Monday, 2020-01-20

*** tetsuro has joined #openstack-release00:31
*** dave-mccowan has joined #openstack-release00:37
*** dave-mccowan has quit IRC01:45
*** stephenfin has quit IRC01:50
*** stephenfin has joined #openstack-release01:50
*** smcginnis|PTO has quit IRC01:51
*** smcginnis|PTO has joined #openstack-release01:52
*** prometheanfire has quit IRC02:39
*** diablo_rojo has joined #openstack-release02:43
*** prometheanfire has joined #openstack-release03:03
*** udesale has joined #openstack-release04:18
*** udesale_ has joined #openstack-release04:58
*** udesale has quit IRC05:02
*** evrardjp has quit IRC05:34
*** evrardjp has joined #openstack-release05:34
*** soumplis has joined #openstack-release05:44
*** ykarel|away is now known as ykarel06:32
openstackgerritAbhishek Kekane proposed openstack/releases master: Release Glance 19.0.1
*** jbadiapa has joined #openstack-release07:09
*** e0ne has joined #openstack-release07:17
openstackgerritAbhishek Kekane proposed openstack/releases master: Release Glance 18.0.1
*** rpittau|afk is now known as rpittau07:51
*** ykarel is now known as ykarel|lunch07:51
*** diablo_rojo has quit IRC07:55
*** slaweq has joined #openstack-release07:56
*** slaweq has quit IRC08:05
*** jtomasek has joined #openstack-release08:07
*** e0ne has quit IRC08:16
*** tosky has joined #openstack-release08:17
*** slaweq has joined #openstack-release08:23
*** hberaud|gone is now known as hberaud08:44
*** ykarel|lunch is now known as ykarel09:00
*** ricolin has quit IRC09:04
*** ricolin has joined #openstack-release09:15
*** jhesketh has quit IRC09:32
*** jhesketh has joined #openstack-release09:34
*** e0ne has joined #openstack-release09:36
*** hberaud is now known as hberaud|lunch10:30
*** ricolin has quit IRC10:55
*** udesale_ has quit IRC11:02
*** dtantsur|afk is now known as dtantsur11:13
*** rpittau is now known as rpittau|bbl11:20
openstackgerritMerged openstack/releases master: Add openstack-tox-bashate job for scripts
openstackgerritMerged openstack/releases master: Add signature links for independent deliverables
*** ykarel is now known as ykarel|afk11:40
*** hberaud|lunch is now known as hberaud11:48
*** ykarel|afk is now known as ykarel12:24
*** mnaser has quit IRC12:27
*** mnaser has joined #openstack-release12:27
*** slaweq has quit IRC12:48
*** slaweq has joined #openstack-release12:51
*** ykarel is now known as ykarel|afk13:01
*** electrofelix has joined #openstack-release13:01
*** electrofelix has quit IRC13:01
*** smcginnis|PTO is now known as smcginnis13:09
*** rpittau|bbl is now known as rpittau13:11
smcginnisgmann, tonyb: Does EM really make a difference. The tempest issues would still need to be solved. It's not EOL, just EM, so there is still the expectation that tests can be run.13:14
tonyb[m]smcginnis: You're right in that they need to be fixed.  I have a very (very!) vague recollection about tempest and support for releases and EM?13:33
tonyb[m]smcginnis: I'm probably very confused13:33
smcginnistonyb[m]: With EM, we don't do releases anymore. But there is still the expectation that tests can be run against the branch. If tests can't or they fail, that would force our hand to just EOL the branch.13:43
tonybsmcginnis: Yup.  I agree I assumed that gmann was using the EM deadline as the date to have the code merged so that the EM projects will work and not force EOL13:44
openstackgerritTony Breeds proposed openstack/releases master: Add tools/ and get-contacts
openstackgerritTony Breeds proposed openstack/releases master: Use __str__ for printing the contact information
openstackgerritTony Breeds proposed openstack/releases master: Add docs for tools/
*** ykarel|afk is now known as ykarel13:56
*** ricolin has joined #openstack-release13:57
*** ricolin has quit IRC14:07
*** ricolin has joined #openstack-release14:11
gmannsmcginnis: tonyb for EM, we pin the Tempest tag for EM branch testing if master no longer work. py2 drop making the master Tempest not run-able on Xenial node testing which is till Rocky.14:14
* tonyb needs to think on that some14:19
tonyb<- confused14:19
gmanntonyb: on qa policy on pin tempest or py2 drop things.
tonybgmann: Thanks14:21
smcginnisOK, make sense.14:21
smcginnisSounds like we may need to pin even before EM?14:21
gmannwe did for ocata and pike when Tempest master failed to run14:22
*** ricolin_ has joined #openstack-release14:24
*** ricolin_ has quit IRC14:26
*** ricolin has quit IRC14:26
*** ricolin has joined #openstack-release14:26
*** ab-a has joined #openstack-release14:27
*** ab-a has left #openstack-release14:34
*** zxiiro has joined #openstack-release14:36
*** udesale has joined #openstack-release14:42
smcginnisttx: I was striking out last week's info in the tracking etherpad.14:42
smcginnisttx: There was a task for you on there to post the mid-milestone-2 process patch.14:43
smcginnisttx: Was that done? Or should I move that to this week?14:43
openstackgerritMerged openstack/releases master: Release OpenStack-Ansible Rocky
*** ykarel is now known as ykarel|away15:05
*** hberaud is now known as hberaud|school-r15:26
openstackgerritMerged openstack/releases master: Release Glance 18.0.1
*** hberaud|school-r is now known as hberaud15:40
*** mlavalle has joined #openstack-release15:43
*** bauzas has quit IRC15:43
*** diablo_rojo has joined #openstack-release15:58
openstackgerritMerged openstack/releases master: Release Glance 19.0.1
*** udesale has quit IRC16:09
*** openstackgerrit has quit IRC16:13
*** lpetrut has quit IRC16:20
*** KeithMnemonic has joined #openstack-release16:21
ttxsmcginnis: that was done16:21
ttxposted and merged iirc16:22
*** tosky has quit IRC16:23
*** slaweq has quit IRC16:24
*** slaweq has joined #openstack-release16:40
smcginnisttx: Thanks, I couldn't remember.16:42
*** rpittau is now known as rpittau|afk16:44
*** slaweq has quit IRC16:45
*** bauzas has joined #openstack-release16:51
*** ricolin has quit IRC16:55
clarkbrelease team, we appear to be creating universal wheels for python3 only projects17:12
clarkbit appears this may be caused by somethign the release jobs are doing, you may want to hold off on releases until that gets fixed17:13
clarkb(I'm trying to sort out what forces neutron-lib to build a universal wheel despite its config saying not to, and my local builds not building universal wheels eitehr)17:13
clarkb that explicitly sets universal17:15
clarkb is the issue17:16
smcginnisclarkb: I very vaguely remember that being added.17:18
clarkbusing neutron-lib as an example it set the universal=1 flag in setup.cfg until that was no longer true17:19
clarkbbut now we override that package level specification in the job itself17:19
clarkbI think we may want to make the job a bit smarter and have it produce python3 only wheels unless the package sets universal=1 ?17:20
clarkb(the problem with that is we may have to fix project package configs if they don't set universal=1 on stable branches :/)17:20
smcginnisThat seems right. We still want the univeral wheels for stable releases and things like that though.17:20
smcginnisBut I think the flag was added in the job since most of them were not setting any flags in setup.cfg.17:21
clarkbright which means that we get broken packaging out the other end for master17:22
clarkbbasically python3 is here to colelct on our tech debt now17:22
smcginnisIf we don't set it at the job level and some of these older py2 deliverables don't set universal=1, that just makes things a little "less optimal", right?17:23
clarkbsmcginnis: we'll only make python3 wheels for those17:23
clarkbpython2 users would be forced to build their own wheels essentially, which isn't the end of the world but will potentially make installations slightly slower17:24
*** e0ne has quit IRC17:24
smcginnisSo definitely net better to remove that flag from the job.17:24
clarkbanother option is to make the job smarter and have it check for require-python and if that is set to >=3 then drop --universal17:25
clarkband keep --universal by default (but that seems like taking out another loan to pay the current loan)17:25
smcginnisYeah, I'd be concerned about that.17:25
clarkbI guess the other concern with not building python2 wheels is if we have any code that links against C libs you'd have to install headers for them first17:26
clarkbbut I'm 99% sure openstack doesn't build any packages that do that17:27
smcginnisOnly thing I can think of is maybe numpy? Or maybe that's the one I'm remembering just because it's huge.17:27
clarkbsmcginnis: we consume packages that do but we don't build and release any to pypi ourselves17:28
clarkbnumpy is responsible for figuring this out for their packages basically17:28
smcginnisOh, right. Yeah, I doubt any openstack/* deliverables would be an issue.17:28
clarkbso ya I think that would be the downside to removing --universal. python2 consumers would build our packages from sdist which might be a bit slower17:28
smcginnisNone that I've seen at least.17:28
clarkbalso I think --universal isn't compatible with C linking17:28
clarkbbecause python2 and python3 have different ABIs17:29
smcginnisclarkb: Want me to put up a patch to drop that flag?17:29
clarkbfungi: ^ want to sanity check that?17:29
clarkbsmcginnis: ya I think we should then maybe have someone like mordred and fungi and ajaeger sanity check my assertions above17:29
clarkband if packages really really want to be universal they can update their setup.cfg with the universal=1 flag17:29
fungiyes, if we're building any platform-specific wheels they'll have to be separate py2 and py3 wheels anyway afaik17:32
fungibut just about all our projects are pure python17:32
fungibut i agree, we've given projects more than enough time to get their setup.cfg updated to support wheels correctly17:32
fungitime to remove the band-aid17:32
*** dtantsur is now known as dtantsur|afk17:33
*** evrardjp has quit IRC17:34
*** evrardjp has joined #openstack-release17:34
smcginnisPushed up
clarkband this should prevent python2 from trying to use our python3 only wheels which will make python2 happier there17:38
clarkboverall I think this is a good step forward17:38
*** soumplis has quit IRC17:38
*** bnemec has joined #openstack-release17:52
clarkbsmcginnis: fungi: I also wonder if we should do some noop version bump tags to build supercedent wheels17:53
clarkbneutron-lib is the only example i know of on our end suffering from this problem but there may be others17:53
smcginnisclarkb: Can we just pull the wheels for now? Trying to think of what would be the least disruptive.17:54
clarkbsmcginnis: I guess that is an option too. And thinking about it more this is largely a problem for our wheel mirror in CI17:55
smcginnisWe will have new releases for most things soon with milestone 2 coming up anyway.17:55
clarkbsince pypi itself has the requires-python metadata17:55
clarkbwe can probably limp along until new releases happen17:55
smcginnisThat might be best.17:55
smcginnisfungi: I just wrote up something for the ML, but I used my gmail address instead of my gmx. I thought I had subscribed with that one too, but had emails turned off, but I don't see the message coming through. Any way to check if that got caught in the moderation queue?18:02
smcginnisfungi: Ah, nevermind. Now I see it.18:03
*** bnemec has quit IRC18:08
fungimy favorite kind of problem--those that solve themselves18:10
fungiclarkb: yes, i think we should just delete that wheel and any others we spot with the same issue from our cache. i can do that18:11
*** andrein has quit IRC18:11
fungiand add it to the blacklist temporarily until we can start filtering downloaded wheels out of the cache18:11
clarkbfungi: we'll redownload it from pypi. I think we have to remove it from pypi first18:11
clarkbor blacklist it I guess18:11
*** andrein has joined #openstack-release18:11
fungisame way we did setuptools/pip/virtualenv18:11
clarkbya that will work18:12
clarkbfungi: what is the next step on only having wheels that aren't on pypi?18:12
clarkbbecause that would also address the "wheel is on pypi" problem18:12
fungiclarkb: i think i can add my poc to the end of the job, just a sec18:13
fungibut the idea is we set all the `pip wheel` invocations to use an appending log, and then after we loop over all the branches we parse that log to build a list of files which were downloaded and delete them from the wheelhouse18:14
fungithen when the post playbook goes to retrieve them from the build node, the only wheels which shuold be there are the ones we built18:14
clarkbmakes sense18:15
fungiwe can also remove or stub out our blacklist from the copy script once that's working18:15
*** hberaud is now known as hberaud|gone18:20
*** tosky has joined #openstack-release18:50
*** lpetrut has joined #openstack-release18:56
*** slaweq has joined #openstack-release20:11
*** lpetrut has quit IRC21:12
*** slaweq_ has joined #openstack-release21:14
*** slaweq has quit IRC21:15
*** slaweq_ is now known as slaweq21:15
*** mlavalle has quit IRC21:42
*** openstackgerrit has joined #openstack-release22:18
openstackgerritAkihiro Motoki proposed openstack/releases master: Release horizon train 16.1.0
openstackgerritAkihiro Motoki proposed openstack/releases master: Release horizon stein 15.2.0
openstackgerritAkihiro Motoki proposed openstack/releases master: Release horizon rocky 14.1.0
*** jbadiapa has quit IRC22:29
*** slaweq has quit IRC22:36
*** jtomasek has quit IRC22:53
openstackgerritGhanshyam Mann proposed openstack/releases master: Release Tempest 23.0.0
openstackgerritGhanshyam Mann proposed openstack/releases master: Release Patrole 0.8.0
*** tosky has quit IRC23:51

Generated by 2.15.3 by Marius Gedminas - find it at!