| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2017-01-11 | |||
| 15:16:29 | stevemar | i am negative on that | |
| 15:16:37 | stevemar | untested (mostly) | |
| 15:17:15 | dtroyer | roger that… agreed | |
| 15:17:20 | stevemar | dtroyer: want to add a functional test job to osc-lib before releasing? | |
| 15:17:29 | stevemar | or are you comfortable enough to release? | |
| 15:17:37 | stevemar | i think the changes were mostly additive | |
| 15:17:44 | dtroyer | depends on how long it takes to get it added :) | |
| 15:18:25 | dtroyer | I made those compat-changes, the case I am concerned about we get in the dsvm jobs (using env vars for auth) | |
| 15:18:42 | dtroyer | rather, using having the comat bits in osc-lib and also in osc | |
| 15:19:08 | dtroyer | in one case it's doing double work and is not (should not be) harmful | |
| 15:19:27 | stevemar | dtroyer: looks like only one thing is failing with sdk 0.9.12 -- http://logs.openstack.org/50/418650/4/check/gate-osc-dsvm-functional-ubuntu-xenial/abbb8d4/testr_results.html.gz ? | |
| 15:20:03 | dtroyer | woo! | |
| 15:20:15 | dtroyer | qos rule type... | |
| 15:20:37 | dtroyer | is anyone workign on that (I just skimmed the overnight scrollback) | |
| 15:20:52 | dtroyer | I'll jump on it after adding the func tests and releasing osc-lib | |
| 15:20:53 | stevemar | dtroyer: for the functional job in osc-lib... it should test with master osc-lib and master osc? | |
| 15:21:24 | dtroyer | we need that combination somewhere, but I was thinking master osc-lib with released osc | |
| 15:21:42 | dtroyer | we've found that dsvm jobs are insufficient for validating we don't break anything | |
| 15:21:47 | stevemar | right | |
| 15:22:28 | stevemar | dtroyer: i did the combination before but i don't recall how to do it, may take a few hours | |
| 15:22:55 | dtroyer | release osc + master osc-lib would be the default, right? | |
| 15:23:05 | dtroyer | ie, no extra work | |
| 15:29:08 | stevemar | i think so, but we'll need to run the functional tests *in* the osc repo, so i'm not sure how that works :\ | |
| 15:29:34 | dtroyer | ugh, right | |
| 15:31:04 | stevemar | dtroyer: and we'll need to probably run the master version of those tests? | |
| 15:31:26 | stevemar | dtroyer: you know, we could have another job in the OSC gate that tests osc-lib master and osc master | |
| 15:31:39 | stevemar | instead of osc-lib release and osc master | |
| 15:31:49 | stevemar | that'll be *really* easy | |
| 15:31:53 | stevemar | i think | |
| 15:32:11 | dtroyer | lets do it that way… in gate, check or experimental so it can go on-demand? | |
| 15:32:15 | stevemar | just copy+paste the old job and force it to install osc-lib in master | |
| 15:32:28 | stevemar | non-voting check for now i think | |
| 15:33:28 | dtroyer | kk | |
| 15:37:28 | stevemar | dtroyer: aside from copying https://github.com/openstack/python-openstackclient/blob/master/openstackclient/tests/functional/post_test_hook.sh and the 'functional' job | |
| 15:37:33 | stevemar | sorry | |
| 15:37:52 | stevemar | dtroyer: let me tell you in steps what i was going to do, see if you can think of a simpler way | |
| 15:38:00 | dtroyer | ok | |
| 15:38:25 | stevemar | 1) copy the functional job here, but call a different post_test_hook.sh, maybe post_test_osc_lib_hook.sh ? https://github.com/openstack-infra/project-config/blob/e172a5e4107f1774a77e212d82218086ed39329e/jenkins/jobs/osc.yaml#L2 | |
| 15:38:58 | stevemar | 2) copy https://github.com/openstack/python-openstackclient/blob/master/openstackclient/tests/functional/post_test_hook.sh to the new name and have it clone and install osc-lib master | |
| 15:39:09 | stevemar | 3) make money | |
| 15:40:14 | stevemar | i could get fancy and maybe set an ENV variable somewhere that says we're using osc-lib from master, but ... i know this will work | |
| 15:41:28 | dtroyer | I don't know the job stuff well enough to know the shortcuts, your plan sounds like it will work | |
| 15:41:38 | dtroyer | should we commit the shell script first? | |
| 15:41:43 | dtroyer | also, stable jobs? | |
| 15:43:42 | stevemar | the shell script will be in osc no? | |
| 15:43:53 | dtroyer | yes | |
| 15:44:05 | stevemar | oh i thought you were asking about getting that in before osc-lib is released | |
| 15:44:07 | stevemar | ignore me | |
| 15:44:15 | dtroyer | should we also do sdk? can we do both in the same job? | |
| 15:44:21 | dtroyer | so many questions! | |
| 15:45:16 | dtroyer | a) I was thinking we'd want the new *-hook.sh merged before defining the job that calls it | |
| 15:45:43 | dtroyer | b) I'm still thinking through how this affects (or if it should?) stable jobs | |
| 15:46:09 | dtroyer | c) we should also do the sdk this way, can we do both in the same job in a sane manner? | |
| 15:47:32 | stevemar | dtroyer: i think we can | |
| 15:49:58 | dtroyer | o-c-c too? | |
| 15:50:05 | dtroyer | how much can we add??? :) | |
| 15:50:15 | dtroyer | those are the three primary break-risk libs for OSC | |
| 16:27:41 | dtroyer | stevemar: I just got the same results in my local OSC functional tests (same devstack-related failures) for both current and master osc-lib. Are you OK with pushing the release now or do you still want to wait intil the new job is in? | |
| 16:31:39 | stevemar | dtroyer: i'm OK with pushing | |
| 16:32:11 | stevemar | dtroyer: i'm trying to figure out the pip command to install the sdk | |
| 16:32:24 | briancurtin | pip install openstacksdk | |
| 16:32:35 | stevemar | hehe | |
| 16:32:50 | stevemar | briancurtin: from master instead? | |
| 16:32:58 | briancurtin | what? | |
| 16:33:08 | stevemar | i was thinking ... pip install -q -U -e "git+https://git.openstack.org/python-openstacksdk/openstacksdk.git#egg=openstacksdk" | |
| 16:33:34 | briancurtin | that probably works. or maybe egg=openstack. i can’t remember exactly what goes in the egg part | |
| 16:33:58 | stevemar | briancurtin: thats where i'm stumbling :) | |
| 16:34:20 | briancurtin | the PyPI package name is openstacksdk but the package that python itself deals with is called openstack, so maybe egg=openstack | |
| 16:35:12 | dtroyer | looks like egg=python_openstackclient is used for that case, so pypi package name? | |
| 16:35:24 | dtroyer | with '-' -> '_' mapping | |
| 16:36:06 | stevemar | dtroyer: sorta, pip install -q -U -e "git+https://git.openstack.org/openstack/osc-lib.git#egg=osc_lib" works for osc-lib | |
| 16:36:37 | dtroyer | so I'd expect #egg=puthon_openstacksdk to be the magic | |
| 16:37:10 | dtroyer | s/pu/py/ | |
| 16:38:53 | dtroyer | pip install -q -U -e "git+https://git.openstack.org/openstack/python-openstacksdk.git#egg=python_openstacksdk" | |
| 16:38:59 | dtroyer | appears to have worked for me | |
| 16:40:21 | stevemar | *smh* i had an extra openstack in there | |
| 16:41:14 | reedip_outofmemo | briancurtin : ping | |
| 16:41:22 | briancurtin | reedip_outofmemo: what’s up | |
| 16:42:12 | reedip_outofmemo | briancurtin : hi , busy ? | |
| 16:42:20 | openstackgerrit | Steve Martinelli proposed openstack/python-openstackclient: add support for running tests with master dependencies https://review.openstack.org/419032 | |
| 16:42:23 | briancurtin | reedip_outofmemo: no, what’s up | |
| 16:42:26 | stevemar | dtroyer: ^ | |
| 16:42:30 | reedip_outofmemo | briancurtin : it was related to https://review.openstack.org/#/c/418729/ | |
| 16:42:42 | briancurtin | yeah, what is the question? | |
| 16:43:13 | reedip_outofmemo | briancurtin : earlier, quota show for openstackclient ( before 0.9.12 version ) was working smoothly | |
| 16:43:29 | reedip_outofmemo | with 0.9.12, there has been a change and quota show now fails | |
| 16:43:39 | openstack | Launchpad bug 1655537 in python-openstackclient "Cannot see Network quota using "openstack quota show"" [Undecided,New] - Assigned to Reedip (reedip-banerjee) | |
| 16:43:39 | reedip_outofmemo | briancurtin :https://bugs.launchpad.net/python-openstackclient/+bug/1655537 | |
| 16:44:23 | reedip_outofmemo | briancurtin: the reason is though the Compute and Volume quotas are returned as dict, network quota is returned as an object | |
| 16:44:23 | briancurtin | reedip_outofmemo: there was no change in 0.9.12, i think it was in 0.9.11, but this is an OSC problem and not an SDK problem | |
| 16:44:48 | reedip_outofmemo | briancurtin : ok , one question | |
| 16:45:03 | reedip_outofmemo | briancurtin : is the compute client and volume client part of Openstack SDK ? | |
| 16:45:09 | briancurtin | reedip_outofmemo: openstacksdk hasn’t explicitly returned dictionaries ever, though some time ago it did return objects that can act like dictionaries. compute hasn’t done that for probably 8 months | |
| 16:45:11 | reedip_outofmemo | briancurtin : like Network client | |
| 16:45:32 | briancurtin | reedip_outofmemo: I’m not sure, i don’t work on OSC. i think OSC only uses openstacksdk for the network related commands though, not for compute and volume | |
| 16:46:04 | reedip_outofmemo | stevemar : any suggestions related to above query ? | |
| 16:46:17 | dtroyer | reedip: briancurtin is correct re compute and volume quotas | |
| 16:46:41 | reedip_outofmemo | dtroyer : ok, so sdk is used only for Networks?????? | |
| 16:46:54 | dtroyer | the root problem is that OSC was not ready for the 0.9.11 changes in the SDK because our functioanl tests were producing false positives for a time | |
| 16:47:00 | dtroyer | reedip: yes | |