Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-08-19
14:56:24 sean-k-mooney stephenfin: one example of something that is broke is we do not update the neutron port mac address for pf migration
14:56:52 stephenfin Cool. I can assert that kind of stuff in the test
14:56:54 sean-k-mooney that is broken for both cold and live migration
14:57:15 sean-k-mooney detach on livem migation with PFs is also broken
14:57:39 stephenfin can you attach PFs via neutron?
14:57:47 sean-k-mooney not after the vm is booted
14:57:49 sean-k-mooney but on boot yes
14:57:57 sean-k-mooney vnic_type=direct-phsyical
14:58:28 sean-k-mooney no it was rammed through as a feature without accounting form a bunch of basic edgecease
14:58:44 sean-k-mooney and broke backwards compatiablei of the pci alias
14:59:06 stephenfin sounds like CPU pinning and PCI in general
15:01:40 sean-k-mooney not really
15:14:38 iurygregory sean-k-mooney, hey I did the changes directly to the node (in setup time) and at least all non uefi/multinode jobs are green =)
15:23:05 sean-k-mooney cool
16:05:32 stephenfin gmann: suuuuper dumb question. I'm attempting to enable policy by default in all tests but seeing policy failures https://review.opendev.org/#/c/745906/
16:06:13 stephenfin gmann: some of the rules are using PROJECT_MEMBER_OR_SYSTEM_ADMIN but I'm seeing requests using the non-admin user rejected
16:06:35 stephenfin I assume I need to toggle that "enable new system-scoped policy" config option to get this working?
16:08:16 gmann stephenfin: with PROJECT_MEMBER_OR_SYSTEM_ADMIN check_str we have is to validate the 'system:all' for SYSTEM_ADMIN part and projectid&&member role for PROJECT_MEMBER so non admin would not pass if it is different project
16:08:33 gmann stephenfin: are you removing the old defaults also? if not then it should keep working
16:08:37 gmann let me check the patch
16:08:53 stephenfin gmann: okay, that's what I was expecting but it doesn't happen
16:09:19 gmann if you enable enforce_scope then it will start failing. as it need scope checks also
16:09:19 stephenfin if you could have a look, I'd be grateful :) I think it's a good idea (enabling policy by default)
16:10:19 gmann stephenfin: yeah, that is really good idea to run tests with actual policy. though we might need to do lot of work on test side though
16:10:23 gmann checking..
16:10:38 stephenfin not as much as you'd fear
16:13:34 gmann stephenfin: if you try to do with admin (super access role) then yes no much changes but tests with real policy will be to use reader, member, admin roles. GET APIs by reader role not admin and so on
16:17:18 stephenfin yeah, to be fair, I've only tested with legacy policy config and I've been blasting through things with admin credentials where necessary /o\
16:25:39 openstackgerrit Lee Yarwood proposed openstack/nova master: libvirt: Fix indentation in test__video_model_supported https://review.opendev.org/746979
16:25:39 openstackgerrit Lee Yarwood proposed openstack/nova master: libvirt: Add libvirt version mocks in test_private_destroy_* tests https://review.opendev.org/746980
16:25:40 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Bump MIN_{LIBVIRT,QEMU}_VERSION and NEXT_MIN_{LIBVIRT,QEMU}_VERSION https://review.opendev.org/746981
16:25:40 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_LIBVIRT_FILE_BACKED_DISCARD_VERSION https://review.opendev.org/746982
16:25:41 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_{LIBVIRT,QEMU}_NATIVE_TLS_VERSION https://review.opendev.org/746983
16:25:41 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_LIBVIRT_BETTER_SIGKILL_HANDLING https://review.opendev.org/746984
16:25:42 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_LIBVIRT_VIDEO_MODEL_VERSIONS https://review.opendev.org/746985
16:25:42 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_{LIBVIRT,QEMU}_PMEM_SUPPORT https://review.opendev.org/746986
16:28:06 gmann stephenfin: ok. for legacy policy, that is ok.
16:33:25 gmann stephenfin: those failing test missing ADMIN_API=True, i checked admin_action failed tests. because by default it is non admin API request, sample tests need to set ADMIN_API for admin APIs - https://github.com/openstack/nova/blob/cff7382fb3de4eef7aeddc34b6ae3409cbe50ea0/nova/tests/functional/integrated_helpers.py#L1036
16:33:54 stephenfin cool, that's what I've done in most cases
16:35:40 gmann and for GET APIs where default is admin_or_owner they are failing for missing admin fields in response which confirms this and no other policy issue
16:38:11 stephenfin so I'm going to run all the servers tests as admin to get those extra fields
16:38:23 stephenfin but anything else, I'll just drop the admin-only fields from the response
16:38:43 gmann stephenfin: from func sample tests?
16:38:52 stephenfin yup
16:39:26 gmann i think we should not, functional sample tests has complete response verification and we use that in api-ref also.
16:40:04 gmann making servers tests as admin is ok i think
16:40:36 stephenfin Sure, for this like 'GET /servers/{id}'
16:40:39 gmann and other tests also as admin if it involve admin APi
16:41:00 stephenfin what about things like the rescue action?
16:41:35 stephenfin that's currently being tested as a non-admin user but the response we've documented in an admin-type response
16:42:01 stephenfin I'm thinking those admin fields aren't important in that instance and we should keep showing a non-admin response
16:42:21 stephenfin but maybe not
16:43:04 gmann rescue does not add admin-only fields. you mean rebuild?
16:43:15 stephenfin Sorry, I mean in the response
16:43:38 stephenfin actually, never mind
16:43:46 gmann yeah in response rescue return empty dict or admin_password
16:44:15 stephenfin this is what I was referring to https://github.com/openstack/nova/blob/master/nova/tests/functional/api_sample_tests/test_rescue.py#L94
16:44:23 stephenfin but that's only for internal validation purposes
16:44:36 stephenfin that sample response doesn't need admin-fields in it
16:46:03 gmann stephenfin: ohk i got now. yes we can keep that with non-admin. if sample is not complete then yes keep testing that with non-admin.
16:47:51 gmann stephenfin: even such test can just check state from response explicitly instead of comparing it with sample file response.
17:19:13 openstackgerrit Stephen Finucane proposed openstack/nova master: functional: Move single-use function to its caller https://review.opendev.org/745904
17:19:13 openstackgerrit Stephen Finucane proposed openstack/nova master: functional: Drop '_api' suffix from placement fixture https://review.opendev.org/745905
17:19:14 openstackgerrit Stephen Finucane proposed openstack/nova master: functional: Enable real policy fixture by default https://review.opendev.org/745906
17:19:14 openstackgerrit Stephen Finucane proposed openstack/nova master: functional: Add and use 'GlanceFixture' https://review.opendev.org/745907
17:19:15 openstackgerrit Stephen Finucane proposed openstack/nova master: tests: Remove '_FakeImageService' https://review.opendev.org/745908
17:19:27 stephenfin gmann: ^
17:19:49 stephenfin I might have missed one or two tests (there were 109 failures, after all) but that should be pretty correct now :)
17:21:08 gmann stephenfin: ok, waiting for gate result
17:31:50 stephenfin dansmith: melwitt: Can you folks look at https://review.opendev.org/#/c/744958/ again today?
17:32:14 openstackgerrit Stephen Finucane proposed openstack/nova master: compute: Add type hints for resize functions https://review.opendev.org/745341
17:32:14 openstackgerrit Stephen Finucane proposed openstack/nova master: WIP: compute: Add more type hints for resize functions https://review.opendev.org/745675
17:55:25 artom stephenfin, lyarwood, did any of you start backporting Stephen's func tests refactors to train? Or did we want to socialize that a bit first?
17:56:51 artom Nothing that I can find in gerrit, at any rate. Shall I start?
17:58:52 lyarwood artom: I've not had a chance to get back to it so feel free to start
17:59:45 lyarwood elod: https://review.opendev.org/#/c/746529/ - that reminds me can you take a look at this series from artom when you have time please
18:01:13 artom lyarwood, yep, thanks for poking elod for me
18:01:13 dansmith artom: I think stephenfin did what you asked here, so can you throw your +1 back on it? https://review.opendev.org/#/c/744950/4
18:01:21 dansmith er, s/back//
18:06:45 lyarwood melwitt: https://review.opendev.org/#/c/745512/ - trivial test cleanup if you have time today btw
18:20:39 openstackgerrit Lee Yarwood proposed openstack/nova master: rbd: Move rbd_utils out of libvirt driver under nova.storage https://review.opendev.org/746904
18:20:39 openstackgerrit Lee Yarwood proposed openstack/nova master: Add ability to download Glance images into the libvirt image cache via RBD https://review.opendev.org/574301
18:23:42 melwitt lyarwood: took too long to find what I was looking for, but commented on your patch that there's a fixture for doing that
18:25:52 dansmith so that would only work with eventlet,
18:25:53 dansmith which seems like an odd thing to favor
18:26:08 yankcrime hey folks, i think i'm hitting this bug with an instance that was created circa rocky, but now a user has stopped and is attempting to start it again for this first time under ussuri: https://bugs.launchpad.net/nova/+bug/1864020
18:26:08 openstack Launchpad bug 1864020 in OpenStack Compute (nova) "libvirt.libvirtError: Requested operation is not valid: format of backing image %s of image %s was not specified in the image metadata (See https://libvirt.org/kbase/backing_chains.html for troubleshooting)" [Undecided,Fix committed] - Assigned to Lee Yarwood (lyarwood)
18:26:34 yankcrime if i'm reading right, it's related to libvirt 6.0 and that fix only applies to new instances that are created with this vesion of libvirt
18:30:34 lyarwood yankcrime: yeah I thought libvirt then relaxed this in a later release?
18:31:27 melwitt dansmith: yeah I guess loopingcall uses eventlet for its looping https://github.com/openstack/oslo.service/blob/master/oslo_service/loopingcall.py
18:32:56 lyarwood yankcrime: sorry need to drop for the day now but if this isn't resolved by a later release of libvirt then you might need to convert the disks in place to include the format details of the backing file
18:33:05 dansmith yeah that seems weird to me -- if we were to convert some of our stuff to real threads as we've discussed before, that'd be a problem
18:34:16 yankcrime lyarwood: yeah that's the conclusion i'm coming to, wanted to make sure i wasn't missing anything
18:34:37 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_LIBVIRT_FILE_BACKED_DISCARD_VERSION https://review.opendev.org/746982
18:34:38 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_{LIBVIRT,QEMU}_NATIVE_TLS_VERSION https://review.opendev.org/746983
18:34:38 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_LIBVIRT_BETTER_SIGKILL_HANDLING https://review.opendev.org/746984
18:34:39 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_LIBVIRT_VIDEO_MODEL_VERSIONS https://review.opendev.org/746985
18:34:39 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Remove MIN_{LIBVIRT,QEMU}_PMEM_SUPPORT https://review.opendev.org/746986
18:34:58 lyarwood yankcrime: yup sorry, I was sure the libvirt folks backed down after we complained about this but I can't seem to find the change
18:35:14 lyarwood yankcrime: kashyap is the person to speak to about this, I'll follow up with him in the morning

Earlier   Later