| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-10-18 | |||
| 11:17:15 | artom | ygk_12345, you mean you want to see a os.exec('qemu') in the code? You won't find it | |
| 11:17:29 | artom | ygk_12345, that's in the libvirt code itself | |
| 11:17:40 | ygk_12345 | @artom where exactly ? | |
| 11:17:40 | artom | ygk_12345, Nova just defines a domain with some XML | |
| 11:17:55 | artom | ygk_12345, I dunno, use grep, or ask in #libvirt | |
| 11:18:01 | artom | But mostly grep | |
| 11:18:25 | cdent | "But mostly grep" is my mantra | |
| 11:18:43 | ygk_12345 | @artom I have installed libvirt-python and used recursive grep but could not find the qemu-kvm pattern | |
| 11:18:49 | artom | Mostly Grep is my rap name | |
| 11:19:06 | cdent | oh, that's a good one | |
| 11:19:07 | artom | ygk_12345, that's because those are just the Python bindings for libvirt | |
| 11:19:22 | artom | ygk_12345, you want libvirtd | |
| 11:21:21 | artom | We form wht Unix Krew with Oh Snap it's Sed, Aw man it's Awk, and Who Let Perl In? | |
| 11:21:26 | artom | *the Unix | |
| 11:24:31 | ygk_12345 | @artom things at the OS level are weird but interesting | |
| 11:26:09 | cdent | ygk_12345: out of curiosity, why are you looking for the exec call? are you trying to solve a problem, exploring, something else? | |
| 11:27:07 | ygk_12345 | @cdent I am trying to understand and trace the vm create event through nova code :) | |
| 11:27:36 | cdent | I think what artom said is the key bit: nova creates some xml and gives it to libvirt, libvirt does the rest | |
| 11:28:25 | ygk_12345 | @cdent so its libvirt API internally | |
| 11:28:57 | ygk_12345 | @cdent awesome | |
| 11:29:28 | ygk_12345 | @cdent u see no limits to curiosity until we hit the hardware | |
| 11:29:37 | cdent | :) | |
| 11:29:57 | cdent | plenty of room for curiosity on the hardware too | |
| 11:30:45 | ygk_12345 | @cdent appreciate Linus Torvalds for his first version of Linux | |
| 12:17:24 | ygk_12345 | HI all ? | |
| 12:17:26 | ygk_12345 | which wsgi framework is used in nova ? | |
| 12:19:42 | sean-k-mooney | we kind of have our own https://github.com/openstack/nova/blob/master/nova/api/wsgi.py | |
| 12:19:56 | ygk_12345 | ok | |
| 12:20:32 | ygk_12345 | does it have a standalone web server ? | |
| 12:20:46 | sean-k-mooney | it can run in several different modes | |
| 12:21:06 | sean-k-mooney | it can run on the built in python webserver with eventlets | |
| 12:21:27 | sean-k-mooney | or you can run the wsgi service via mod_wsgi or uwsgi | |
| 12:21:58 | ygk_12345 | as of now when we install nova-api, in which mode deso it erun ? | |
| 12:22:04 | ygk_12345 | *does it | |
| 12:22:21 | sean-k-mooney | the nova api console script uses https://github.com/openstack/nova/blob/master/nova/cmd/api.py | |
| 12:22:26 | sean-k-mooney | which uses eventlet | |
| 12:22:27 | sean-k-mooney | but | |
| 12:22:31 | sean-k-mooney | if you use devstack | |
| 12:22:36 | sean-k-mooney | or most installers | |
| 12:22:44 | sean-k-mooney | the nova-api service is not calling that script | |
| 12:22:58 | sean-k-mooney | its running the service under uwsgi i think | |
| 12:23:05 | ygk_12345 | oh | |
| 12:25:18 | sean-k-mooney | ygk_12345: cdent know more about this then i but we moved to testing with extrenal wsgi server implemenation a few release ago | |
| 12:25:43 | ygk_12345 | oh ok | |
| 13:42:28 | KeithMnemonic | mriedem Good day, what is the best was to get some reviews on the patch you submitted for me ? https://review.opendev.org/#/c/683008/ | |
| 13:43:29 | mriedem | KeithMnemonic: umm | |
| 13:43:45 | mriedem | you know we won't release that right? | |
| 13:43:58 | mriedem | since pike is in extended maintenance mode. we can merge stuff to pike but it won't be released. | |
| 13:44:08 | mriedem | so if you really need it you could probably just cherry pick it downstream | |
| 13:46:36 | KeithMnemonic | ok, i can do that, i was hoping it would get merged at least , is that not an option anymore even it if it is not released | |
| 13:47:13 | mriedem | it could be merged upstream but finding people to care about pike at this point upstream is hard, and that patch isn't exactly trivial | |
| 13:47:16 | mriedem | with all of the conflicts | |
| 13:48:16 | dansmith | mriedem: api patch landed, can you kick this in? https://review.opendev.org/#/c/687141/ | |
| 13:48:16 | KeithMnemonic | ok let me try and cherry pick it locally | |
| 13:49:42 | mriedem | dansmith: were you going to reply to andrey again before doing so? | |
| 13:50:16 | mriedem | he is a core so i feel sort of bad if i just overrule him without a reply | |
| 13:51:03 | dansmith | okay | |
| 13:54:16 | mdbooth | Oh, that's evil | |
| 13:54:52 | mdbooth | In python2, if your module is in nova.tests.functional and you 'import fixtures', you get nova.tests.functional.fixtures. In python3 you get 'fixtures'. | |
| 13:55:13 | mdbooth | Apart from renaming nova.tests.functional.fixtures, I'm not immediately sure wtf to do about that. | |
| 13:57:55 | mdbooth | Fun fact 2: emojis in review comments elicit a 500 from gerrit. | |
| 13:58:01 | sean-k-mooney | mdbooth: that to do with relitive imports and what happens if you have a module in the local director that alisas a module in your path | |
| 13:58:23 | mdbooth | sean-k-mooney: Any idea how to disable in py2 in a manner which doesn't break py3? | |
| 13:58:50 | sean-k-mooney | there might be a futre thing for this. absolute import or something like htat | |
| 13:59:21 | melwitt | KeithMnemonic, mriedem: I started reviewing that patch yesterday but not done yet. it's taking forever bc of all the conflicts. I'll finish reviewing today | |
| 13:59:36 | KeithMnemonic | thanks! | |
| 13:59:38 | sean-k-mooney | https://www.python.org/dev/peps/pep-0328/ | |
| 13:59:45 | sean-k-mooney | from __future__ import absolute_import | |
| 14:00:10 | sean-k-mooney | but that shoudl be enabled by default in py 2.7 | |
| 14:00:28 | mdbooth | Apparently not | |
| 14:05:36 | bauzas | mdbooth: to make it clear, nobody should just import any module like "import <module>" but rather doing "from <where> import <module>" | |
| 14:06:08 | bauzas | if I'm reviewing any change like this, I'd -1 it | |
| 14:09:00 | melwitt | mdbooth: there shouldn't be a module nova.tests.functional.fixtures module, there's one in nova.tests.fixtures though | |
| 14:09:11 | melwitt | I think that's how this hasn't come up before | |
| 14:11:38 | efried_afk | mdbooth: you need | |
| 14:11:38 | efried_afk | from __future__ import absolute_import | |
| 14:12:03 | efried | oh, I'm late to the party. | |
| 14:12:25 | efried | if we could just merge the "kill py2" patch, we could stop doing this ick | |
| 14:14:20 | sean-k-mooney | yes | |
| 14:14:35 | sean-k-mooney | although we shoudl still keep it in mind if we backport things | |
| 14:27:28 | melwitt | bauzas: well, if you need to import the base fixtures module (https://pypi.org/project/fixtures/) there's no other way than 'import fixtures', right? | |
| 14:34:55 | mriedem | melwitt: nova.tests.functional.fixtures is where the placement fixture is | |
| 14:35:22 | mriedem | it got moved there after ripping placement out of nova | |
| 14:36:02 | melwitt | oh, I see. I forgot it was just 'fixtures' there | |
| 14:36:09 | melwitt | mah bad | |
| 14:43:44 | openstackgerrit | Eric Fried proposed openstack/nova master: Always trait the compute node RP with COMPUTE_NODE https://review.opendev.org/688979 | |
| 14:43:45 | openstackgerrit | Eric Fried proposed openstack/nova master: ItemsMatcher: mock call list arg in any order https://review.opendev.org/689487 | |
| 14:44:14 | efried | gibi: You'll like ^ :) | |
| 14:52:46 | bauzas | melwitt: sorry was afk, I meant to not import any module by relative | |
| 14:53:34 | bauzas | any import should be absolute | |
| 14:53:46 | bauzas | that's it | |
| 14:53:48 | melwitt | bauzas: np. I know, I'm saying how do you do the "from |
|
| 14:54:15 | bauzas | melwitt: "import fixtures" is absolute for it :) | |
| 14:54:24 | bauzas | I meant for packages | |
| 14:54:44 | bauzas | fixtures is both the package and the module you wanna import | |
| 14:54:53 | bauzas | nova.tests.unit.fixtures isn't | |
| 14:55:15 | bauzas | http://sametmax.com/les-imports-en-python/ | |
| 14:56:16 | melwitt | right, but if your test is in nova/tests/functional/my_test.py and you need 'import fixtures' and there is a nova/tests/functional/fixtures.py it would pick up the latter, we now know | |
| 14:56:50 | melwitt | anyway, just saying I don't see a way to avoid the problem other than using 'from __future__ import absolute_import' as has been mentioned earlier | |
| 14:58:33 | mdbooth | Thanks, all! I'll update the test. | |