Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-09-13
17:21:27 efried mikal Long-term, if nothing else, once we bring snapshot support in-tree, we would be able to make a case to put the whole open-and-stream method into a nova.privsep.powervm module in nova proper.
17:22:42 mikal efried: its ok, I've given up on the dream of replacing temporary_chown with privsep, as I can't find a way to make that work
17:22:55 mikal efried: I updated the patch in nova to say that thing
17:23:12 efried mikal Saw that -- we can't have a privseppy context manager?
17:54:48 openstackgerrit Merged openstack/nova master: doc: fix flavor notes https://review.openstack.org/502112
17:55:31 openstackgerrit Merged openstack/nova master: Enable test_iscsi_volume in live migration job https://review.openstack.org/459316
18:51:55 openstackgerrit Balazs Gibizer proposed openstack/nova master: Moving more utils to ServerResourceAllocationTestBase https://review.openstack.org/499539
18:51:55 openstackgerrit Balazs Gibizer proposed openstack/nova master: factor out compute service start in ServerMovingTest https://review.openstack.org/503037
18:51:56 openstackgerrit Balazs Gibizer proposed openstack/nova master: Test resource allocation during soft delete https://review.openstack.org/495159
19:03:22 kfarr gibi here is the patch I mentioned: https://review.openstack.org/#/c/439855/
19:03:25 kfarr Thank you for taking a look!
19:24:00 gibi kfarr: thanks, I will check
19:47:13 melwitt so, it looks like the ability to run unit tests with a regex is broken recently? I'm doing 'tox -epy27 test_blah' and it's running osprofiler first and then running all tests after
19:51:30 cdent melwitt: does ‘tox -epy27 — test_blah’ work
19:51:47 cdent if so it may be the thing that mtreinish just posted to os-dev about
19:51:57 melwitt cdent: 'tox -epy27 -- --regex test_blah' doesn't work. lemme try your suggestion
19:53:11 melwitt that runs everything too
19:53:43 mtreinish melwitt: hmm I think there is a bug in the nova tox.ini
19:53:58 mtreinish melwitt: try dropping the single quotes around posargs on the py27 job
19:54:06 mtreinish and run 'tox -epy27 -- --regex test_blah'
19:55:07 mtreinish that's probably my fault for not catching the quotes in https://review.openstack.org/#/c/501842/
19:55:26 cdent hmm, I don’t have this problem, maybe it is related to version of tox?
19:55:49 melwitt oh, hm. let me make sure I have the latest version
19:56:01 mtreinish cdent: does your tox venv have os-testr 1.0.0?
19:56:39 cdent maybe not; slow network has limited my refreshes of late
20:00:39 cdent melwitt, mtreinish: okay, yeah, with a refreshed venv I get the same bad behavior
20:01:21 bauzas FWIW, I don't usually pass args this way when calling the specific py27 tox target
20:01:35 bauzas I'd rather do "tox -epy27 <my_pattern>"
20:01:37 melwitt cdent, mtreinish: yep, that was it. removed the single quotes from tox.ini and then use 'tox -epy27 -- --regex test_blah' works
20:02:06 bauzas oh, a regex, then nvm
20:02:08 melwitt bauzas: yeah, same. but mtreinish says in the commit message of that change that that was apparently working by accident
20:03:53 bauzas lemme look tox.ini
20:04:10 bauzas posargs are treated differently depending on the target
20:07:29 melwitt mtreinish: does the new version of ostestr also do the stdout capture differently? it used to be in a separate section for stdout and now it's just going inline with everything else?
20:07:56 bauzas we quote the posargs for ostestr actually
20:08:34 bauzas so that means that if you pass something like "-- --regex foo", it'll litterally call out ostestr '-- --regex foo'
20:13:09 mriedem mikal: https://review.openstack.org/#/c/503164/
20:13:11 mriedem come on
20:14:14 mikal mriedem: I replied just now
20:32:14 mtreinish bauzas: yeah we need to set the env variables in the tox.ini
20:32:24 mtreinish the capture fixtures depend on an env var being set
20:32:34 mtreinish and they were done magically by default in .testr.conf
20:32:39 mtreinish which isn't used anymore
20:32:44 bauzas hah!
20:32:59 mtreinish personally I just drop the env variable checks in the base test class because no one ever wants to turn it off
20:34:42 mtreinish bauzas, melwitt: I'll push a patch up to fix the things, one sec
20:42:13 melwitt mtreinish: ah, thanks. didn't realize the stdout env var stuff was being done in .testr.conf
20:42:27 mtreinish melwitt: heh, yeah no one does
20:42:28 openstackgerrit Matthew Treinish proposed openstack/nova master: Fix test runner config issues with os-testr 1.0.0 https://review.openstack.org/503811
20:42:33 mtreinish melwitt, cdent, mikal: ^^^
20:43:55 mtreinish you probably want to test that locally (I just pushed it up)
20:44:06 melwitt mtreinish: yep trying it out now
20:46:31 mtreinish bauzas: ^^^ oops forgot about you :)
20:46:56 bauzas heh, np
20:47:07 bauzas waiting for melwitt's test :p
20:47:33 melwitt mtreinish: cool, works for py27 env. are the other posargs with single quotes in tox.ini okay the way they are?
20:47:50 bauzas I guess so
20:47:57 melwitt like what about the functional env
20:48:04 mtreinish I wasn't sure how the pretty_tox.sh script does things
20:48:20 melwitt yeah, I was thinking the same. the functional env is a direct ostestr call tho
20:48:22 mtreinish with quotes it treats posargs as a single arg which breaks for something like --regex foo
20:48:34 mtreinish which needs to be 2 args for argparse to understand it
20:49:07 mtreinish hmm, I'm not sure if that fixes the capture fixtures though. I saw things outside the capture in a local run
20:49:30 melwitt it worked for what I'm working on (wherein I added a couple of print statements)
20:49:53 mtreinish melwitt: ah ok, maybe it was just something else then
20:49:54 melwitt gonna try running -efunctional with a regex to see if that is okay
20:49:59 mtreinish ok
20:50:09 dansmith mikal: this might be a good set for your special list: https://review.openstack.org/#/c/502152
20:50:24 dansmith mikal: the unregister.* ones
20:52:53 mikal dansmith: I live to obey
20:54:00 openstackgerrit Merged openstack/nova master: Move execs of tee to privsep. https://review.openstack.org/489438
20:57:26 mriedem dtantsur|afk: TheJulia: are we going nova/ironic in our room? (Ballroom A)
20:57:31 mriedem s/going/doing/
20:59:19 dansmith cdent: why do you need to move the custom namespace thing in this patch? https://review.openstack.org/#/c/502155/3
20:59:30 dansmith oh sorry, I can read
20:59:34 cdent import loop
20:59:51 cdent i hadn’t done it, and then the tests blew up and trying to fix it uncovered that
21:00:52 melwitt mtreinish: I commented on the review
21:01:38 mtreinish melwitt: yeah, I saw that I'm pushing a new rev to answer mikal's complaints and fix the quotes on functional tests
21:02:15 melwitt mtreinish: oh cool. also on api-samples I would think
21:03:03 mtreinish yep
21:03:32 openstackgerrit Matthew Treinish proposed openstack/nova master: Fix test runner config issues with os-testr 1.0.0 https://review.openstack.org/503811
21:03:49 mtreinish melwitt: ^^^ I think that should work, I accidently did tox -r so the tests are still waiting
21:03:53 mtreinish for me locally
21:04:08 mtreinish mikal: ^^^ test that and see if you still want to stab me
21:05:58 melwitt stestr: error: invalid choice: 'nova.tests.unit.test_profiler' (choose from 'run', 'list', 'slowest', 'failing', 'last', 'init', 'load')
21:10:21 openstackgerrit Eric Fried proposed openstack/nova master: WIP: dac_admin.openfile context manager https://review.openstack.org/503823
21:11:27 mtreinish melwitt: yeah I forgot to add run after stestr
21:11:30 mtreinish one sec
21:12:01 openstackgerrit Lee Yarwood proposed openstack/nova stable/pike: Move hash ring initialization to init_host() for ironic https://review.openstack.org/502082
21:12:21 openstackgerrit Matthew Treinish proposed openstack/nova master: Fix test runner config issues with os-testr 1.0.0 https://review.openstack.org/503811
21:12:27 mtreinish melwitt: ^^^
21:15:59 mtreinish melwitt: it works for me locally
21:16:14 melwitt yeah, working for me too. thanks
21:17:50 mtreinish melwitt: and a bonus with that patch is if you do tox -epy27 -- --no-discover nova.tests.unit.test_utils it will save some time because it will skip discovery
21:18:31 mtreinish well, at least for the second run. the osprofiler thing will still do discovery
21:18:41 mtreinish oh, actually I can add that flag and make that a bit faster
21:18:56 mtreinish I'll do that in a follow up though to land the fix faster
21:19:09 melwitt yeah, it's a bummer to have osprofiler run first for me, bc I usually just ctrl-c that part
21:19:50 bauzas long known rant about osprofiler :)

Earlier   Later