| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-08-06 | |||
| 09:09:52 | lyarwood | gibi / bauzas / stephenfin ; ^ okay I think that's a pretty trivial fix to the functional failures, I don't get why it's suddenly an issue but as this RPC call to the conductor then calls down to the source I'm assuming we've added some additional overhead somewhere and were not creeping over the 1 second timeout in the test. | |
| 09:10:10 | lyarwood | are now* | |
| 09:12:58 | gibi | lyarwood: will check soon | |
| 09:42:57 | kashyap | lyarwood: What's the "best" manual way to change the video model for an existing instance? (Asking for a friend.) Is the following painful approach the only way? | |
| 09:43:30 | kashyap | lyarwood: (1) Stop the instance. (2) snapshot it (what kind?). (3) Update the metadata on the snapshot image to use 'virtio' display dev. (4) Then rebuild the instance using the snapshot | |
| 09:43:34 | kashyap | Any better way than that? | |
| 09:43:54 | kashyap | (Besides the future goodness of 'nova-manage' command akin to machine_type update) | |
| 09:44:27 | lyarwood | if you want to persist the instance data then yeah I think that's the only way | |
| 09:44:42 | lyarwood | I don't think resize would change it | |
| 09:45:16 | kashyap | I see. | |
| 09:45:41 | kashyap | And for new instances, obviously start with: openstack image set --property hw_video_model=virtio [...] | |
| 09:46:54 | lyarwood | yeah | |
| 09:48:00 | kashyap | Or: | |
| 09:48:01 | kashyap | $ openstack server create --flavor m1.micro \ | |
| 09:48:01 | kashyap | --image Fedora34 \ | |
| 09:48:01 | kashyap | --property 'hw_video_model=virtio' vm1 | |
| 09:48:05 | kashyap | Thanks! | |
| 09:52:10 | lyarwood | I don't think the second command is valid? | |
| 09:54:53 | lyarwood | oh it's metadata | |
| 09:55:07 | lyarwood | so yeah that isn't going to do anything kashyap | |
| 09:55:53 | kashyap | lyarwood: Oh, I see. So what's the way again to set the prop while creating the server? | |
| 09:56:35 | lyarwood | For new servers you would update the associated image | |
| 09:56:49 | lyarwood | for the existing server case you'd update the snapshot image | |
| 09:57:03 | lyarwood | and then in both create a new server from the image | |
| 09:57:36 | opendevreview | Merged openstack/nova master: compute: Avoid duplicate BDMs during reserve_block_device_name https://review.opendev.org/c/openstack/nova/+/801990 | |
| 09:58:19 | stephenfin | kashyap: Echoing what lyarwood said, we don't expose this kind of knob via flavor extra specs. It's only possible via image metadata properties which rn require a rebuild to change | |
| 09:58:59 | kashyap | stephenfin: lyarwood: Thank you very much, both. I don't have this top off my head. | |
| 09:59:02 | stephenfin | kashyap: you can find a full list of all supported extra specs here https://docs.openstack.org/nova/latest/configuration/extra-specs.html | |
| 09:59:05 | kashyap | I'll correct my mistake on a bugzilla | |
| 10:01:51 | opendevreview | Lee Yarwood proposed openstack/nova stable/wallaby: Add functional test for bug 1937375 https://review.opendev.org/c/openstack/nova/+/803717 | |
| 10:01:52 | opendevreview | Lee Yarwood proposed openstack/nova stable/wallaby: compute: Avoid duplicate BDMs during reserve_block_device_name https://review.opendev.org/c/openstack/nova/+/803718 | |
| 10:02:04 | pjakuszew | kashyap: I wrote a script which modifes extra_specs on existing instances, though I consider it dangerous | |
| 10:02:38 | kashyap | pjakuszew: Heh, understood. The intersting question is - did it work reliably for you, without any nasty side-effects? | |
| 10:03:28 | opendevreview | Lee Yarwood proposed openstack/nova stable/victoria: Add functional test for bug 1937375 https://review.opendev.org/c/openstack/nova/+/803720 | |
| 10:03:29 | opendevreview | Lee Yarwood proposed openstack/nova stable/victoria: compute: Avoid duplicate BDMs during reserve_block_device_name https://review.opendev.org/c/openstack/nova/+/803721 | |
| 10:04:59 | pjakuszew | kashyap: In my case it was a case of adding a cpu_type key to a flavor, then syncing the flavor metadata to instance and reqspec extra_specs - in your case (hw_video_model=virtio) I guess you will have to do a hard reboot so the libvirt xmls will be regenerated | |
| 10:05:32 | kashyap | pjakuszew: Yep; in this case hard-reboot is unavoidable | |
| 10:06:44 | pjakuszew | let me check it, one sec | |
| 10:13:47 | kashyap | pjakuszew: Note, though: I don't have any environment to try this out. I'm just documenting some stuff due to a fallout downstream | |
| 10:13:53 | kashyap | But thank you! :) | |
| 10:15:22 | pjakuszew | kashyap: sure, I'm checking the differences between the xmls with and without properties right now | |
| 10:15:43 | pjakuszew | so I can check if the metadata change was properly propagated after hard reboot later | |
| 10:16:39 | opendevreview | Lee Yarwood proposed openstack/nova stable/ussuri: Add functional test for bug 1937375 https://review.opendev.org/c/openstack/nova/+/803722 | |
| 10:16:40 | opendevreview | Lee Yarwood proposed openstack/nova stable/ussuri: compute: Avoid duplicate BDMs during reserve_block_device_name https://review.opendev.org/c/openstack/nova/+/803723 | |
| 10:17:28 | bauzas | wow, I can see light at the end of the tunnel : https://events.linuxfoundation.org/kubecon-cloudnativecon-north-america/register/ | |
| 10:21:47 | kashyap | pjakuszew: Thank you; do keep me posted. And post a link to your script - perhaps that can be added in some dangerous_utils/ directory | |
| 10:22:21 | kashyap | bauzas: That's still living in la-la-land for LF -- I doubt many will register until the dust settles with the Delta variant | |
| 10:22:59 | bauzas | kashyap: yup and the travel ban is still effective, but that's a fresh change at least for dometic flights | |
| 10:24:06 | kashyap | bauzas: Domestic flights in the US, you mean? | |
| 10:24:13 | bauzas | yup | |
| 10:24:33 | bauzas | this gives us some direction | |
| 10:24:48 | pjakuszew | kashyap: this is something I wanted to add into nova_manage in a more accessible form, actually | |
| 10:25:05 | opendevreview | Merged openstack/nova master: func: Increase rpc_response_timeout in TestMultiCellMigrate tests https://review.opendev.org/c/openstack/nova/+/803714 | |
| 10:25:22 | bauzas | conferences could run for a mixed virtual+physical system until restrictions slowly improve | |
| 10:25:43 | kashyap | Mixed confs suck like hell :-( It frustrates both in-person and the remotees | |
| 10:26:10 | kashyap | "Can you hear me? Oh, I was talking to myself. Anybody there? Oh, you all went for a coffee break? Never mind" | |
| 10:34:02 | pjakuszew | kashyap: are these server properties visible in the libvirt xml in your case? looks like adding that property both under server create --property and manually on another one doesn't seem to propagate | |
| 10:34:26 | kashyap | pjakuszew: They are not; see stephenfin's comment earlier | |
| 10:34:51 | pjakuszew | oh I overlooked that | |
| 10:35:10 | kashyap | pjakuszew: Only way for existing instances are what lyarwood mentioned earlier (snapshot the image; adjust the property; and then boot an instance off the updated snapshot image) | |
| 10:43:57 | opendevreview | Lee Yarwood proposed openstack/nova stable/train: Add functional test for bug 1937375 https://review.opendev.org/c/openstack/nova/+/803727 | |
| 10:43:58 | opendevreview | Lee Yarwood proposed openstack/nova stable/train: compute: Avoid duplicate BDMs during reserve_block_device_name https://review.opendev.org/c/openstack/nova/+/803728 | |
| 10:50:08 | opendevreview | Merged openstack/nova master: fup: Move _wait_for_volume_attach into InstanceHelperMixin https://review.opendev.org/c/openstack/nova/+/802623 | |
| 11:50:55 | lyarwood | stephenfin / gibi ; https://review.opendev.org/c/openstack/nova/+/802801/ could I get a +W on this now it's passing again with the above func test landed | |
| 12:58:45 | opendevreview | Balazs Gibizer proposed openstack/nova master: [nova-manage]support extended resource request https://review.opendev.org/c/openstack/nova/+/802060 | |
| 13:15:32 | bauzas | stephenfin: hola, before I leave, I'd appreciate if you could just look at the last rev for https://review.opendev.org/q/topic:%22bp%252Fgeneric-mdevs%22+(status:open%20OR%20status:merged) we need a second pair of eyes | |
| 13:15:44 | bauzas | since you already looked at it | |
| 13:27:21 | sean-k-mooney | bauzas: im off today but i will sync with the rest of the team and see if i can push the mdev work forward while your on pto | |
| 13:27:38 | kashyap | sean-k-mooney: Don't taint your free day with work... | |
| 13:28:17 | sean-k-mooney | :) that is why im not going to review today | |
| 13:28:38 | sean-k-mooney | anyway i have almost finished what i need to do on my laptop soo ill be gone soonish | |
| 13:29:29 | bauzas | kashyap: by the way, I apologize for you, my co-worker for the mdev series ;) | |
| 13:29:43 | bauzas | kashyap: we hadn't time to discuss about this blueprint :p | |
| 13:30:06 | bauzas | sean-k-mooney: as you want, anyway I'll be back on Tuesday, 2 days before FF | |
| 13:30:09 | kashyap | bauzas: I was supposed to pair on it, right? Yeah, downstream-- stuff completely engulfed me :-( | |
| 13:30:15 | bauzas | ;) | |
| 13:30:21 | bauzas | kashyap: heh no worries | |
| 13:31:25 | stephenfin | bauzas: will do | |
| 13:31:31 | bauzas | ta | |
| 13:33:30 | kashyap | bauzas: I was reading the mdev class code yesterday and then had to reboot the laptop and my memory flushed it. I know, I sound lame. | |
| 13:36:34 | kashyap | bauzas: Did you get around to testing it on real hardware? Or "don't ask?" | |
| 13:37:00 | bauzas | kashyap: I did | |
| 13:37:19 | bauzas | I even bubbled the hostname in some pastebin, whoops | |
| 13:37:26 | bauzas | :) | |
| 13:37:56 | bauzas | I was smart enough to keep the hardware that was given for me for testing the Ampere GPU :p | |
| 13:38:13 | bauzas | so, technically, I tested this series against MIG-backed vGPUs ;) | |
| 13:38:14 | kashyap | I see; yeah, I know the hardware acquis on this PITA | |
| 13:38:46 | kashyap | s/on this/here is a/ | |
| 13:39:04 | bauzas | hopefully I should still have this server when I'm back :p | |
| 13:39:53 | kashyap | bauzas: The only thing I know about MIG is that it can divide the GPU into several "mini GPUs". | |
| 13:40:36 | kashyap | s/mini/flexible/ | |
| 13:52:36 | gibi | I associate to a fighter jet when I see MIG. Can we virtualize fighter jets these days? :) | |
| 13:55:08 | kashyap | gibi: Heh, I too had to overwrite my memory to refer to a GPU instead of a fighter-jet | |
| 13:58:40 | bauzas | gibi: lolelly | |
| 13:59:39 | gibi | we had MIGs in the past, nowadays we have Grippens | |
| 14:00:23 | opendevreview | Balazs Gibizer proposed openstack/nova master: Block servers with vGPU and device profile in heal_allocations https://review.opendev.org/c/openstack/nova/+/803603 | |
| 14:01:31 | gibi | bauzas: fixed the failing unit tests there ^^ | |
| 14:08:17 | opendevreview | Stephen Finucane proposed openstack/nova master: api: Rename 'parameter_types.hostname' -> 'fqdn' https://review.opendev.org/c/openstack/nova/+/778549 | |
| 14:08:17 | opendevreview | Stephen Finucane proposed openstack/nova master: tests: Speed up 'servers' API tests https://review.opendev.org/c/openstack/nova/+/778732 | |
| 14:08:18 | opendevreview | Stephen Finucane proposed openstack/nova master: api: Add support for 'hostname' parameter https://review.opendev.org/c/openstack/nova/+/778550 | |