount16:21
noonedeadpunkNope, it would place it now to `/etc/systemd/system/var-lib-machines.mount` the one in /lib is provided by systemd-containerd16:24
noonedeadpunkSo not to mess up with it (as it will be overriden on package update) we will place ours to /etc/16:24
noonedeadpunkwhich should be a valid override of it16:25
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-ceph_client stable/victoria: Disable setuptools auto discovery  https://review.opendev.org/c/openstack/openstack-ansible-ceph_client/+/84294916:30
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-ceph_client stable/victoria: Use global package_state  https://review.opendev.org/c/openstack/openstack-ansible-ceph_client/+/84240816:31
*** dviroel|lunch is now known as dviroel16:58
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-tests stable/wallaby: Fix rich version for repos that don't have test-requirements  https://review.opendev.org/c/openstack/openstack-ansible-tests/+/84317517:11
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-ceph_client stable/wallaby: Use global package_state  https://review.opendev.org/c/openstack/openstack-ansible-ceph_client/+/84240717:12
opendevreviewMerged openstack/ansible-role-python_venv_build stable/train: Remove tripleo-ci-centos-7-standalone-os-tempest from gate layout  https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/83816917:12
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Allow lxc container backing store to be defined as part of the scenario  https://review.opendev.org/c/openstack/openstack-ansible/+/84301117:15
opendevreviewJonathan Rosser proposed openstack/openstack-ansible master: Add zuul jobs and template for testing different LXC backing stores  https://review.opendev.org/c/openstack/openstack-ansible/+/84302717:16
spatelnoonedeadpunk +118:36
opendevreviewMerged openstack/openstack-ansible-os_keystone stable/wallaby: add oauth support  https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/83278819:39
spatelnoonedeadpunk what is wrong here - https://paste.opendev.org/show/bNm4iaj7UgMyVoip8WDD/19:52
spatelsince i upgraded to 23.3.0 i started getting this error 19:52
spatelsome kind of conflict 19:53
opendevreviewMerged openstack/openstack-ansible-os_octavia stable/wallaby: Fix condition for deleting old amp images  https://review.opendev.org/c/openstack/openstack-ansible-os_octavia/+/84293619:55
NeilHanlonlooks like it wants to downgrade librados2 and librdb120:02
NeilHanlonspatel 👆20:03
NeilHanlonseems like maybe a broken dep or something weird in the ubuntu packages, if I'm reading it correct 🤔20:04
jrosserspatel: can you show `apt policy ceph-common` for that node?20:22
spatelsorry i was in meeting, i am back now.. let me show you20:26
spateljrosser https://paste.opendev.org/show/bdYKdTELLejqADWyHhwz/20:27
jrosserspatel: there is an explanation here i think in the commit message https://review.opendev.org/c/openstack/openstack-ansible-openstack_hosts/+/83865720:32
jrosserthere should be an apt log on your hosts to show you what was installed, and what was upgraded/downgraded20:34
jrosseryou could also look into this as well https://review.opendev.org/c/openstack/openstack-ansible-openstack_hosts/+/838855/320:36
jrosserand add `APT::Install-Recommends "false";` to your hosts apt config20:36
spatelhmm20:37
jrosserthose patches are all to master as we saw the Quincy release of ceph from UCA interfere with expecting to get Octopus from ceph.com20:37
jrosseryou may have a similar but slightly different situation on a stable branch20:37
jrossera combination of looking in the apt log, and what `apt policy` shows you should let you work out whats going on20:38
spatellet me first try APT::Install-Recommends "false"20:38
jrosserwell, it's a little late now20:39
spatelwhy?20:39
jrosseris that a compute node or a controller?20:39
spatelcompute node20:39
jrosserinteresting20:40
jrosserso those patches i showed you fixed a thing that happened when LXC got installed, it has a 'recommended' dependancy of something which in turn required librbd20:40
spatelIn this cloud we have 180 compute nodes life was good and last week i have upgraded to 23.3.0 and today trying to add 30 more compute nodes but encounter this error 20:40
jrosserso a very indirect dependancy20:40
jrosserand we figured out that was due to installing the lxc-templates package really early in the deployment20:41
jrosserlong long before the ceph_client role had made config that only packages from download.ceph.com were to be used20:42
jrosserso $something thats being installed on your compute node is pulling in librbd20:42
jrosserand then later along comes ceph_client and sets up the repo priorities properly, and the previously installed librbd now needs to be downgraded to match the new apt policy20:43
spatelAPT::Install-Recommends "false" didn't help anyway 20:43
opendevreviewMerged openstack/openstack-ansible-os_neutron master: Support dns_domain_keywords extension  https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/84210220:44
jrosserif a too new librbd has already been installed, then yes, it won;t make any difference20:44
jrosserwhich branch (name?) too late for me to parse release numbers :)20:45
spatelhow about if i remove that apt_pin ?20:45
spatelwallaby / 23.3.020:45
jrosserremove?20:45
spatellibrbd120:45
spatelThis one currently installed 20:46
spatelii  librbd1                               16.2.7-0ubuntu0.21.04.1~cloud2        amd64        RADOS block device client library20:46
spatelOSA trying to install 15.2.x 20:46
jrosserit would be very helpful to know why that got installed20:46
jrosserwhich is where looking in the apt log will help you20:47
spatelLet me check apt logs 20:47
jrosserbtw unless you find a specific thing changed in 23.3.0 it is unlikley to be something we changed in openstack-ansible20:48
jrosserit is much more likley to be some external factors like new pacakges in UCA which we don't have an appropriate apt pin for early enough in the deployment20:49
spateljrosser this is how it got installed by OSA - https://paste.opendev.org/show/bmbFY9ezTRIgtgpB7rAt/20:52
spatellook like qemu-* packages install that lib20:54
jrosserright20:57
jrosserso i think this is a bug20:57
jrosserwhich has probably been there for a long time20:58
jrosserwould be interesting to compare with the apt log on one of your previous hosts20:58
spateli can check that also wait.. 20:58
jrosserwhat happens is that in our CI job for ceph the ceph OSD is setup before nova, so the apt pins are placed on the *same node* as what will be the compute node, early in the deploy20:59
jrosserand then when it comes time to install librbd for qemu-*, it's already present from the ceph-osd setup, and meets the apt-policy becasue the pins were already placed21:00
jrosseri have the apt log from CI here and we can see it has installed librbd to the correct version before qemu https://zuul.opendev.org/t/openstack/build/f7abcc2628b34f779a9e55f7467c2eed/log/logs/host/apt/history.log.txt#56-6021:02
jrosseroh wait, thats also an upgrade..... `Upgrade: librbd1:amd64 (15.2.16-0ubuntu0.20.04.1, 16.2.7-0ubuntu0.21.04.1~cloud2)`21:03
jrosserhmmm21:03
spatelThis is my old box and that package comes from difference place - https://paste.opendev.org/show/ba8RbMsY53RTPr1vyPAn/21:05
jrossercan you show what apt-policy says there?21:06
spatelhttps://paste.opendev.org/show/b6808PTwTBSOLV1u1yyq/21:09
spatelThis is old box 21:09
spatelhttps://paste.opendev.org/show/bdYKdTELLejqADWyHhwz/21:10
spatelThis is new box21:10
*** dviroel is now known as dviroel|out21:11
spatelhere is the librbd comparison - https://paste.opendev.org/show/bmdFvO44YmuX4yn41Y4r/21:12
spatelI have to leave now but i will keep poking and try to report bug if its real issue. 21:12
jrosserit is odd that those packages have 21.04 in the name21:12
jrosseri have started a bug, i'll drop the link here for you21:13
spatelThank you! 21:13
spatelI may go offline and won't see link but anyway will followup with you tomorrow :)21:13
spatelthank you for helping out 21:13
jrosserhttps://bugs.launchpad.net/openstack-ansible/+bug/197564321:16
jrosseri really do wonder about the 21.04 in the name, thats just not feeling right21:16
spatel+121:17
NeilHanlon👀https://www.reddit.com/r/Python/comments/uwhzkj/i_think_the_ctx_package_on_pypi_has_been_hacked/22:49

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!