| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2020-09-03 | |||
| 13:55:58 | lyarwood | dansmith: yeah sorry did you have a draft change somewhere for that? | |
| 13:56:05 | dansmith | lyarwood: yeah | |
| 13:56:13 | dansmith | lyarwood: https://review.opendev.org/#/c/731577/ | |
| 13:56:29 | lyarwood | dansmith: cool thanks, I'll try to sort something out shortly | |
| 13:56:45 | dansmith | lyarwood: thanks, sorry to play the "shame is a powerful motivator" card :P | |
| 13:57:34 | lyarwood | dansmith: np I said I would do this and totally forgot so it's valid :) | |
| 13:57:42 | artom | gibi, was a func test for https://review.opendev.org/#/c/748453/ ever discussed? | |
| 13:57:44 | dansmith | :) | |
| 13:59:41 | sean-k-mooney | artom: not for that but there are some func tests later in the series | |
| 13:59:52 | sean-k-mooney | in https://review.opendev.org/#/c/740995/14 | |
| 14:00:24 | CeeMac | stephenfin: I managed to find a workable solution! | |
| 14:00:29 | sean-k-mooney | for what its worth i do think the first 4 patches in the series are ready to merge if we can get more core eyes on them | |
| 14:00:51 | sean-k-mooney | the following 3 im re reviewing | |
| 14:01:03 | artom | sean-k-mooney, afaict those tests don't have both SRIOV and NUMA in the same instance | |
| 14:01:12 | stephenfin | CeeMac: Nice. What did you do? | |
| 14:01:15 | artom | And don't test live migration | |
| 14:01:33 | artom | sean-k-mooney, would be reasonable to ask for that, no? | |
| 14:01:49 | sean-k-mooney | it would be valid | |
| 14:01:52 | artom | We should have the framework(s) in place already... | |
| 14:02:07 | sean-k-mooney | i cloud re run my sriov testing with numa guest too i guess | |
| 14:02:21 | sean-k-mooney | i tested live migration with seriov but not of a numa instance | |
| 14:03:13 | sean-k-mooney | i still suspect "Make PCI claim NUMA aware during live migration" is incomplte but its strictly better then what we did before | |
| 14:03:40 | CeeMac | stephenfin: as the instance is volume backed, I removed the instance (keeping the volume), set the hw_qemu_guest_agent=yes image-property on the volume. Launched a new instance from the volume settings hw_qemu_guest_agent=yes as customer meta-data. | |
| 14:04:03 | lyarwood | elod / melwitt ; thanks for the reviews again! https://review.opendev.org/#/q/topic:bug/1889108+branch:stable/stein should be ready now if you have time | |
| 14:04:07 | CeeMac | unix socket is created, guest-ping works from host :D | |
| 14:04:43 | sean-k-mooney | CeeMac: ya that shoudl work | |
| 14:05:08 | sean-k-mooney | since we will pull the volume metadata for the new isntance | |
| 14:05:09 | stephenfin | CeeMac: Great to hear :) If you've time and are set up for Gerrit, that'd be a nice little footnote to include in the 'doc/source/admin/configuration/hypervisor-kvm.rst' doc. | |
| 14:05:26 | stephenfin | If not, at least you're sorted | |
| 14:06:03 | CeeMac | stephenfin: i have gerrit set up | |
| 14:06:13 | CeeMac | whats the process for updating the doc? | |
| 14:06:21 | CeeMac | haven't done that yet | |
| 14:06:41 | stephenfin | CeeMac: you'll want to clone the nova repo and edit the source doc at 'doc/source/admin/configuration/hypervisor-kvm.rst' | |
| 14:06:57 | CeeMac | just master? | |
| 14:07:12 | stephenfin | That's rendered and results in https://docs.openstack.org/nova/latest/admin/configuration/hypervisor-kvm | |
| 14:07:23 | stephenfin | Yeah, we can cherry-pick it back then | |
| 14:07:27 | elod | lyarwood: will do! to tell you the truth it's a bit too many refactoring patch backport for my taste :/ | |
| 14:07:30 | stephenfin | backport it | |
| 14:08:32 | CeeMac | i'll give it a go! | |
| 14:09:13 | lyarwood | elod: yeah I appreciate that, some of these should've already been backported as bugfixes tbh | |
| 14:09:26 | lyarwood | elod: but the func changes are required to actually test this correctly | |
| 14:09:35 | lyarwood | elod: and it's all pretty trivial IMHO | |
| 14:11:11 | openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: Bump MIN_{LIBVIRT,QEMU}_VERSION and NEXT_MIN_{LIBVIRT,QEMU}_VERSION https://review.opendev.org/746981 | |
| 14:11:11 | openstackgerrit | Lee Yarwood proposed openstack/nova master: fakelibvirt: Use versionutils to set min versions found in the driver https://review.opendev.org/749707 | |
| 14:11:12 | openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: Remove MIN_{LIBVIRT,QEMU}_NATIVE_TLS_VERSION https://review.opendev.org/746983 | |
| 14:11:12 | openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: Remove MIN_LIBVIRT_FILE_BACKED_DISCARD_VERSION https://review.opendev.org/746982 | |
| 14:11:13 | openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: Remove MIN_LIBVIRT_VIDEO_MODEL_VERSIONS https://review.opendev.org/746985 | |
| 14:11:13 | openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: Remove MIN_LIBVIRT_BETTER_SIGKILL_HANDLING https://review.opendev.org/746984 | |
| 14:11:14 | openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: Remove MIN_{LIBVIRT,QEMU}_PMEM_SUPPORT https://review.opendev.org/746986 | |
| 14:15:30 | elod | lyarwood: ok, i see. I'll try to overcome my taste o:) | |
| 14:16:46 | CeeMac | stephenfin: is there a topic naming convention for branching? | |
| 14:17:13 | artom | sean-k-mooney, wait, we didn't add any functional tests for SRIOV live migration? | |
| 14:17:30 | artom | sean-k-mooney, all I see is the PCI stuff - ie, not Neutron ports, but flavor-driven | |
| 14:17:32 | sean-k-mooney | artom: not without stephenfin patch set | |
| 14:17:32 | stephenfin | CeeMac: If it's a bug, bug/<bug-id>. Because you likely want this backported, I'd suggest opening one if you can? | |
| 14:17:42 | stephenfin | artom: I have it done but it didn't exist previously, no | |
| 14:17:49 | stephenfin | It was a lot of work, in fairness | |
| 14:17:55 | artom | stephenfin, it merged? | |
| 14:17:58 | sean-k-mooney | no | |
| 14:17:58 | artom | Link plz? | |
| 14:18:04 | CeeMac | stephenfin: sure I can open a bug for this | |
| 14:18:07 | sean-k-mooney | its conflicting wiht many things | |
| 14:18:16 | stephenfin | https://review.opendev.org/#/q/status:open+project:openstack/nova+branch:master+topic:sriov-functional-tests | |
| 14:18:27 | CeeMac | i'll grab a coffee first! | |
| 14:18:40 | sean-k-mooney | artom: specificaly its conflicting with both my live migration bug fix your live migration bug fix and gibi sriov attach/detach series | |
| 14:18:43 | stephenfin | CeeMac: Sure thing :) | |
| 14:19:40 | artom | sean-k-mooney, sounds about right | |
| 14:19:44 | sean-k-mooney | the conflict are not surpriseing since we are all modifying similar parts fo the code | |
| 14:19:51 | sean-k-mooney | and adding tests | |
| 14:19:57 | artom | stephenfin, wait, how does https://review.opendev.org/#/c/746950/6 pass if you're depending on gibi's NUMA SRIOV live migration fix? | |
| 14:20:30 | sean-k-mooney | it does not depend on it | |
| 14:20:30 | stephenfin | artom: I'm not. His fix is for a certain edge condition I'm not triggering | |
| 14:20:51 | artom | Err, *you're not | |
| 14:20:51 | stephenfin | I did try to add a test for it but clearly didn't understand the bug enough | |
| 14:20:54 | stephenfin | *well enough | |
| 14:21:04 | sean-k-mooney | stephenfin: by edgecase you mean the fact we dont consider the numa toplogy of the guest at all right now | |
| 14:21:14 | artom | Also, having NUMA and SRIOV in the same instance is not an edge condition :) | |
| 14:21:19 | artom | Unless I'm way off base on that bug | |
| 14:21:32 | stephenfin | sean-k-mooney: it's to do with the claiming bit right? | |
| 14:21:38 | sean-k-mooney | yes | |
| 14:21:46 | sean-k-mooney | when we claim we pass None currenlty | |
| 14:21:50 | stephenfin | okay, so I have https://review.opendev.org/#/c/749360/ | |
| 14:22:34 | artom | stephenfin, ah, you're not using NUMA | |
| 14:22:36 | stephenfin | but the guest NUMA topology appears to be getting correctly recalculated to ensure the instance lands on the same host NUMA node as the PCI device | |
| 14:22:36 | artom | Just SRIOV | |
| 14:22:42 | artom | Not in the same test, at least | |
| 14:22:47 | stephenfin | I am in that one | |
| 14:22:51 | stephenfin | I'm using CPU pinning | |
| 14:22:55 | stephenfin | that's NUMA | |
| 14:23:14 | stephenfin | https://review.opendev.org/#/c/749360/1/nova/tests/functional/libvirt/test_pci_sriov_servers.py@363 | |
| 14:24:08 | sean-k-mooney | it could be passing jsut out of luck | |
| 14:24:21 | artom | Interesting, I wonder if there's a specific assertion that's missing | |
| 14:24:32 | sean-k-mooney | you would need to force a case where cpus are only avaiable on one node and the pci device is on another | |
| 14:24:38 | sean-k-mooney | with the bug it would boot | |
| 14:24:46 | stephenfin | sean-k-mooney: Yeah, I think that's what I'm missing | |
| 14:24:56 | openstackgerrit | Lee Yarwood proposed openstack/nova stable/train: compute: Validate a BDMs disk_bus when provided https://review.opendev.org/744552 | |
| 14:25:09 | stephenfin | But I wasn't able to come up with a clever way to do that since conf is global :-( | |
| 14:25:12 | sean-k-mooney | by the way gibi's fix i dont thnk it a full fix | |
| 14:25:53 | sean-k-mooney | since we are claiming the devices at a different point then we are generating the numa toplogy there is always a race with outher vms unless this is all happening under a common lock | |
| 14:26:00 | sean-k-mooney | and i dont think it is | |