| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2026-07-09 | |||
| 00:27:58 | opendevreview | Takashi Kajinami proposed openstack/nova master: Use domain capabilities to detect SEV-SNP support https://review.opendev.org/c/openstack/nova/+/996316 | |
| 00:52:40 | sean-k-mooney | gibi: https://review.opendev.org/c/openstack/futurist/+/996567 and the followups may be of interst. | |
| 02:24:28 | opendevreview | Takashi Kajinami proposed openstack/nova master: libvirt: Remove redundant version check for virtio-fs https://review.opendev.org/c/openstack/nova/+/996496 | |
| 02:41:05 | opendevreview | Ghanshyam Maan proposed openstack/nova master: Call manager graceful_shutdown while RPC server waits https://review.opendev.org/c/openstack/nova/+/996583 | |
| 07:15:43 | frickler | anyone around for stable reviews of the OSSN-0101 fix? https://review.opendev.org/c/openstack/nova/+/995956 etc. | |
| 07:25:33 | opendevreview | Bhavana proposed openstack/nova master: Add Minimal Displacement Solver for MIG GPU Repartitioning https://review.opendev.org/c/openstack/nova/+/996058 | |
| 07:54:35 | gibi | frickler: +2 | |
| 07:54:47 | gibi | elodilles: could you look at these backports? https://review.opendev.org/c/openstack/nova/+/995956 | |
| 07:57:31 | opendevreview | Balazs Gibizer proposed openstack/nova master: HACK:Speed up functional test in threading mode https://review.opendev.org/c/openstack/nova/+/996500 | |
| 08:14:26 | opendevreview | Thibaut Démaret proposed openstack/nova master: libvirt: add disk rotation_rate support for local disks https://review.opendev.org/c/openstack/nova/+/979693 | |
| 08:39:04 | elodilles | gibi: +2'd it. will follow up with the backports to older branches when this merges | |
| 09:30:16 | gibi | elodilles: thanks | |
| 10:43:54 | ralonsoh | hi folks, a quick review for https://review.opendev.org/c/openstack/os-vif/+/995933 will be appreciated | |
| 10:43:55 | ralonsoh | thanks! | |
| 11:00:03 | opendevreview | Merged openstack/nova stable/2026.1: Fix mutating global config in websocket proxy https://review.opendev.org/c/openstack/nova/+/995956 | |
| 11:01:35 | opendevreview | Lajos Katona proposed openstack/nova master: Use SDK for Neutron networks https://review.opendev.org/c/openstack/nova/+/928022 | |
| 11:01:35 | opendevreview | Lajos Katona proposed openstack/nova master: Use SDK for Neutron subnets https://review.opendev.org/c/openstack/nova/+/962190 | |
| 11:53:00 | opendevreview | Ashish Gupta proposed openstack/placement master: tests: Support file-backed SQLite URL in placement fixtures https://review.opendev.org/c/openstack/placement/+/993106 | |
| 12:31:57 | opendevreview | Ashish Gupta proposed openstack/nova master: tests: use file-backed Placement SQLite in functional threading https://review.opendev.org/c/openstack/nova/+/992581 | |
| 12:34:54 | opendevreview | Ashish Gupta proposed openstack/nova master: tests: Use per-database write locks instead of global lock https://review.opendev.org/c/openstack/nova/+/992862 | |
| 13:14:19 | gibi | gmaan: sean-k-mooney: the threading functional test reproduces a seemingly valid API race condition between volume attach and attachment show https://bugs.launchpad.net/nova/+bug/2160224 | |
| 13:15:05 | sean-k-mooney | :) thats fun | |
| 13:15:42 | gibi | we create the BDM in one transaction and populate the attachment_id in second one | |
| 13:15:53 | gibi | in between the two there is a BDM but no attachment_id in it | |
| 13:15:59 | sean-k-mooney | i need to refine the futurist patches not only based on your feed back but to have them relibaliy pass the unit tests :) but there are a lot of optimisation opterutnities there | |
| 13:16:19 | gibi | but the API response schema on showing the attachment insists on having a not non attachment_id in the response | |
| 13:16:29 | gibi | ~not None | |
| 13:16:46 | sean-k-mooney | we proably should weaken it slightly | |
| 13:16:50 | gibi | I will push a fix to return 404 until the BDM has a valid attachement_id | |
| 13:16:58 | sean-k-mooney | if the servier has been built it shoudl be non non | |
| 13:17:00 | gibi | that does not need an response schema chang | |
| 13:17:40 | sean-k-mooney | oh for just GET servers/<server_id>/os-volume_attachments/<volume_id> | |
| 13:17:44 | gibi | yepp | |
| 13:17:48 | sean-k-mooney | so this isnt an issue in server show | |
| 13:17:51 | gibi | nope | |
| 13:17:59 | sean-k-mooney | its an isssue int he aggachemts proxy api | |
| 13:18:14 | sean-k-mooney | ok then ya a 404 makes sense before its finished attaching | |
| 13:18:23 | gibi | but I'm not sure what nova does if creating an attachment in cinder fails | |
| 13:18:38 | gibi | if we cannot show the attachment then I guess we cannot delete it either | |
| 13:18:57 | gibi | i.e. the user cannot show the attachment as we will return 404 | |
| 13:19:05 | gibi | hm | |
| 13:19:39 | gibi | but deleting an attachment only needs the volume_id not the attachement_id | |
| 13:19:46 | gibi | then the user can delete a half baked attachemtn | |
| 13:19:51 | gibi | the it is OK | |
| 13:20:05 | gibi | I can return 404 in show until the attachment_id is None | |
| 14:26:49 | opendevreview | Balazs Gibizer proposed openstack/nova master: HACK:Speed up functional test in threading mode https://review.opendev.org/c/openstack/nova/+/996500 | |
| 14:34:58 | opendevreview | Merged openstack/nova stable/2025.2: Fix mutating global config in websocket proxy https://review.opendev.org/c/openstack/nova/+/995957 | |
| 14:59:51 | sean-k-mooney | gibi: melwitt do we want to proceed with https://review.opendev.org/c/openstack/nova/+/969538 for now and we can loop back to adress_space vs rss vs both later? | |
| 15:02:10 | sean-k-mooney | i am not sure i have the capsity to track that and rather then forget about it i think im ok to upgrade to +2w if ye are both happy with the current appoch | |
| 15:11:55 | gibi | lets land it and get back to it later if needed | |
| 16:01:37 | melwitt | thanks gibi and sean-k-mooney, if you prefer rss I can quickly change it to that, I am good either way | |
| 16:05:53 | sean-k-mooney | melwitt: while i do in general i dont know htat i want to have to try and validate that does not break anythign else. it shoudl not but lets leave it for a follow up | |
| 16:06:19 | melwitt | sean-k-mooney: yeah that's kinda where I was at too. thanks | |
| 16:07:22 | sean-k-mooney | cool noted in the review +2w | |
| 16:08:41 | melwitt | super thanks :) | |
| 16:10:17 | sean-k-mooney | one thing to note is this test is not failign because of debian 12 and now 13 | |
| 16:11:07 | sean-k-mooney | so unless we try this on ubuntu 24.04 we wont know if its fully "fixed" i.e. if tweadkin the config option in the ci job is enough | |
| 16:16:16 | melwitt | yes that is true | |
| 16:17:52 | melwitt | IIRC the error was pretty clear it had hit the process limit, that combined with how ironic and cinder also have to configure it made me reasonably confident this would work | |
| 16:18:32 | sean-k-mooney | it shoudl but the defautl value might need to be higher, thats why i mentioned the job config. so one thing that might be interesting to try is how does ubuntu 26.04 and ceph intereact | |
| 16:19:00 | sean-k-mooney | although since debian 13 is fine | |
| 16:19:08 | sean-k-mooney | i think 26.04 will work as well | |
| 16:19:11 | melwitt | yeah, it was just the encryption overhead makes a 1G image slightly larger than 1G so the failure makes sense | |
| 16:32:14 | opendevreview | Ashish Gupta proposed openstack/placement master: tests: Support file-backed SQLite URL in placement fixtures https://review.opendev.org/c/openstack/placement/+/993106 | |
| 16:59:48 | gouthamr | melwitt: bauzas gibi: could you folks review this OSSN for me: https://review.opendev.org/c/openstack/security-doc/+/996187 | |
| 17:11:01 | melwitt | gouthamr: done | |
| 17:11:20 | gouthamr | thank you melwitt! | |
| 17:11:29 | gmaan | gibi: sean-k-mooney: on volume attachment id as None, agree on fixing the API and not schema. if atachment_id is not available we cannot guarantee that attachment is done in cinder side or not so returning 404 make sense . | |
| 17:13:33 | gmaan | gibi: sean-k-mooney now I am wondering that we used to return 200 before schema restriction which converted it to 400. now if I consider case where schema restriction is not enabled (actual production env) when by fixing this bug we will be converting 200 to 404 so we need microversion bump. | |
| 17:13:58 | gmaan | even tough it is bug fix, but need to take care of interoperability with version bump | |
| 17:14:50 | sean-k-mooney | gmaan: well if the attchment id is not avaibel the attchment is defnitly not don in nova yet | |
| 17:15:14 | sean-k-mooney | so regardless of cidner we have not compelted the attachment | |
| 17:15:30 | sean-k-mooney | so a 404 until that is fully compelt makes sense i think | |
| 17:17:04 | sean-k-mooney | the 200->404 change is worht consdiering but i agree that it should not need a microverion | |
| 17:19:04 | gmaan | think about users who used to call this API to monitor if attachment is ready until attachment_id is not None they will be broken from 200 to 404 so IMO it definitely need microversion | |
| 17:19:34 | gmaan | like server show to keep calling until server is active or ip addresses are pupulated in that | |
| 17:24:37 | sean-k-mooney | gmaan: right so that is alreayd broken | |
| 17:24:52 | sean-k-mooney | becuase it will be listed there before the attchment in nova is completed | |
| 17:25:08 | sean-k-mooney | gmaan: the only way to do that and not ahve a race today is via server show | |
| 17:26:45 | sean-k-mooney | gmaan: becasue of https://bugs.launchpad.net/nova/+bug/2058928 | |
| 17:28:33 | gmaan | yes, that action finish time could be helpful | |
| 17:29:41 | sean-k-mooney | the api gibi is trying to fix just tell you that we have complete attachign the device in cinder but not to the vm or nova when the id is populated | |
| 17:30:22 | sean-k-mooney | tehre are actuly 2 more step to completing the volume attachment on the nova side although populating the attachment id in the bdms is near the end | |
| 20:49:19 | opendevreview | Merged openstack/nova master: Make QEMU_IMG_LIMITS process limits configurable https://review.opendev.org/c/openstack/nova/+/969538 | |
| 21:09:42 | opendevreview | Merged openstack/nova master: tests: file-backed SQLite with WAL in threading mode for Database and CellDatabases Fixtures https://review.opendev.org/c/openstack/nova/+/988583 | |
| #openstack-nova - 2026-07-10 | |||
| 04:01:22 | opendevreview | Ghanshyam Maan proposed openstack/nova master: Task tracking mechanism for graceful shutdown https://review.opendev.org/c/openstack/nova/+/996299 | |
| 04:03:40 | opendevreview | Ghanshyam Maan proposed openstack/nova master: Call manager graceful_shutdown while RPC server waits https://review.opendev.org/c/openstack/nova/+/996583 | |
| 05:10:53 | opendevreview | Fabian Wiesel proposed openstack/nova master: vmware: Read iso file in binary mode during upload https://review.opendev.org/c/openstack/nova/+/996738 | |
| 09:06:24 | gibi | gmaan: sean-k-mooney: I responed in the bug about the attachment show API issue | |
| 09:06:59 | gibi | A proper client should handle 404 already while the volume is being attached as nova returns 404 when BDM is nto yet exists. | |
| 09:07:16 | sean-k-mooney | yes it should | |
| 09:07:24 | gibi | so returning 404 while BDM exists but without attachment_id is backward compatible from that client perspectvie | |
| 09:07:54 | sean-k-mooney | i dont think a 200 -> 404 shoudl require a microversion as it should alreay be getting the 404 before while we are attaching and retrying | |
| 09:08:10 | sean-k-mooney | so i agree that is backwards compatiable | |
| 09:08:39 | sean-k-mooney | but also tahat api cant be used to know fi the volume is actully attached today | |
| 09:09:07 | sean-k-mooney | although it will get closer to working for that with your proposed chagne | |
| 09:09:48 | gibi | yeah just because we have attachment_id it does not tell us if the compute finished attaching the volume to the VM | |
| 09:10:01 | gibi | the versioned notification is better in this regard as that is sent from the compute side | |
| 09:10:02 | sean-k-mooney | if i recall the oder correctly , the volume gets attached in cinder, then we udate the bdms and finally attach it to the vm | |
| 09:10:07 | gibi | yepp | |