| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-05-09 | |||
| 15:46:42 | mnaser | mriedem: would it be appended to the 'existing' cpu flags or it would have to match a hypervisor that has those cpu flags (i.e I would have a bunch of nested-virt enabled computes, and those that aren't) | |
| 15:48:07 | mriedem | mnaser: my understanding from https://specs.openstack.org/openstack/nova-specs/specs/rocky/implemented/report-cpu-features-as-traits.html is that the compute reports the cpu features as traits, and you can configure each compute with additional flags using https://docs.openstack.org/nova/latest/configuration/config.html#libvirt.cpu_model_extra_flags | |
| 15:48:48 | mriedem | i think cpu_model_extra_flags get reported as traits as well, but i would have to dig into the code to confirm | |
| 15:49:32 | mnaser | but if I add `cpu_model_extra_flags` that would then make every instance get nested virt | |
| 15:49:39 | mnaser | (on that host) | |
| 15:49:49 | mriedem | oh i see, you want per-instance | |
| 15:50:01 | mnaser | yeah, that way I don't have to maintain a whole bunch of host aggrgates | |
| 15:50:06 | mriedem | well, geez, that sounds like starlingx craziness bro! | |
| 15:50:28 | mnaser | is that something they do / looking to do ? | |
| 15:50:36 | mriedem | i want to say yes | |
| 15:51:19 | mriedem | i thought somewhat related to https://review.opendev.org/#/c/637834/ but not really what you want | |
| 15:51:29 | mnaser | it looks like the workflow of google is per-image based -- https://cloud.google.com/compute/docs/instances/enable-nested-virtualization-vm-instances | |
| 15:53:08 | mriedem | yar, i know we have things like hw_vif_model, hw_pointer_model, hw_scsi_model but no hw_cpu_model | |
| 15:54:30 | mnaser | yeah I guess the weird thing is `hw_cpu_model` actually ties into physical infra instead of virtual only | |
| 15:54:32 | mriedem | not sure where vcpu_model comes from on the instance | |
| 15:54:56 | mnaser | i.e. `hw_cpu_model` depends on the physical CPU having a certain capability so it might not always work | |
| 15:55:10 | mnaser | whereas the rest will probably always work as they are virtualized things | |
| 15:55:36 | mriedem | https://review.opendev.org/#/c/148138/ | |
| 15:57:40 | mriedem | allowing the user to specify the cpu features they want sounds like something kashyap might be aware of if someone is working on it | |
| 15:58:51 | mnaser | because right now I think the assumption is: if a host advertises support for a specific trait, it will always boot vms with that cpu flag | |
| 15:59:33 | mriedem | mnaser: ah here you go, starlingx did have a spec for it at one point https://review.opendev.org/#/c/168982/ (liberty) | |
| 15:59:47 | mriedem | thank my handy dandy starlingx diffs spreadsheet | |
| 16:00:02 | mnaser | where as I'm trying to have it be: if a host advertises support for a specific trait, it can be an opt-in one, so all vms boot without it, unless it's requested explicitly | |
| 16:01:42 | mnaser | yeah.. this seems to pretty much cover the majority of the use case, but huge scope | |
| 16:01:54 | mriedem | last week starlingx claimed to be down to 11 critical patches outside of nova and when i asked about that they said their were all for numa live migration, so if they have support for the user specifying cpu model in the flavor/image, i'm not sure what they are doing now | |
| 16:01:58 | mriedem | and cfriesen isn't around | |
| 16:02:12 | mriedem | note that spec also predates traits and all that by years | |
| 16:02:46 | mnaser | I was thinking maybe just add something like cpu_models_optin or something .. compute manager will advertise all traits (including ones in cpu_models_optin) and when it gets a vm, it'll decide to add/remove that flag if the flavor/image has that trait added | |
| 16:03:19 | mnaser | which would make the scheduler side of things fairly untouched, and it would be a virt-level thing... and the traits stuff will still work too | |
| 16:04:24 | mnaser | for t in requested_traits: if t in cpu_optional_features then add cpu flag else noop -- pretty much, | |
| 16:14:07 | openstackgerrit | Brian Haley proposed openstack/nova stable/queens: Teardown networking when rolling back live migration even if shared disk https://review.opendev.org/658149 | |
| 16:18:56 | dansmith | mriedem: I took that claim to mean they have only 11 patches of stuff planned to merge into nova and the rest is "value add" they're planning to just keep in their fork | |
| 16:19:17 | dansmith | which is noticeably different from the original stance on the "we don't want to maintain a fork" stuff that none of us believed anyway :) | |
| 16:19:42 | mriedem | yeah maybe | |
| 16:22:50 | mriedem | i know they have lots of other stuff like this, like specifying vif_model when attaching a port | |
| 16:25:39 | openstackgerrit | Dan Smith proposed openstack/nova master: Add extra logging to request filters https://review.opendev.org/657629 | |
| 16:30:02 | openstackgerrit | Surya Seetharaman proposed openstack/nova master: Disable limit if affinity(anti)/same(different)host is requested https://review.opendev.org/658110 | |
| 16:32:02 | mriedem | gibi: comments and questions and nits in your 'move ops for qos ports' spec https://review.opendev.org/#/c/652608/ but nothing worth holding it back | |
| 16:35:12 | roukoswarf | hey all, i have a fun one... so i boot an instance using an image and a volume, works great, hard resets fine. but i boot an instance with a volume as the boot disk, add a volume, then hard reset, and then it has "no bootable device" till the volume is detached | |
| 16:36:53 | roukoswarf | i cant find a decernable difference in the libvirt cfg that would cause it to not check all disks for mbr... but maybe theres some cfg im missing? replication steps are pretty simple, boot from a volume, add a volume, do a full power reset on the vm, fail. | |
| 16:57:00 | openstackgerrit | Dan Smith proposed openstack/nova master: Add extra logging to request filters https://review.opendev.org/657629 | |
| 17:04:57 | openstackgerrit | Mohammed Naser proposed openstack/nova master: [rfc] ability to use optional cpu flags https://review.opendev.org/658158 | |
| 17:05:15 | mnaser | mriedem: ^ well, I'd love to hear comments, I know this probably needs a spec and all sorts of other stuff.. but the 'general' idea | |
| 17:35:55 | openstackgerrit | Merged openstack/nova master: libvirt: auto detach/attach sriov ports on migration https://review.opendev.org/629589 | |
| 18:44:24 | openstackgerrit | Artom Lifshitz proposed openstack/python-novaclient master: Use SHA56 instead of MD5 in completion cache https://review.opendev.org/658181 | |
| 18:55:34 | openstackgerrit | Merged openstack/nova master: Remove ComputeDriver.macs_for_instance method https://review.opendev.org/652737 | |
| 18:59:18 | mriedem | https://review.opendev.org/#/q/topic:bp/libvirt-neutron-sriov-livemigration+(status:open+OR+status:merged) is all merged now but i'm not sure what docs are needed or where | |
| 19:00:26 | mriedem | ah yes https://docs.openstack.org/neutron/latest/admin/config-sriov.html#known-limitations | |
| 19:02:57 | mriedem | stephenfin: is ^ something you want to handle while sean is out on pto? | |
| 19:11:21 | openstackgerrit | Dustin Cowles proposed openstack/nova master: Use SDK instead of ironicclient for validating instance and node https://review.opendev.org/656028 | |
| 19:18:50 | artom | mriedem, sean's back Monday, and stephenfin is probably drinking and/or sleeping now | |
| 19:19:48 | mriedem | yeah, it was for his bouncer | |
| 19:19:58 | mriedem | irc, not the "bloke" at the "pub" | |
| 19:20:22 | artom | (Yeah, not sure why I thought you didn't know his tz) | |
| 19:22:51 | openstackgerrit | Merged openstack/nova master: Add --instance option to heal_allocations https://review.opendev.org/651945 | |
| 19:35:54 | openstackgerrit | Artom Lifshitz proposed openstack/python-novaclient master: Use SHA256 instead of MD5 in completion cache https://review.opendev.org/658181 | |
| 20:17:40 | redkrieg | I'm having an issue with lxd on my deployment where creating instances fails with "LXDAPIException: Profile is currently in use" in nova-compute.log. is there any good way to troubleshoot what could be causing that? | |
| 20:35:19 | openstackgerrit | Merged openstack/nova master: Update the contributor doc for macos https://review.opendev.org/658137 | |
| 21:01:11 | mriedem | efried: nova meeting today? | |
| 21:01:22 | efried | yes, sorry, starting... | |
| 21:01:23 | edleafe | heh, was just typing that :) | |
| 21:01:33 | mriedem | redkrieg: lxd isn't in tree, you'd have to ask in their channel | |
| 21:01:51 | redkrieg | cool, thanks | |
| 21:01:56 | mriedem | redkrieg: https://github.com/openstack/nova-lxd#support-and-discussions | |
| 21:03:05 | redkrieg | that helps, thank you mriedem. I'm doing a reinstall right now so I'll follow up there if I still have an issue | |
| 21:03:30 | mriedem | np | |
| 22:05:30 | openstackgerrit | Matt Riedemann proposed openstack/nova master: DNM: Test theory behind bug 1822884 https://review.opendev.org/649464 | |
| 22:05:32 | openstack | bug 1822884 in OpenStack Compute (nova) "live migration fails due to port binding duplicate key entry in post_live_migrate" [Undecided,In progress] https://launchpad.net/bugs/1822884 - Assigned to sean mooney (sean-k-mooney) | |
| 22:21:19 | mriedem | sorry | |
| 22:21:27 | mriedem | not sorry | |
| 22:22:18 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Update usage in RT.drop_move_claim during confirm resize https://review.opendev.org/641806 | |
| 22:22:19 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add Instance.hidden field https://review.opendev.org/631123 | |
| 22:22:19 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add InstanceAction/Event create() method https://review.opendev.org/614036 | |
| 22:22:19 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add Migration.cross_cell_move and get_by_uuid https://review.opendev.org/614012 | |
| 22:22:20 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add TargetDBSetupTask https://review.opendev.org/627892 | |
| 22:22:21 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Execute TargetDBSetupTask https://review.opendev.org/633853 | |
| 22:22:21 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add CrossCellMigrationTask https://review.opendev.org/631581 | |
| 22:22:22 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add prep_snapshot_based_resize_at_dest compute method https://review.opendev.org/633293 | |
| 22:22:22 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add can_connect_volume() compute driver method https://review.opendev.org/621313 | |
| 22:22:23 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add prep_snapshot_based_resize_at_source compute method https://review.opendev.org/634832 | |
| 22:22:23 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add PrepResizeAtDestTask https://review.opendev.org/627890 | |
| 22:22:24 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add PrepResizeAtSourceTask https://review.opendev.org/627891 | |
| 22:22:24 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add nova.compute.utils.delete_image https://review.opendev.org/637605 | |
| 22:22:25 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add power_on kwarg to ComputeDriver.spawn() method https://review.opendev.org/642590 | |
| 22:22:25 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Refactor ComputeManager.remove_volume_connection https://review.opendev.org/642183 | |
| 22:22:26 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add FinishResizeAtDestTask https://review.opendev.org/635646 | |
| 22:22:26 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add finish_snapshot_based_resize_at_dest compute method https://review.opendev.org/635080 | |
| 22:22:27 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add Destination.allow_cross_cell_move field https://review.opendev.org/614035 | |
| 22:22:38 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add cross-cell resize policy rule and enable in API https://review.opendev.org/638269 | |
| 22:22:38 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add CrossCellWeigher https://review.opendev.org/614353 | |
| 22:22:39 | openstackgerrit | Matt Riedemann proposed openstack/nova master: WIP: Enable cross-cell resize in the nova-multi-cell job https://review.opendev.org/656656 | |
| 22:58:54 | redkrieg | I reinstalled one of my compute nodes and thought I had removed it from openstack first but I think it reregistered before I shut it down. is there a good way to get around this? error: | |
| 22:58:58 | redkrieg | 2019-05-09 18:57:34.505 12028 ERROR nova.scheduler.client.report [req-d7fa7b90-ada9-4296-8f2f-977b8e041dfc - - - - -] [req-ce0a1ceb-1e0f-486f-afe6-c1dc9ecdbad5] Failed to create resource provider record in placement API for UUID 5f5eecf0-9805-4b0c-892e-cddb70b6a4eb. Got 409: {"errors": [{"status": 409, "request_id": "req-ce0a1ceb-1e0f-486f-afe6-c1dc9ecdbad5", "detail": "There was a conflict when | |
| 22:59:04 | redkrieg | trying to complete your request.\n\n Conflicting resource provider name: nyc-slxc-1 already exists. ", "title": "Conflict"}]}. | |
| 23:04:33 | cdent | redkrieg: quick and dirty: find the resource provider that represent the older compute node (its uuid) and delete that from placement. you can do that with curl calls to placement, the osc-placement client, or looking into the database directly | |
| 23:12:35 | redkrieg | cdent: thanks, I found the entry and deleted it. looks like it registered correctly after. | |
| 23:13:15 | cdent | redkrieg: great. I think there's been some work or at least discussion on making that less of a problem, but quick and dirty seemed like the right way to go in this case | |
| 23:13:48 | redkrieg | yep, it's not a common scenario. I generally avoid reinstalling when I can. | |
| 23:16:37 | mriedem | the proper thing would have been (1) stop the nova-compute process on the host, (2) delete the compute service in the API which (depending on your release) should cleanup the provider in placement automatically | |
| 23:18:07 | mriedem | https://developer.openstack.org/api-ref/compute/?expanded=delete-compute-service-detail#delete-compute-service | |