| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2019-10-29 | |||
| 14:23:58 | efried | At least from what I can tell, it's doing the same thing. And you said yourself (in the TODO) that it could be removed after 1.4.0 :P | |
| 14:24:17 | dtroyer | Then it should be good, but consider the author | |
| 14:24:26 | efried | hehe | |
| 15:44:54 | umbSublime | with a connection object is it possible to specify the api microversion. Something equivalent to conn.compute.get('/os-hypervisors?with_servers=true', microversion='2.53') | |
| 15:49:45 | efried | umbSublime: Yes, all the methods (including the REST primitives like .get(), .put(), etc.) should take a microversion kwarg. Let me double check how it's spelled... | |
| 15:51:53 | umbSublime | efried: sorry I wasn't really clear, my example works | |
| 15:52:16 | efried | oh, you want to specify the default microversion for the connection so you don't have to put in the kwarg every time? | |
| 15:52:19 | umbSublime | My question was more, is this possible with for example conn.compute.hypervisors() | |
| 15:55:43 | efried | umbSublime: I was wrong, the non-primitives don't take a microversion. You can set the default microversion on the proxy though. | |
| 15:56:39 | umbSublime | Is that something I setup when creating the connection | |
| 15:56:57 | efried | Sorry, conn.compute is the proxy | |
| 15:57:05 | efried | conn.compute.default_microversion = XXX | |
| 15:57:05 | efried | So it should definitely work like: | |
| 15:57:22 | efried | But there's probably a prettier way to do it, via config or something. | |
| 15:57:24 | umbSublime | Ahh I see, thanks | |
| 15:58:10 | efried | dtantsur: do you have that answer off the top ^ ? | |
| 15:58:13 | umbSublime | and what about the 'with_servers=True', I see it works with .get, but with .hypervisors() it get's ingored. | |
| 15:58:34 | umbSublime | I pass it as a kwarg on .hypervisors() | |
| 15:58:57 | dtantsur | I suspect something like default_compute_microversions | |
| 15:59:16 | dtantsur | however, if you need to set it for e.g. conn.compute.hypervisors, then somebody is doing it wrong: other we or you :) | |
| 15:59:36 | dtantsur | openstacksdk is expected to use the latest microversions it knows. if it doesn't - it's a bug. | |
| 15:59:46 | dtantsur | if it does, then changing the default won't help you | |
| 15:59:57 | efried | dtantsur: Didn't mordred just fix something for that? | |
| 16:00:27 | dtantsur | maybe, I haven't been following closely the last couple of weeks | |
| 16:00:38 | dtantsur | (relocation) | |
| 16:00:39 | umbSublime | "somebody is doing it wrong" -> most likely me | |
| 16:00:56 | dtantsur | umbSublime: it may help if you describe the problem you're facing | |
| 16:01:16 | efried | https://review.opendev.org/#/c/685999/ implies we weren't getting the compute microversion we were expecting. | |
| 16:01:57 | umbSublime | dtantsur: I have a meeting coming, up but I'll be back with the deets | |
| 16:02:11 | efried | umbSublime: What version of keystoneauth are you using? | |
| 16:04:18 | efried | it's possible that without https://review.opendev.org/#/c/685042/ you're getting 2.1 instead of 'latest'. | |
| 16:04:21 | efried | dtantsur: ^ | |
| 16:05:29 | dtantsur | yeah, this may be the cause | |
| 16:05:43 | dtantsur | these microversions, never liked them :) | |
| 16:06:23 | efried | umbSublime: So you may want to make sure you have ksa 3.18.0 (it looks like the Train fix hasn't merged yet) | |
| 16:06:34 | efried | and see if that fixes the issue | |
| 16:07:37 | efried | umbSublime: As for `with_servers`, that was added in the 2.53 microversion of the compute API, but we might not have brought SDK up to parity yet. | |
| 16:12:50 | efried | yeah. | |
| 16:13:00 | efried | umbSublime: You should propose a patch to add that :P | |
| 16:23:05 | openstackgerrit | Eric Fried proposed openstack/python-openstackclient master: DNM: CI check https://review.opendev.org/691910 | |
| 16:41:46 | openstackgerrit | Merged openstack/openstacksdk master: Add support for Node tainted field https://review.opendev.org/690755 | |
| 17:13:59 | umbSublime | efried: I'm using keystone v3. As for the microversion I get from the proxy how can i check this? i can only see 'version=2' in vars(conn.compute) | |
| 17:14:58 | umbSublime | v3.11 for ksa | |
| 18:40:01 | mriedem | dtroyer: can we get this simple osc change in https://review.opendev.org/#/c/691039/ to unblock novaclient 16.0.0 from being in u-c https://review.opendev.org/#/c/690097/ | |
| 20:38:01 | efried | mriedem: something changed between then and now, osc isn't building: https://review.opendev.org/#/c/691910/ | |
| 20:38:11 | efried | so your patch won't clear the gate, is my point. | |
| 20:38:47 | mriedem | but not just mine, any you're saying | |
| 20:38:52 | efried | And btw, I have no idea what's going on there. I can't repro "locally" | |
| 20:39:04 | efried | yes, noticed it on one of mine, seemed unrelated, so I put that up to confirm. | |
| 20:39:19 | efried | mebbe you can do your logstash thing and find the culprit? | |
| 20:40:28 | mriedem | my guess is some py2/3 thing was dropped/released | |
| 20:42:00 | mriedem | http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22TypeError%3A%20Unicode-objects%20must%20be%20encoded%20before%20hashing%5C%22%20AND%20tags%3A%5C%22console%5C%22&from=7d isn't super useful | |
| 20:43:38 | efried | I guess it wouldn't be, since nobody's been building anything in osc except for me. | |
| 20:43:51 | mriedem | well, could be openstack-requirements for a dependency | |
| 20:43:55 | mriedem | but i'm not seeing anything obvious | |
| 20:44:03 | efried | but it shows that whatever it is only affects python-openstackclient. | |
| 20:44:13 | mriedem | this hashlib library that's blowing up here https://github.com/openstack/python-openstackclient/blob/4bde9af89251431791fc8d69fe09d5e17a8fba8f/openstackclient/tests/functional/compute/v2/test_agent.py#L24 hasn't changed in forever on pyi | |
| 20:44:15 | mriedem | *pypi | |
| 20:44:40 | mriedem | https://zuul.opendev.org/t/openstack/build/49ff6a5645554095b3fb51bce497aff5/log/tox/functional-0.log hmm is that in the standard lib now? | |
| 20:45:01 | mriedem | this job must use py36? | |
| 20:45:04 | efried | did that job recently cut over to py3? | |
| 20:45:14 | mriedem | that's my guess | |
| 20:45:40 | efried | well, its incarnation in python-openstackclient didn't, so unless it has an ancestor that did... | |
| 20:45:46 | mriedem | https://docs.python.org/3.6/library/hashlib.html#hashlib.hash.update | |
| 20:46:01 | mriedem | that's entirely possible, parent projects are moving to py3 in ussuri | |
| 20:46:16 | efried | devstack-tox-functional | |
| 20:47:27 | mriedem | seems ok https://opendev.org/openstack/devstack/src/branch/master/.zuul.yaml#L566 | |
| 20:48:03 | mriedem | parent: multinode | |
| 20:49:46 | efried | ...which is abstract, whatever that means. | |
| 20:49:56 | mriedem | means it just has to be extended i think | |
| 20:50:10 | efried | well, yeah | |
| 20:50:33 | efried | btw, attempting to run simply tox -e functional on my devstack node deadlocks. | |
| 20:50:46 | mriedem | i don't see anything recent in openstack-zuul-jobs that looks related | |
| 20:50:49 | mriedem | or devstack | |
| 20:51:03 | efried | yeah | |
| 20:51:35 | mriedem | the job does have USE_PYTHON3=False in it | |
| 20:51:38 | mriedem | so it should be using py2 | |
| 20:51:52 | efried | but isn't? | |
| 20:52:54 | mriedem | looks like it's using py3 https://zuul.opendev.org/t/openstack/build/49ff6a5645554095b3fb51bce497aff5/log/job-output.txt#30632 | |
| 20:53:10 | mriedem | $ /usr/bin/python3 -m virtualenv --no-download --python /usr/bin/python3 functional | |
| 20:53:21 | efried | I'll push a fix. | |
| 20:53:35 | efried | ...without for the moment caring that we "should be running py2". Because we really shouldn't. | |
| 20:54:18 | mriedem | i think you just need to put USE_PYTHON3: True in here https://github.com/openstack/python-openstackclient/blob/master/.zuul.yaml#L76 | |
| 20:54:45 | efried | but the problem is that it *is* using py3 and md5().update() needs an encoded string. | |
| 20:55:54 | mriedem | well, a quick fix would be setting basepython = python2 in the tox.ini functional env i think if the job isn't using py3, | |
| 20:56:01 | openstackgerrit | Eric Fried proposed openstack/python-openstackclient master: Fix functional tests for py3 https://review.opendev.org/691980 | |
| 20:56:02 | mriedem | but if the parent is now using py3, yeah idk what's going on there that changed | |
| 20:56:23 | efried | mriedem: That ought to fix at least that error, we'll see if something else crops up. | |
| 20:57:57 | efried | dtroyer: FYI: TL;DR: python-openstackclient builds broke some time between 10/24 and this morning when the functional jobs somehow got cut over to py3. Fixing via https://review.opendev.org/691980 | |
| 20:59:40 | efried | actually, I can narrow it down further. I have a successful build Oct 28 19:31 (that's US Central time) | |
| 20:59:51 | mordred | efried: how did we manage to break builds? | |
| 21:00:02 | efried | mordred: haven't figured that out yet. | |
| 21:00:18 | efried | As far as mriedem and I can tell, "something" cut over to py3 | |
| 21:00:29 | mordred | "awesome" | |
| 21:00:53 | efried | If this build is running py2, that'll be confirmed: https://zuul.opendev.org/t/openstack/build/ad5f0ee04fed4285b9f5429debd9b32d/log/job-output.txt | |
| 21:01:17 | openstackgerrit | Pierre Prinetti proposed openstack/python-openstackclient master: versions: Fix 'versions show' help message https://review.opendev.org/686408 | |
| 21:02:33 | efried | I'm not good at definitively determining whether a job is running under py2 or py3, but afaict that one is py2, and today's broken ones are py3 | |
| 21:02:35 | mordred | efried: that one seems to be py2 - I see a bunch of false for use python3 - and an uninstall of python3 package | |
| 21:02:46 | efried | yup | |
| 21:02:47 | efried | so | |