| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-03-05 | |||
| 11:33:10 | kashyap | lyarwood: Unless I'm terribly blind, I see that you've added about the SATA bus aspect, but I can't find anything about the PCIe root ports. :) But no prob | |
| 11:34:30 | stephenfin | lyarwood: The patch to add '--block-device' to 'openstack server create' landed and I'm adding functional tests. It seems boot_index is not mandatory | |
| 11:35:19 | stephenfin | If I omit it, I guess it defaults to -1. The guest correctly boots from the image I'm specifying alongside (cirros) | |
| 11:36:01 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP zuul: Replace nova-grenade-multinode with grenade-multinode https://review.opendev.org/c/openstack/nova/+/778885 | |
| 11:36:51 | lyarwood | stephenfin: are you providing both --image and --block-device? | |
| 11:37:07 | stephenfin | yes | |
| 11:37:13 | lyarwood | stephenfin: I think osc and novaclient fudge things in the past to always boot from the --image you provided | |
| 11:37:42 | lyarwood | stephenfin: in terms of the api behaviour | |
| 11:38:02 | lyarwood | stephenfin: I think n-api also ensures we boot from the image, let me check | |
| 11:39:57 | stephenfin | There's also additional weirdness. I was copying this behaviour https://github.com/openstack/python-novaclient/blob/master/novaclient/v2/servers.py#L775-L779 | |
| 11:40:30 | stephenfin | However, that seems broken. nova rejects those requests with 'Block Device Mapping is Invalid: Mapping image to local is not supported.' | |
| 11:41:22 | stephenfin | openstack server create ... --image cirros-0.5.1-x86_64-disk --block-device device_name=vdb,source_type=image,destination_type=volume,uuid=$UUID ... | |
| 11:41:47 | gibi | stephenfin: ooh that image to local error was one of my early nova contribution ;) I0c38134e2aacca83d98ed0b341f15c480afe14f1 | |
| 11:42:03 | stephenfin | that works just fine, so I'm going to change it to default destination type to volume when source type is image instead, since that seems more logical/correct | |
| 11:42:42 | stephenfin | gibi: we can't be blamed for the mistakes of our younger selves ;) | |
| 11:43:03 | gibi | indeed | |
| 11:43:57 | stephenfin | ah, wait, I think I misunderstood that code | |
| 11:44:14 | lyarwood | right | |
| 11:44:25 | lyarwood | image to local is valid | |
| 11:44:33 | lyarwood | I'm not sure what we end up doing to hit that | |
| 11:44:45 | lyarwood | do we provide an imageRef *and* image->local bdm? | |
| 11:45:00 | stephenfin | I think that's transforming the '--image' request to a bdm dict | |
| 11:45:20 | lyarwood | yeah it should do and I think that needs to drop the imageRef as a result? | |
| 11:46:18 | lyarwood | with imageRef we end up with an image based local bdm so I'm not sure why we would block this in the API | |
| 11:46:28 | lyarwood | maybe because it skips imageRef | |
| 11:46:42 | stephenfin | http://paste.openstack.org/show/803266/ | |
| 11:46:55 | stephenfin | that's what the OSC code currently ends up generating | |
| 11:47:24 | stephenfin | in response to http://paste.openstack.org/show/803267/ | |
| 11:47:49 | lyarwood | stephenfin: and n-api rejects that? | |
| 11:47:55 | stephenfin | nope, that's fine | |
| 11:48:00 | lyarwood | oh | |
| 11:48:03 | lyarwood | okay | |
| 11:48:06 | stephenfin | if I add destination_type=local to the '--block-device' arg though | |
| 11:48:10 | stephenfin | the API rejects _that_ | |
| 11:48:27 | lyarwood | oh right | |
| 11:48:29 | lyarwood | yeah | |
| 11:48:39 | lyarwood | there can only be one | |
| 11:48:42 | lyarwood | I think | |
| 11:49:29 | stephenfin | Yeah, quite possibly | |
| 11:49:33 | stephenfin | I'm also using the same image twice | |
| 11:49:41 | stephenfin | i don't know if that's an issue or not | |
| 11:49:41 | stephenfin | http://paste.openstack.org/show/803268/ | |
| 11:49:45 | lyarwood | https://github.com/openstack/nova/blob/c8a6f8d2e502e139c2212fb20cc0f89de816aeb1/nova/block_device.py#L193-L206 | |
| 11:49:56 | stephenfin | plus other differences as you can see | |
| 11:51:06 | stephenfin | okay, so the logic should be... | |
| 11:51:33 | gibi | as far as I remember I blocked image to local mapping as it was not implemented (for other than for a single image) | |
| 11:51:36 | stephenfin | if --image is specified, use destination_type=volume since this is an additional image | |
| 11:52:10 | lyarwood | destination_type shouldn't be optional right? | |
| 11:52:32 | lyarwood | huh it is | |
| 11:52:37 | lyarwood | I thought it wasn't in the API | |
| 11:52:46 | lyarwood | so yeah default to volume if --image is provided | |
| 11:52:52 | lyarwood | and destination_type isn't | |
| 11:52:56 | stephenfin | they're all optional except UUID | |
| 11:53:12 | lyarwood | and reject the use of image as a destination_type if --image is already provided in the cli | |
| 11:53:20 | stephenfin | of local you mean | |
| 11:53:25 | lyarwood | yeah sorry | |
| 11:53:27 | lyarwood | local | |
| 11:53:35 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Add device event constants to fakelibvirt https://review.opendev.org/c/openstack/nova/+/778887 | |
| 11:53:45 | stephenfin | --image + --block-device source_type=image,destination_type=local ==> FAIL | |
| 11:54:02 | lyarwood | I can add a note in the api docs about this | |
| 11:54:11 | stephenfin | actually, for OSC I might just leave this to the server | |
| 11:54:16 | stephenfin | as you suggested in the review | |
| 11:54:39 | lyarwood | ah damn past me | |
| 11:54:42 | stephenfin | simply using a sensible destination_type (volume) would be good | |
| 11:54:46 | lyarwood | but that makes sense | |
| 11:54:46 | stephenfin | enough | |
| 11:54:51 | lyarwood | yeah | |
| 12:26:50 | openstackgerrit | Lee Yarwood proposed openstack/nova master: zuul: Replace grenade and nova-grenade-multinode with grenade-multinode https://review.opendev.org/c/openstack/nova/+/778885 | |
| 12:27:19 | lyarwood | ^ tosky / gmann ; I'm not sure what you both think about this approach but it could allow us to finish our move to zuulv3 before M3 | |
| 12:27:56 | lyarwood | tl;dr - move everything to grenade-multinode and add coverage for ceph later once that multinode base job is finished | |
| 12:55:07 | sean-k-mooney | lyarwood: that wont actully remove grenade unless we remove it form the integrated-gate-compute template | |
| 12:55:42 | sean-k-mooney | it wil still be added by integrated-gate-compute and it will no longer use the irrelevant-files list | |
| 12:56:02 | sean-k-mooney | so what you really need to do is also remove the integrated-gate-compute template | |
| 12:56:51 | sean-k-mooney | oh never mind | |
| 12:57:04 | sean-k-mooney | https://github.com/openstack/tempest/blob/master/zuul.d/integrated-gate.yaml#L442-L454 the patch to remvoe grenade form that has been merged | |
| 13:05:19 | gmann | yeah that is already done | |
| 13:05:32 | gmann | lyarwood: ack, thanks, I will check during weekend or on Monday | |
| 13:06:28 | lyarwood | sean-k-mooney: it's already removed from that template | |
| 13:06:35 | lyarwood | sean-k-mooney: yeah | |
| 13:07:16 | tosky | lyarwood: that's entirely up to what you (nova) think it's better | |
| 13:08:28 | lyarwood | I'd rather avoid duplication and move to zuulv3 now instead of waiting to get the ceph coverage in place | |
| 13:09:47 | openstackgerrit | Lee Yarwood proposed openstack/nova master: zuul: Remove nova-dsvm-multinode-base https://review.opendev.org/c/openstack/nova/+/778908 | |
| 13:09:48 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP zuul: Create and use nova-libvirt-irrelevant-files https://review.opendev.org/c/openstack/nova/+/778909 | |
| 13:11:53 | stephenfin | lyarwood: Low priority, but you may like 'Either a URI-style path (\'file:\\\\{path}\') to a JSON file ' | |
| 13:11:55 | stephenfin | whoops | |
| 13:12:01 | stephenfin | https://review.opendev.org/c/openstack/python-openstackclient/+/778910 | |
| 13:13:12 | lyarwood | stephenfin: nice idea | |
| 13:13:35 | stephenfin | I stole it https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/block-device-mapping-concepts.html | |
| 13:13:47 | stephenfin | (search for --block-device-mappings) | |
| 13:18:51 | sean-k-mooney | speaking of osc https://review.opendev.org/c/openstack/python-openstackclient/+/778571 | |
| 13:18:54 | openstackgerrit | Lee Yarwood proposed openstack/nova master: zuul: Create and use nova-libvirt-irrelevant-files https://review.opendev.org/c/openstack/nova/+/778909 | |
| 13:18:56 | sean-k-mooney | that is for vdpa | |
| 13:19:27 | sean-k-mooney | stephenfin: care to take a look i dont know if i need a release note for that or anything else | |
| 13:19:40 | sean-k-mooney | stephenfin: i just did then minium i needed to not have to use curl | |
| 13:20:30 | stephenfin | sean-k-mooney: will do | |
| 13:20:45 | stephenfin | that's merged in neutron now? | |
| 13:20:53 | sean-k-mooney | yep | |
| 13:20:56 | stephenfin | cool | |
| 13:21:06 | stephenfin | will review fully after lunch but a small release note would be A+ | |
| 13:21:21 | stephenfin | and a link to the neutron change that added it (for context) | |