Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2025-07-22
16:14:44 gibi hey thats me :)
16:15:17 gibi we are slowly landing patches from the scheduler series
16:15:45 gibi I got a nice set of reviews from bauzas on the doc patch. I have to go back there and touch up the doc
16:16:50 gibi sambork's patch https://review.opendev.org/c/openstack/nova/+/949754 logic looks good to me but I found some extra cleanup pieces and a bit of test issues
16:17:35 gibi and I'm following Dan's series starting https://review.opendev.org/c/openstack/nova/+/954990/4
16:18:16 gibi I still have the intention to go back making our unit tests run with threading
16:18:19 gibi that is it
16:18:31 gibi #topic Open discussion
16:18:34 gibi (sp-bmilanov) Bug #2092391: duplication instances when nova compute service restart: https://bugs.launchpad.net/nova/+bug/2092391
16:18:46 sp-bmilanov hi :)
16:19:07 gibi I guess this is a review request for https://review.opendev.org/c/openstack/nova/+/938223
16:19:10 gibi am I correct?
16:19:16 sp-bmilanov not exactly
16:19:21 gibi ohh
16:19:25 gibi then tell us :)
16:20:00 sp-bmilanov I wonder if it would be better to bring this up when more core people are around but still -- we hit this bug recently and it was not due to a graceful Nova agent shutdown
16:20:24 gibi what was the trigger?
16:20:55 sp-bmilanov the tldr; is that during a migration, if a nova-agent crashes at the correct moment, it is possible to have the same VM running on the source and destination hypervisor
16:21:22 dansmith I think it has already been noted on that bug that nova-compute doesn't really have any graceful shutdown support, and what the bug describes during a live migration is pretty much expected at the moment
16:22:14 gibi even with graceful shutdown a crash would not be handled
16:22:29 sp-bmilanov dansmith: right, I read Sean's comment as "it is not supported to ask nova-compute to shutdown during live migration"
16:22:31 dansmith gibi: I think the problem is likely on restart we re-activate the instance
16:22:36 gibi so I guess we need a solution where a compute starting up can fix the situation
16:22:46 gibi dansmith: yeah
16:22:54 sp-bmilanov yes, as gibi said, it's about when it crashes
16:23:08 dansmith and the review mentioned above would only be the non-crash situation
16:23:24 gibi yepp
16:23:45 gibi I feel that nova-compute during statup can be smarter about this to remove the VM duplication
16:24:03 sp-bmilanov the VM was seen in an error state after the nova-compute got back up because of a mismatch in what libvirt was reporting and the contents of the nova DB
16:25:51 sp-bmilanov a teammate suggested it would be better to have this as an separate error state which has more obstacles to get around until you are able to start the VM again
16:26:43 sp-bmilanov else nova-compute recreates the libvirt domain on VM start on the source hypervisor
16:26:53 gibi whichever compute puts the VM to error could be smarter and try to abort the live migration I guess
16:26:55 dansmith any solution for this is going to be something we need to document as a spec I think, because there are a lot of factors in play.. it's hard to know what to do when a live migration fails and in the past, we've basically said "lean on the operator to clean it up"
16:27:14 gibi dansmith: make sense
16:27:18 gibi it is complicated
16:27:46 dansmith preventing nova from re-starting on startup if it's not sure is good, but barriers to prevent the user from doing something bad are part of the complexityt
16:28:21 gibi I agree
16:28:45 gibi Also having a spec would force us to load context around this codepath (I don't have it loaded)
16:29:25 gibi sp-bmilanov: could you draft a spec even if it is just the problem statement with more details about what exactly happening and why
16:29:46 dansmith +1
16:29:55 gibi I think that would help us brainstorming on a list of potential solutions
16:30:11 sp-bmilanov sure can
16:30:19 gibi cool. thanks.
16:31:06 sp-bmilanov thanks gibi dansmith!
16:31:08 gibi Is there anything else to discuss?
16:32:32 gibi then thanks for joining today. Next week we will have Uggla back.
16:32:35 opendevmeet Log: https://meetings.opendev.org/meetings/nova/2025/nova.2025-07-22-16.00.log.html
16:32:35 opendevmeet Minutes (text): https://meetings.opendev.org/meetings/nova/2025/nova.2025-07-22-16.00.txt
16:32:35 opendevmeet Minutes: https://meetings.opendev.org/meetings/nova/2025/nova.2025-07-22-16.00.html
16:32:35 opendevmeet Meeting ended Tue Jul 22 16:32:35 2025 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)
16:32:35 gibi #endmeeting
16:32:40 elodilles thanks o/
16:33:46 gmaan o/
16:36:16 fwiesel o/
16:47:59 opendevreview Merged openstack/nova master: Rename DEFAULT_GREEN_POOL to DEFAULT_EXECUTOR https://review.opendev.org/c/openstack/nova/+/948086
16:48:13 opendevreview Merged openstack/nova master: Make the default executor configurable https://review.opendev.org/c/openstack/nova/+/948087
18:13:14 opendevreview Merged openstack/nova master: api: Add response body schemas for server group APIs https://review.opendev.org/c/openstack/nova/+/952281
19:25:50 mikal Heya. https://review.opendev.org/q/topic:%22libvirt-vdi%22 has been sitting with a single set of +2s for a couple of weeks. Does anyone have a minute to take a look at them please?
21:14:37 opendevreview Merged openstack/nova master: api: Address issues with server group APIs https://review.opendev.org/c/openstack/nova/+/953281
21:49:32 opendevreview Ghanshyam proposed openstack/nova master: Add project manager role in Nova API policy rule https://review.opendev.org/c/openstack/nova/+/953063
#openstack-nova - 2025-07-23
06:17:53 opendevreview Jonas Schäfer proposed openstack/nova master: Preserve vTPM state between power off and power on https://review.opendev.org/c/openstack/nova/+/955657
06:19:48 opendevreview Jonas Schäfer proposed openstack/nova master: Preserve vTPM state between power off and power on https://review.opendev.org/c/openstack/nova/+/955657
06:20:11 jssfr dansmith, there you go ^ :)
10:51:31 opendevreview Ian Watson proposed openstack/nova master: Implementing get_num_instances for ironic virt driver https://review.opendev.org/c/openstack/nova/+/955685
11:05:22 opendevreview Ian Watson proposed openstack/nova master: Closes-Bug: #2117697 Removing provider_tree removal upon 409 in set_aggregates_for_provider for race condition https://review.opendev.org/c/openstack/nova/+/955686
11:07:19 opendevreview Ian Watson proposed openstack/nova master: Closes-Bug: 2117700 Implementing get_num_instances for ironic virt driver https://review.opendev.org/c/openstack/nova/+/955685
11:13:29 opendevreview Ian Watson proposed openstack/nova master: Implementing get_num_instances for ironic virt driver https://review.opendev.org/c/openstack/nova/+/955685
11:14:53 opendevreview Ian Watson proposed openstack/nova master: Removing provider_tree removal upon 409 in set_aggregates_for_provider for race condition https://review.opendev.org/c/openstack/nova/+/955686
11:36:32 manumoh team, I facing a strange behaviour after migrating octavia to wallaby - my loadbalancer creations are failing if the used network (both lb-mgmt-network and the provider-network used as vrrp-net) contains more than one subnet. the amphora vm port has ips from both subnet and is causing the vms become not accessible. while checking the ports associated with the vm is see the following
11:36:58 manumoh https://www.irccloud.com/pastebin/aw4PlXn9/
11:38:38 manumoh I can see octavia makes a POST call to nova to create the vm with network uuid , does this create port with multiple ips?
11:39:20 manumoh any help would be great
11:54:07 opendevreview Ian Watson proposed openstack/nova master: Changing how placement updates are trigger by seeing if a full reshape has occured and whether any changes to the provider tree has taken place. Closes-Bug: #2117701 https://review.opendev.org/c/openstack/nova/+/955687
12:09:13 opendevreview Ian Watson proposed openstack/nova master: Adding a new method migrations_supported and using method to determine whether or not to issue get migrations api calls as part of resource tracker. Closes-Bug: #2117708 https://review.opendev.org/c/openstack/nova/+/955690
13:02:34 noonedeadpunk qemu
13:03:05 noonedeadpunk that ^ supposed to be search in the channel history :D
13:15:11 opendevreview Kamil Sambor proposed openstack/nova master: Replace eventlet.event.Event with threading.Event https://review.opendev.org/c/openstack/nova/+/949754
14:11:57 opendevreview sean mooney proposed openstack/nova master: Revert "Always delete NVRAM files when deleting instances" https://review.opendev.org/c/openstack/nova/+/922354
14:12:02 opendevreview sean mooney proposed openstack/nova master: Revert "Always delete NVRAM files when deleting instances" https://review.opendev.org/c/openstack/nova/+/922354
14:29:00 gibi eventlet sync starts in 1 minute https://meet.google.com/bcy-uqoz-hje
14:45:45 sean-k-mooney just saw this now i see folks have droped
14:52:26 opendevreview Kamil Sambor proposed openstack/nova master: Replace eventlet.event.Event with threading.Event https://review.opendev.org/c/openstack/nova/+/949754
15:53:16 gmaan stephenfin: fixed the doc comments, please check https://review.opendev.org/c/openstack/nova/+/953063
15:54:32 gibi sean-k-mooney: yeah we finished in 13 minutes the agenda doc has the details what we discussed
16:15:28 sean-k-mooney gibi: no worries, i didnt have time to look at the work since last week anyway
16:15:40 sean-k-mooney gibi: so i dint have anythign to dicuss
16:16:32 gmaan sean-k-mooney: I merged the manager role and new policy changes into single change, please check https://review.opendev.org/c/openstack/nova/+/953063
16:17:08 gmaan that gives complete view of how changes looks like, thanks for that comment.
16:18:11 sean-k-mooney gmaan: sure. im having a hard time tryign to decice what to work on today so im happy to take a look
16:19:26 gmaan sean-k-mooney: thanks, I think good timing from me :)
16:21:13 sean-k-mooney normally i have no issue with a lack fo prioirty in my life because im pretty good at setting my own but right now im like, there are too many things i could be working on so its nice to be handed a review and say can you look at this
16:21:47 gmaan ++
16:23:20 gibi sean-k-mooney: OK cool
17:14:28 opendevreview Merged openstack/nova master: libvirt: Add objects and notifications for sound model. https://review.opendev.org/c/openstack/nova/+/926126
17:30:01 sean-k-mooney gmaan: so i need ot look at the testign of the poicy change tomorrow but lookign at the code change i think that is pretty good. i left one of two nit comments but nothign that need to be changed
17:30:39 gmaan sean-k-mooney: ack, I have linked tempest change in same topic
17:30:39 sean-k-mooney gmaan: you already have tempest coverage for this too correct. ill try and look at that too tomorow to get a full picture of the testing
17:30:48 gmaan sure, thanks
17:54:45 opendevreview sean mooney proposed openstack/nova master: [DNM] testing using debian for ceph jobs https://review.opendev.org/c/openstack/nova/+/955716
#openstack-nova - 2025-07-24
00:45:04 mikal stephenfin: I can't find an example in the nova codebase of a unit test which tests parsing for a non-string flavor extra spec. I'll do one up, but I think its a thing we've not deeply tested in the past unless every other extra spec is always a string or something.

Earlier   Later