Earlier  
Posted Nick Remark
#openstack-nova - 2021-06-04
08:53:12 lyarwood so I think we need to add a dump of the instance console when we fail to detach
08:53:28 lyarwood as this might be an issue within the guestOS but without that it's impossible to say
08:53:41 lyarwood n-cpu and libvirt appear to be doing the correct things at the correct time
08:57:22 kashyap Hm
09:21:50 lyarwood kashyap: https://review.opendev.org/c/openstack/tempest/+/794757 I'll try some runs with this to see what the guest is up to
09:23:43 kashyap lyarwood: Nice; that's quick!
09:56:49 opendevreview Lee Yarwood proposed openstack/nova master: libvirt: Set driver_iommu when attaching virtio devices to SEV instance https://review.opendev.org/c/openstack/nova/+/794639
09:56:50 opendevreview Lee Yarwood proposed openstack/nova master: DNM testing tempest volume detach failure capture of console https://review.opendev.org/c/openstack/nova/+/794766
10:18:58 opendevreview Merged openstack/placement master: Add 'cryptography' package to test-requirements.txt https://review.opendev.org/c/openstack/placement/+/787289
10:20:03 opendevreview Merged openstack/nova master: Fix RequestLevelParams persistence handling in RequestSpec https://review.opendev.org/c/openstack/nova/+/791502
12:40:55 kashyap sean-k-mooney: Hey, I'm trying to add a new video device model ... "bochs" display
12:41:19 kashyap sean-k-mooney: It is recommended by the QEMU folks for UEFI guests as a safe alternative
12:41:59 opendevreview Kashyap Chamarthy proposed openstack/nova master: Allow 'bochs' as a display device option https://review.opendev.org/c/openstack/nova/+/794796
12:42:28 kashyap sean-k-mooney: --^ When you get a minute, let me know what is missing there :)
12:42:54 sean-k-mooney sure
12:43:03 kashyap sean-k-mooney: A blueprint is an overkill for this; shall I just file a bug?
12:43:35 sean-k-mooney no this is a feature
12:43:52 kashyap sean-k-mooney: I think I also need to bump the hash for ImageMetaProps?
12:43:54 sean-k-mooney when i added virtio i had to actully do a spec if i recal
12:44:02 sean-k-mooney you need to do a few thing
12:44:03 kashyap sean-k-mooney: Okay, I'll just file a blueprint.
12:44:12 sean-k-mooney you need to create a new trait in os-traits
12:44:39 sean-k-mooney you will alos need to update i think the notification objects but i have not reviewed the patch yet
12:44:40 kashyap Oh, trait too
12:44:57 kashyap (I haven't added the notification object)
12:45:14 sean-k-mooney https://github.com/openstack/os-traits/blob/master/os_traits/compute/graphics.py#L15
12:45:30 sean-k-mooney os just MODEL_BOCHS
12:45:33 opendevreview Lee Yarwood proposed openstack/nova-specs master: Add attachmentId to responses of the os-volume_attachments API https://review.opendev.org/c/openstack/nova-specs/+/794799
12:46:08 sean-k-mooney kashyap: we have a placement prefilter that allows you to schdule to nodes that have support for the model
12:47:06 kashyap sean-k-mooney: Wouldn't all nodes have support for the model in a given deployment, given that min_libvirt_version satisifes that requirement?
12:47:56 opendevreview Lee Yarwood proposed openstack/nova-specs master: Add attachmentId to responses of the os-volume_attachments API https://review.opendev.org/c/openstack/nova-specs/+/794799
12:49:43 sean-k-mooney oh and technially a change to glance https://review.opendev.org/c/openstack/glance/+/647735 but that is more a nice to have
12:49:56 kashyap sean-k-mooney: Got a link to your virtio-gpu patch?
12:50:04 sean-k-mooney kashyap: this is how i added virtio-gpu https://review.opendev.org/c/openstack/nova/+/647733
12:50:05 kashyap sean-k-mooney: Noted on the Glance part
12:51:41 sean-k-mooney technially i just added all the video moeles that were not already supported when i did that change not just virtio but you could be able to just follow the same steps
12:55:22 kashyap sean-k-mooney: So, here we go, the specless-bp: https://blueprints.launchpad.net/nova/+spec/add-bochs-display-device
12:55:40 sean-k-mooney kashyap: do you need to add a test for uefi fo this model type
12:55:54 sean-k-mooney e.g. hw_fireware_type=uefi
12:56:24 sean-k-mooney we proably should assuming it wont work with a bios boot
12:57:05 kashyap sean-k-mooney: For BIOS boot, "this device might be useable as well" is what the QEMU dev writes
12:57:14 sean-k-mooney kashyap: ack on the blueprint that hsould be more or less enough we can disucss it at the meeting on tueday
12:57:23 kashyap Okido
12:57:36 sean-k-mooney kashyap: ok so we dont need to put a gard in place at the api
12:58:02 sean-k-mooney in valdiate_falvor_image_nostatus
12:58:03 kashyap sean-k-mooney: BTW, a silly question: what are the hashes based on, when bumping the ImageMetaPros?
12:58:45 sean-k-mooney they are computed based in the filed defintions of the the object by olso.versioned.objects
12:58:45 kashyap (When bumping it in unit/test_objects.py, i.e.)
12:59:01 kashyap I see, based on the string length, looks like md5
12:59:17 sean-k-mooney that or sha1 proably
13:01:07 sean-k-mooney they are just used in tests so we can change the hash algoritiom freely provide we update the tests
13:01:22 kashyap sean-k-mooney: md5() for sure :-) --
13:01:25 kashyap fingerprint = '%s-%s' % (obj_class.VERSION, md5(
13:01:31 kashyap (Source: https://github.com/openstack/oslo.versionedobjects/blob/master/oslo_versionedobjects/fixture.py#L274)
13:02:10 sean-k-mooney ah ya it is md5
13:02:34 sean-k-mooney and it has usedforsecurity=Fals
13:02:45 sean-k-mooney so no issue with fips
13:05:01 kashyap sean-k-mooney: BTW, what would be be the full expanded trait here, for video models, when an operator configures it?
13:05:31 sean-k-mooney COMPUTE_GRAPHICS_MODEL_BOCHS
13:05:48 sean-k-mooney its in COMPUTE since its an atribute of what the hypervior can vituralise
13:06:04 sean-k-mooney HW_ is used for descibing atributes of the phsyical hardware
13:06:45 kashyap I see
13:06:57 sean-k-mooney so just add MODEL_BOCHS here https://github.com/openstack/os-traits/blob/master/os_traits/compute/graphics.py#L19
13:07:19 sean-k-mooney like https://github.com/openstack/os-traits/commit/95668fdd59bf526277874ca50c4800ebd3aafb26
13:08:46 kashyap sean-k-mooney: Yep; done. I recall also reworking the layout of CPU-related traits, which was more confusing before :D
13:09:50 sean-k-mooney common x86 ones are here https://github.com/openstack/os-traits/blob/master/os_traits/hw/cpu/x86/__init__.py
13:10:14 sean-k-mooney but we have vendor specific ones in there own module https://github.com/openstack/os-traits/blob/master/os_traits/hw/cpu/x86/intel.py https://github.com/openstack/os-traits/blob/master/os_traits/hw/cpu/x86/amd.py
13:11:29 sean-k-mooney im not sure why https://github.com/openstack/os-traits/blob/master/os_traits/hw/cpu/amd.py is here though
13:11:33 kashyap sean-k-mooney: Right; it was before haphazard; I reworked it in this commit: https://github.com/openstack/os-traits/commit/93d2ff782f
13:11:48 sean-k-mooney unless amd plan to support SEV on non x86
13:12:21 sean-k-mooney oh right it dperecated
13:12:34 kashyap Yep; see the ASCII "diagram" in the commit message :)
13:13:37 sean-k-mooney so to not break the placemnt prefilter you will need to both have a depends on the os-triats patch and we will have to do an os traits release
13:14:00 sean-k-mooney kashyap: the tempest jobs will pull in the os-traits patch correctly
13:14:06 kashyap I see
13:14:08 sean-k-mooney but the upstream tox jobs wont
13:14:19 sean-k-mooney because of how things are installed int eh venv
13:14:31 kashyap sean-k-mooney: So is this the correct config for operator? — trait:COMPUTE_GRAPHICS_MODEL_BOCHS=true
13:14:46 sean-k-mooney no
13:14:55 kashyap Do correct
13:15:03 sean-k-mooney in the image they set hw_video_model=bochs
13:15:23 sean-k-mooney and the prefilter will generate trait:COMPUTE_GRAPHICS_MODEL_BOCHS=reuqired
13:15:28 sean-k-mooney in the placemnet query
13:15:55 sean-k-mooney kashyap: traits never change how an xml is generated directly
13:16:21 kashyap sean-k-mooney: Ah, right; I first wrote "required", and then wrote 'true'
13:16:43 sean-k-mooney adding trait:COMPUTE_GRAPHICS_MODEL_BOCHS=reuqired to the flavor for example woudl jsut ensure it laned on a host with supprot for it but we would still default to cirrus graphics
13:17:18 opendevreview Kashyap Chamarthy proposed openstack/os-traits master: Add a trait for the "bochs" display device https://review.opendev.org/c/openstack/os-traits/+/794807
13:39:22 opendevreview Balazs Gibizer proposed openstack/nova master: Reject server create with extended resource req https://review.opendev.org/c/openstack/nova/+/793619
13:39:25 opendevreview Balazs Gibizer proposed openstack/nova master: Reject server operations with extended resource req https://review.opendev.org/c/openstack/nova/+/793620
13:39:48 kashyap sean-k-mooney: By notification object, you mean this one, right? notification_samples/common_payloads/ImageMetaPropsPayload.json
13:40:14 sean-k-mooney well that the api sample
13:42:52 opendevreview Balazs Gibizer proposed openstack/nova master: Add same_subtree field to RequestLevelParams https://review.opendev.org/c/openstack/nova/+/791503
13:42:53 opendevreview Balazs Gibizer proposed openstack/nova master: Bump min placement microversion to 1.36 https://review.opendev.org/c/openstack/nova/+/791504
13:42:53 sean-k-mooney i mean this https://github.com/openstack/nova/blob/master/nova/notifications/objects/image.py#L22
13:43:13 sean-k-mooney well actully https://github.com/openstack/nova/blob/master/nova/notifications/objects/image.py#L107 proably
13:43:40 sean-k-mooney you will need to bump this to 1.8 https://github.com/openstack/nova/blob/master/nova/notifications/objects/image.py#L127
13:43:47 sean-k-mooney and add the version comment
13:46:55 opendevreview Balazs Gibizer proposed openstack/nova master: Support same_subtree in allocation_canadidate query https://review.opendev.org/c/openstack/nova/+/791505
13:49:05 opendevreview Balazs Gibizer proposed openstack/nova master: Support the new port resource_request format https://review.opendev.org/c/openstack/nova/+/787208
13:51:21 opendevreview Balazs Gibizer proposed openstack/nova master: Transfer RequestLevelParams from ports to scheduling https://review.opendev.org/c/openstack/nova/+/791506

Earlier   Later