Friday, 2020-11-13

*** tinwood has quit IRC02:10
*** tinwood has joined #openstack-release02:13
*** armax has quit IRC03:47
*** rm_work has quit IRC04:58
*** rm_work has joined #openstack-release04:58
*** evrardjp has quit IRC05:33
*** evrardjp has joined #openstack-release05:33
*** tosky has joined #openstack-release06:24
*** elod has joined #openstack-release06:45
*** rpittau|afk is now known as rpittau07:20
*** slaweq has joined #openstack-release07:25
*** tkajinam has quit IRC08:08
*** sboyron has joined #openstack-release08:11
*** tkajinam has joined #openstack-release08:12
hberaudfungi: o/ FYI last evening we encoutered 3 failing failures with Indeed this patches released masakari is stable branches (stein, train, ussuri), and the upload to pypi failed for the three jobs (retry limit), I can see this error in the three logs "No package matching 'python-libvirt' is available" and after that many tasks are skipped as they are09:03
hberaudconditional tasks.09:03
ttxhmm could be a mirror issue... probably better to hold on further releases until investigated09:04
hberaudI agree09:04
ttxah hm09:05
ttxI think it's linked to the move to focal. Focal has python3-libvirt09:06
hberaudI was wondering if it could be a side effect of our move to focal09:06
hberaudyeah agree09:07
ttxno python-libvirt09:07
hberaudbut I can't find bindep in masakari's repo09:07
ttxI'm a bit unclear why we do install the bindeps from those projects those09:07
hberaudIf I correctly understand python-libvirt is pulled from bindep, isn't?09:08
hberaud(in this case)09:08
hberaudah necessarily I wasn't looking in the right repo :)09:09
ttxstill unclear why it would pull python-libvirt09:09
hberaudno idea09:09
hberaudI agree it's a bit weird09:09
ttxdoes not seem to come from libvirt-dev09:10
ttxstable branches still have it09:11
hberaudso I think we just have to put a new line that correspond to focal09:12
hberaudbut we also need to ensure that other repo doesn't have this kind of pkg that need to be pulled09:12
hberaudah good catch09:13
ttxand it's a bit tricky to find as codesearch.o.o shows master branch09:13
hberaudI need to dig on beagle to see if we can pass a branch, but I think it's similar to codesearch.o.o as it is based on it09:14
ttxYeah so we should probably hold on all stable releases09:15
ttxvictoria and wallaby should be fine as they were migrated to focal09:15
ttxothers we should probably doublecheck their bindep before approving09:15
hberaudyep I think we should only hold stein, train, ussuri09:15
hberaudso AFAIK beagle doesn't accept a branch param09:16
ttxand we should find a way to inspect those stable bindeps and catch any other issue09:16
hberaudit could be: 1. find projects that use bindep 2. clone them in a batch mode with a script 3. search on stein, train, ussuri for similar issue09:17
hberaudit could be easy to code that09:17
hberaudI think that 98% of projects that use bindep on master already had bindep on stable branches09:19
hberaudand beagle could be used in 1 to generate the input list for 209:19
ttx!/story/2007865 has all the tasks that gmann pushed to fix it in master, but I can't seem to be able to load it as it's too busy09:20
hberaudI'll spend on it today09:20
hberaudonce their bindep will be fixed I suppose we should to bump to a new version, these failing versions will be lost for forever, isn't?09:23
ttxTL;DR for people joining now: we have issues with now-focal release jobs on stable releases as they install old bindeps that may not be focal-compliant (example: masakari asking for python-libvirt instead of python3-libvirt in ussuri). We should hold on stable releases in ussuri and before09:23
ttxhberaud: yes09:23
* hberaud afk09:24
ttxhberaud: i wonder if it's doable to fix the stable bindeps actually -- those branches are not using focal yet, so it might break them. Let's wait for fungi, see if he has suggestions09:32
hberaudwithin 1 hour I'll will send an email related to masakari's failure, to inform the project team about this09:40
*** whoami-rajat__ has quit IRC09:48
*** brinzhang_ has quit IRC10:16
*** icey has quit IRC10:30
*** icey has joined #openstack-release10:36
*** e0ne has joined #openstack-release11:35
openstackgerritRadosław Piliszek proposed openstack/releases master: Re-release Masakari Monitors
openstackgerritRadosław Piliszek proposed openstack/releases master: [masakari] Transition Stein to EM
fungii half expected this, but wanted to see if it actually became a problem.... do we really need to install packages from bindep.txt just to build sdists/wheels?13:53
fungiif so, zuul can now theoretically match tags to branches so we should be able to have branch matchers run point releases for older series on older nodesets13:54
fungibut unless folks think we're likely to need additional distro packages just to make dists, then it might be easier to run those jobs without the bindep roles (or no-op/short-circuiting those roles somehow, maybe by specifying /dev/null as the package list)13:56
hberaudAFAIK I don't think we need these packages (from bindep)14:02
*** rpittau is now known as rpittau|afk14:49
gmannhberaud: ttx stable branch before victoria run on Bionic (and xenial on <=rocky) so it should find it14:53
hberaudgmann: but publish-openstack-artifacts was moved to focal few days ago =>
gmannhberaud: we should run on Focal only for victoria onwards via brnahc variant like we do for all the common jobs used for stable and master14:57
gmannfungi: ^^14:57
fungigmann: yes, but these jobs are triggered by tags, not by changes on branches. tags aren't inherently tied to any particular branch15:03
fungizuul recently grew a feature to try and guess a relevant branch for a tag, but we haven't really tried it yet and it's never going to be perfect because a tag can point to an object which appears in multiple branches or even an object which appears on no branches15:04
fungiwe can give it a shot and see how it works out, but we need newer versions of some tools (recently the wheel package) because older versions don't necessarily work well with uploading things to pypi15:06
fungithere are some things we can't really freeze when we make a stable branch, and the python packaging ecosystem is one of them15:07
gmannfungi:  ohk, I am not sure if running on Focal can bring up more issue than masakari one only.15:07
fungirunning on bionic brought up an issue, which is why we updated the ubuntu version for the release jobs15:08
fungibut if the only issue from running on bionic is that we're trying to tell bindep to install old package names no longer available on focal, then it's probable we can just fix that by not using the bindep role. we shouldn't have to install project-specific distro packages just to create an sdist tarball or a wheel from them15:10
gmannyeah that make sense. devstack does not install bindep package by default like for lib etc15:12
gmanndoing for release publish jobs will be good until asked by specific inherited job15:13
hberaudnotice that the root cause of all these things are the fact that some repo use a README at the markdown format rather than at the rst format, maybe that moving them to rst could be a straightforward path to follow. It could allow us to restore the `publish-openstack-artifacts` as before, thoughts?15:23
hberaudand it could allow us to keep our envs aligned between the different kinds of jobs15:25
hberaudand also allow us to align/standardize all our readme files15:26
*** ianychoi has joined #openstack-release15:30
ttxwe need t move on to focal anyways, regardless of that README thing15:43
ttxwe should try to run that job without the bindep. i fear it's inherited from deep under though15:44
openstackgerritJakub Libosvar proposed openstack/releases master: Release ovsdbapp 1.7.0
ttxcould be as simple as removing line
ttxwill affect release-openstack-python test-release-openstack and publish-openstack-python-branch-tarball15:51
ttxsounds reasonable, but not to test on a Friday15:52
*** armax has joined #openstack-release15:57
openstackgerritMerged openstack/releases master: [horizon] Transition Stein to EM
* smcginnis caught up16:00
smcginnisWe could test with release-test, but it really will be a case of needing to be cautious with every release for a while to make sure there aren't any project specific gotchas that we aren't aware of.16:01
fungihberaud: if memory serves, the reason some projects use markdown readme files is for consistency with specific outside ecosystems they're used in16:09
smcginnisThat's my understand as well. Ansible Galaxy being one of them -
openstackgerritDaniel Bengtsson proposed openstack/releases master: New bugfix release for oslo.messaging for victoria.
clarkbfungi: re bindep, I think you would need system packages if you have setup requires that build against them16:48
clarkbchnaces of that seem slim, but it is theoretically possible16:48
fungiright, i think a16:49
fungiit's likely all our projects have only setuptools and pbr as setup-requires16:50
clarkbyup extremely unlikely for openstack in particular due to use of pbr16:50
fungiclarkb: i've replied on the ml thread as well, but looking through the logs of a release-openstack-python build i suspect we can drop the bindep and ensure-tox roles from playbooks/pti-python-tarball/pre.yaml17:22
fungiwe don't actually use tox nor install the project being packaged within the job17:23
clarkbmakes sense17:23
fungiwe probably need to keep ensure-pip because that makes sure we have wheel (sort of a side effect)17:23
fungii think those roles were just holdovers from back when we still used tox to invoke setup.py17:24
ttxworst case scenario some corner case job fails17:34
fungii'll propose that cleanup as soon as i knock out the ovh log upload changes. just about caught up17:34
*** e0ne has quit IRC17:36
*** bnemec is now known as beekneemech17:37
hberaudfungi: ack17:50
hberaudttx: +117:51
fungi Clean up pti-python-tarball python-branch-tarball18:14
fungihberaud: ttx: smcginnis: gmann: clarkb: ^18:15
*** lifeless has quit IRC19:31
*** lifeless has joined #openstack-release19:47
*** slaweq has quit IRC20:03
*** sboyron has quit IRC20:59
*** sboyron has joined #openstack-release21:14
*** sboyron has quit IRC22:49
*** armax has quit IRC23:10
*** armax has joined #openstack-release23:13
*** corvus has quit IRC23:22
*** jeblair has joined #openstack-release23:25
*** jeblair is now known as corvus23:26
*** tosky has quit IRC23:43

Generated by 2.17.2 by Marius Gedminas - find it at!