| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-06-20 | |||
| 17:16:57 | stephenfin | Right, but other projects use sphinx for things other than building docs | |
| 17:17:06 | stephenfin | for example, oslo_config.sphinxext | |
| 17:17:07 | sean-k-mooney1 | theat project supprot building docs via tox | |
| 17:17:26 | stephenfin | For those projects, we still need to support Python2 | |
| 17:17:35 | sean-k-mooney1 | why | |
| 17:17:54 | sean-k-mooney1 | they can move to running that under python3 | |
| 17:18:12 | sean-k-mooney1 | they need to do that this cycle anyway | |
| 17:18:41 | stephenfin | They can't drop py27 tests | |
| 17:19:29 | stephenfin | and if the project specifies sphinx as a requirement (vs. a test-requirement), it's going to pull in sphinx | |
| 17:19:30 | sean-k-mooney1 | well i guess not until U | |
| 17:19:36 | bnemec | It's kind of irrelevant. The requirements job isn't smart enough to differentiate, so you need the cap whether you support python 2 or not. | |
| 17:19:43 | sean-k-mooney1 | stephenfin: sure | |
| 17:19:53 | sean-k-mooney1 | but upper contratits caps it | |
| 17:19:57 | stephenfin | so attempting to install something like oslo.config on Python 2.7 would no longer be possible | |
| 17:20:00 | sean-k-mooney1 | so it should not be an issue | |
| 17:20:18 | sean-k-mooney1 | it is if you use upper-constratints | |
| 17:20:26 | sean-k-mooney1 | it cap sphix at 1.8.5 | |
| 17:20:37 | stephenfin | I don't know the specifics, but it's not as simple as that unfortunately | |
| 17:20:42 | stephenfin | See what bnemec said | |
| 17:20:51 | stephenfin | The requirements gate job will fail until we fix this | |
| 17:21:05 | sean-k-mooney1 | then we should wait to fix the gate | |
| 17:21:29 | stephenfin | Perhaps. I never looked into it, tbh | |
| 17:21:48 | stephenfin | Merging these patches seemed easier, particularly given we can drop them again in < 4 (3?) months | |
| 17:21:51 | sean-k-mooney1 | locally master works and im about to push a test patch to verify the gate | |
| 17:21:59 | bnemec | That's a discussion for the requirements team. | |
| 17:22:10 | bnemec | I know they're aware that this is bad, but I don't think they have a solution yet. | |
| 17:22:22 | openstackgerrit | sean mooney proposed openstack/os-vif master: testing docs job https://review.opendev.org/666651 | |
| 17:22:47 | sean-k-mooney1 | bnemec: well if i build with tox on master or with ^ its works | |
| 17:23:02 | bnemec | That won't demonstrate the problem. Make a change to a requirements file. | |
| 17:23:18 | bnemec | Without Stephen's patch it will fail the requirements job. | |
| 17:23:33 | openstackgerrit | Stephen Finucane proposed openstack/nova master: docs: Rework all things metadata'y https://review.opendev.org/640730 | |
| 17:23:34 | sean-k-mooney1 | then the requirement job is broken | |
| 17:23:34 | openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Rename 'configuration drive' to 'config drive' https://review.opendev.org/641723 | |
| 17:23:37 | sean-k-mooney1 | not the repo | |
| 17:23:49 | stephenfin | sean-k-mooney1: That's kind of what we're saying | |
| 17:23:50 | stephenfin | but | |
| 17:23:56 | bnemec | Again, that's an argument to have with the requirements team. :-) | |
| 17:23:57 | stephenfin | <bnemec> I know they're aware that this is bad, but I don't think they have a solution yet. | |
| 17:24:07 | sean-k-mooney1 | i have noticed and reported that the job does not handel the python version stuff properly | |
| 17:25:04 | sean-k-mooney1 | we spoke about this at the ptg when we wer fixing lower constraits | |
| 17:25:54 | stephenfin | sean-k-mooney1: If you fancy looking into this, by all means. However, if you want to save yourself a few hours of mostly thankless work, this patch and others like it are a good band-aid for this cycle :) | |
| 17:25:59 | stephenfin | and then, come next cycle | |
| 17:26:04 | stephenfin | Boom. Bye bye Python 2.7 | |
| 17:26:07 | openstackgerrit | sean mooney proposed openstack/os-vif master: testing docs job https://review.opendev.org/666651 | |
| 17:26:29 | aspiers | ffs devstack why do you hate me so? | |
| 17:26:43 | aspiers | good riddance to 2.7 | |
| 17:26:46 | stephenfin | aspiers: Don't worry, it hates everyone | |
| 17:26:56 | aspiers | it hates me more | |
| 17:27:06 | aspiers | it's even refusing to download stuff from files.pythonhosted.org! | |
| 17:27:07 | stephenfin | Good riddance to all manner of encoding issues | |
| 17:27:18 | aspiers | some SSL Certificate verification failure bullshit | |
| 17:27:37 | aspiers | SSLError: HTTPSConnectionPool(host='files.pythonhosted.org', port=443): Max retries exceeded with url: /packages/09/f1/93603a26daf7a993a0acbbcfd32afce8b2fdf30a765d5651571ab635969b/etcd3-0.10.0.tar.gz (Caused by SSLError(SSLError(1, u'[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed (_ssl.c:661)'),)) | |
| 17:27:47 | aspiers | no matter that it used to work fine | |
| 17:28:08 | aspiers | pip install of anything fails | |
| 17:29:53 | stephenfin | I've seen that before but only against internal stuff | |
| 17:29:54 | sean-k-mooney1 | stephenfin: its a quick fix but its also againt the sprit of https://github.com/openstack/governance/blob/master/reference/pti/python.rst#constraints we opt in to the requiremetn team managing the upper-constiraits i dont like circumventing that | |
| 17:30:37 | sean-k-mooney1 | lets see if the bug is repoduced in the gate first | |
| 17:30:42 | stephenfin | aspiers: The solution internally is setting 'export REQUESTS_CA_BUNDLE=<path to some certs>' but I doubt that's applicable to you | |
| 17:30:46 | sean-k-mooney1 | if so then ill file a bug and we can merge it | |
| 17:31:31 | stephenfin | sean-k-mooney1: Fair. To be clear I think this will happen any time a package we use drops Python 2 support | |
| 17:31:40 | stephenfin | Again, until U | |
| 17:32:17 | sean-k-mooney1 | if that is the cases i really dont want to have to manage this in every repo and we should fix the requiremetns job | |
| 17:32:25 | aspiers | stephenfin: openssl s_client -connect files.pythonhosted.org:443 -prexit | |
| 17:32:29 | aspiers | gives this: | |
| 17:32:33 | aspiers | Verify return code: 9 (certificate is not yet valid) | |
| 17:33:18 | aspiers | cya tomorrow o/ | |
| 17:33:36 | efried | jroll: see latest response before you start hacking at the bp template | |
| 17:33:43 | efried | s/bp/spec/ | |
| 17:47:44 | jroll | efried: you severely underestimate my work queue if you think I start on anything within the hour :P | |
| 17:48:00 | jroll | oh, that's good news, thanks | |
| 17:48:10 | efried | yahyoubetcha | |
| 17:49:17 | efried | jroll: justification for no spec: no API change, no upgrade impact, no RPC/OVO changes, no new config, etc | |
| 17:49:40 | efried | mriedem may feel differently | |
| 17:49:57 | jroll | efried: yeah that's fair. I don't mind small specs anyway, so I'm open to either direction | |
| 17:50:22 | efried | heh, except nova-specs will bust you if you don't do the template just right. | |
| 17:50:26 | efried | but sure. | |
| 17:50:40 | sean-k-mooney1 | stephenfin: so http://logs.openstack.org/51/666651/2/check/requirements-check/c8c82c5/job-output.txt.gz#_2019-06-20_17_39_08_900979 is the same issue i spoke to the requirements folks about comming up to RC2 back like 3 months ago | |
| 17:51:14 | jroll | efried: I'm used to nit-picky CI :P | |
| 17:51:40 | sean-k-mooney1 | i hit it when trying to fix nova lower contratits job and found i could not actully use phytion version markers unless they matched upper constritns exactly | |
| 17:53:37 | sean-k-mooney1 | im pretty sure the problem is this | |
| 17:53:39 | sean-k-mooney1 | https://github.com/openstack/requirements/blob/master/global-requirements.txt#L459-L460 | |
| 17:53:54 | sean-k-mooney1 | we should not be listing the max versions there | |
| 17:54:11 | sean-k-mooney1 | that is what is causeign the conflict | |
| 17:55:25 | sean-k-mooney1 | so https://github.com/openstack/requirements/commit/00b2bcf7d664b1526b4eefe157c33113206d6251 is likely the reason all the other porject are broken | |
| 17:55:42 | sean-k-mooney1 | we should have caped just in upper constraitns not in global requiremetns | |
| 17:56:52 | mriedem | sean-k-mooney1: dansmith: it is done https://review.opendev.org/#/c/644881/37 | |
| 17:57:56 | mriedem | efried: jroll: passthrough tpm seems small spec worthy to me, but i can also pass the buck to dansmith | |
| 17:58:08 | sean-k-mooney1 | mriedem: thanks and thanks for rechecking too | |
| 17:58:11 | mriedem | we've had specs for less complicated things i'm sure | |
| 17:58:23 | dansmith | yeah | |
| 18:09:19 | mriedem | there, i have provided at least a few things that would likely need to be mulled over in a spec | |
| 18:09:44 | mriedem | creating a server with a thing and deleting it is easy, moving the damn thing and what happens when the compute is down and we need to cleanup later is the pain in the ass | |
| 18:14:03 | mriedem | stephenfin: remind me why you think it's better to list every option in the group here https://review.opendev.org/#/c/640730/8/doc/source/admin/vendordata.rst@134 rather than just the group itself and let the operator read through the config option docs? | |
| 18:15:39 | mriedem | previously you said, "While this service happens to have most things nicely grouped, that's not always the case. I realize duplication isn't nice but there's value in having everything grouped in one place. I'd like to keep this, if that's okay" but what is that value? and does it outweigh the duplication? | |
| 18:21:01 | mriedem | sean-k-mooney1: john is totally cool with https://review.opendev.org/#/c/647733/ | |
| 18:21:15 | mriedem | the +1 bot must not have -W filtering | |
| 18:22:11 | sean-k-mooney1 | haha ya... | |
| 18:22:53 | sean-k-mooney1 | i will get to adding the missing test before i sign off today | |
| 18:29:06 | efried | jroll: oh well, I tried | |
| 18:48:27 | openstackgerrit | Merged openstack/nova master: Change InstanceFaultRollback handling in _error_out_instance_on_exception https://review.opendev.org/633212 | |