| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2023-12-20 | |||
| 09:04:52 | opendevreview | Merged openstack/nova master: Call Neutron immediately upon _post_live_migration() start https://review.opendev.org/c/openstack/nova/+/883682 | |
| 10:24:09 | opendevreview | Balazs Gibizer proposed openstack/nova stable/zed: Allow live migrate paused instance when post copy is enabled https://review.opendev.org/c/openstack/nova/+/904085 | |
| 10:37:18 | opendevreview | Merged openstack/nova-specs master: Proposes mdev live-migration support in libvirt https://review.opendev.org/c/openstack/nova-specs/+/900636 | |
| 10:49:57 | bauzas | gibi: thanks for the spec review but actually about how to check the destination compute, I was wrong | |
| 10:50:27 | bauzas | gibi: I just checked and we recreate the migrate-data object in the source by using the latest version | |
| 10:50:41 | bauzas | so target_mdevs field should be here | |
| 10:51:07 | bauzas | so I'll create another field for just telling whether we support | |
| 10:51:25 | bauzas | like https://github.com/openstack/nova/blob/master/nova/objects/migrate_data.py#L270 | |
| 10:54:33 | gibi | addign a field there is OK to me. I guess we don't persist those | |
| 11:08:24 | bauzas | nope indeed | |
| 11:09:10 | bauzas | I'll start to upload the mdev-liv-mig series hopefully today EOB | |
| 11:39:20 | damiandabrowski | Hey! I looked into my issue with VMs that are moved to the other AZs during resize/migrate. | |
| 11:39:28 | damiandabrowski | sean-k-mooney: bauzas Thanks for showing me code responsible for preventing this behavior: https://github.com/openstack/nova/blob/d28a55959e50b472e181809b919e11a896f989e3/nova/compute/api.py#L1205 | |
| 11:39:37 | damiandabrowski | The problem is: this method is executed only when instance is created and due to the _get_volumes_for_bdms behavior, it sets proper request specs only when instance is booted from pre-existing volume. | |
| 11:39:43 | damiandabrowski | But if user creates a VM using CLI and --boot-from-volume argument, the VM does not have AZ assigned in requests specs even though its root volume is associated with a specific AZ. | |
| 11:39:49 | damiandabrowski | Should it be considered as a bug? | |
| 12:34:56 | sean-k-mooney[m] | im not sure it could be fixed as a bug | |
| 12:35:50 | sean-k-mooney[m] | i would have expeced boot from volume form image and boot form prexising volume to have the same behavior | |
| 12:36:24 | sean-k-mooney[m] | i.e. both a boot form voluem but in one instance nova is creating the volume and in the other we are using a user provided volume | |
| 12:37:06 | sean-k-mooney[m] | it is expected that we execute this only once at instance creation | |
| 12:42:04 | sean-k-mooney[m] | i assume the ax is not the default az “nova” | |
| 12:42:05 | sean-k-mooney[m] | https://github.com/openstack/nova/blob/d28a55959e50b472e181809b919e11a896f989e3/nova/compute/api.py#L1333 | |
| 12:44:29 | sean-k-mooney[m] | this code https://github.com/openstack/nova/blob/d28a55959e50b472e181809b919e11a896f989e3/nova/compute/api.py#L1399-L1406 will only work with precreated volumes | |
| 12:45:04 | sean-k-mooney[m] | but i tought there was code later to update the request spec if its bfv form image | |
| 12:47:16 | sean-k-mooney[m] | we would have to do that in the conductor after we have created the volume unless that happens on the computer. fi it does then its too late to update the request spec | |
| 12:49:36 | damiandabrowski | yeah exactly, _get_volumes_for_bdms considers only precreated volumes | |
| 12:50:06 | damiandabrowski | and since bfv from image is problematic on my dev environment and bfv from existing volume works fine, we most likely don't have anything that would update request specs later on :/ | |
| 12:50:19 | damiandabrowski | i can file a bug report today if you don't have anything against it | |
| 12:53:55 | sean-k-mooney[m] | so i think this might be a non bakcportable feature not a bug | |
| 12:54:00 | sean-k-mooney[m] | https://github.com/openstack/nova/blob/master/nova/conductor/manager.py#L1646 | |
| 12:54:23 | sean-k-mooney[m] | looking at that method we do not create the cinder volumes before down calling to the compute agent | |
| 12:54:40 | sean-k-mooney[m] | the request specifically is stored in the api db | |
| 12:55:01 | sean-k-mooney[m] | we do not allow upcalls form the compute-agent to the api db in general | |
| 12:56:06 | sean-k-mooney[m] | so to resolve this we would need to move the volume creation to the conductor prior to schduleing | |
| 12:59:43 | sean-k-mooney[m] | alternitively prior to the downcall we could get the selected hosts AZ and update the requet spec here https://github.com/openstack/nova/blob/master/nova/conductor/manager.py#L1779 | |
| 13:01:36 | sean-k-mooney[m] | we have already looked up the instance.az here https://github.com/openstack/nova/blob/master/nova/conductor/manager.py#L1727 | |
| 13:01:46 | sean-k-mooney[m] | so we would need to add | |
| 13:02:06 | sean-k-mooney[m] | if request_spec.is_bfv: | |
| 13:04:01 | sean-k-mooney[m] | if request_spec.az is non and instance.az != conf.default_az: | |
| 13:04:14 | sean-k-mooney[m] | request_spec.az = instance.az | |
| 13:04:24 | sean-k-mooney[m] | request_spec.save() | |
| 13:06:54 | sean-k-mooney[m] | that might work but it would pin the vm to the az of the first host we trided. that may or may not have space to create the volume on that backend and it may still fail for other reasons when we down call meaing we would have to ensure the alternat hosts are also in the same az | |
| 13:07:58 | sean-k-mooney[m] | damiandabrowski: ^ so i guess you can file a bug but i suspect this was not done orginally since nova created ports and volumes are normally only created by the compute agent which is too late to update the request spec. | |
| 13:08:33 | opendevreview | Merged openstack/nova master: Use configuration for single enabled mdev type https://review.opendev.org/c/openstack/nova/+/899406 | |
| 13:11:42 | damiandabrowski | sean-k-mooney: thanks for explanation! that's a lot of information to understand for now, but I'll try to process it all today :D | |
| 13:49:00 | opendevreview | Merged openstack/nova stable/zed: Allow live migrate paused instance when post copy is enabled https://review.opendev.org/c/openstack/nova/+/904085 | |
| 15:03:40 | opendevreview | Merged openstack/nova master: Resolve mypy error https://review.opendev.org/c/openstack/nova/+/903530 | |
| 15:03:48 | opendevreview | Merged openstack/nova master: tox: Use pre-commit for pep8 target https://review.opendev.org/c/openstack/nova/+/896227 | |
| 18:13:21 | melwitt | stephenfin: it looks like the CI failure on https://review.opendev.org/c/openstack/nova/+/903749 is legit ... codespell found a misspelling. not sure how it didn't fail before though | |
| 18:15:18 | sean-k-mooney | interesting my guess is previous runs had not been rebased | |
| 18:15:26 | melwitt | oh it was from the merge of post_live_migration reordering patch heh | |
| 18:15:43 | melwitt | there was one misspelling in that | |
| 18:16:21 | sean-k-mooney | i should check something | |
| 18:16:35 | melwitt | https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L9347 | |
| 18:17:35 | sean-k-mooney | i was wondering if it was https://github.com/openstack/nova/commit/2ffdf2e9173ca2ad0f95ac7f565da64b8b62691f | |
| 18:17:54 | melwitt | I wondered that at first too | |
| 18:17:58 | sean-k-mooney | but precommit is using --all-files before and after | |
| 18:18:17 | sean-k-mooney | i knwo stephen mentioned i shoudl use all-files but i could nto recall if i had made that change or not | |
| 18:18:18 | melwitt | https://zuul.opendev.org/t/openstack/build/de967e204d724b27821268c1fe0c5b75/log/job-output.txt#4132 | |
| 18:19:05 | sean-k-mooney | ok am i can just submit a change to adress the typo quickly if ye like | |
| 18:19:21 | sean-k-mooney | or if you want to do that | |
| 18:19:33 | melwitt | no go ahead | |
| 18:19:33 | sean-k-mooney | then we can proceed with stephens change | |
| 18:22:02 | sean-k-mooney | oh | |
| 18:22:14 | sean-k-mooney | so stephen is bumpiong the codespell version | |
| 18:22:21 | sean-k-mooney | so on master it passes fine | |
| 18:22:30 | melwitt | oh huh | |
| 18:22:46 | sean-k-mooney | but with 2.2.6 presumablly it addes new misspeling to the list | |
| 18:22:51 | melwitt | why would it consider that not a misspelling in the older version 😆 | |
| 18:23:03 | melwitt | yeah, guess so | |
| 18:23:16 | sean-k-mooney | so codespell works by having a list of known incorrect words | |
| 18:23:26 | sean-k-mooney | not a dictonary of correct words | |
| 18:23:33 | melwitt | oh, interesting | |
| 18:23:47 | sean-k-mooney | thats why it tends not to have a lot of false positives | |
| 18:24:33 | sean-k-mooney | brb | |
| 18:29:36 | sean-k-mooney | so stephen is done for today but he is fine wiht use just updating this. im confirming if i can see it locally now | |
| 18:30:41 | sean-k-mooney | oh... | |
| 18:30:48 | sean-k-mooney | so i think i know whats happening | |
| 18:31:02 | sean-k-mooney | so locally this is passing for me evne if i clean the cached copies | |
| 18:31:13 | sean-k-mooney | but the ci does not run on the patch under review. | |
| 18:31:23 | sean-k-mooney | the ci run on the patch under review rebased on top of master | |
| 18:31:39 | sean-k-mooney | and stepehns patch does not have artoms one | |
| 18:32:39 | sean-k-mooney | yep if i rebase stephens patch locally i see the same issue | |
| 18:32:43 | melwitt | ah right | |
| 18:33:06 | sean-k-mooney | and that why we actully run this in ci | |
| 18:34:03 | opendevreview | sean mooney proposed openstack/nova master: pre-commit: Add mypy https://review.opendev.org/c/openstack/nova/+/896228 | |
| 18:34:03 | opendevreview | sean mooney proposed openstack/nova master: pre-commit: Bump linter versions https://review.opendev.org/c/openstack/nova/+/903749 | |
| 18:37:20 | sean-k-mooney | melwitt: ^ care to reapprove | |
| 18:40:57 | melwitt | done | |
| 18:41:41 | sean-k-mooney | by the way i thin kthe merge conflcit with yoru serise was in the patch that is already merged | |
| 18:41:59 | sean-k-mooney | so if that does not merged shortly you could just unparent again if you want too | |
| 18:42:17 | sean-k-mooney | melwitt: are you around tomorow/friday | |
| 18:42:36 | melwitt | sean-k-mooney: I'm around tomorrow | |
| 18:42:57 | sean-k-mooney | ok i might try and review your code tomorrow so if i can find time | |
| 18:43:16 | sean-k-mooney | are you taking much pto over the next few weeks | |
| 18:44:13 | melwitt | we have the year-end recharge Dec 22 until Jan 2 | |
| 18:44:38 | sean-k-mooney | ill be back on the 8th if you think its ready i think it woudl be great to try an land your seriese in early january | |
| 18:45:57 | melwitt | ++ yeah I'm pushing hard to get everything into finished shape by then. it's mostly adding unit tests and stuff like that | |
| 18:46:28 | melwitt | (i.e. the implementation won't change much if at all) | |
| 18:46:36 | sean-k-mooney | well to aid with athat i think it woudl be good to see if we can merge som of the eairler patches. | |