Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2026-01-26
16:21:15 fwiesel s/evenlet/eventlet/
16:21:21 sean-k-mooney fwiesel: not yet
16:21:22 gibi fwiesel: eventually yes
16:21:31 sean-k-mooney maybe next cycle
16:21:36 gibi but not yet for nova-compute
16:21:53 gibi n-sch, n-api already runs without eventlet by default on master
16:21:55 fwiesel Okay, then I won't add it yet. I'll hack something that it runs just on this PR.
16:22:30 sean-k-mooney fwiesel: you could perhaps add a perodic with it once we get it passing
16:22:50 sean-k-mooney i.e. run it with it weekly
16:23:34 fwiesel sean-k-mooney: Sure, should be doable... And report it here in the meeting?
16:23:44 sean-k-mooney that was what i was thinking ya
16:23:57 fwiesel Okay, will do.
16:24:16 gibi thanks
16:25:02 fwiesel That would be all from my side.
16:25:15 Uggla thx fwiesel
16:25:42 Uggla #topic Gibi's news about eventlet removal
16:25:46 gibi o/
16:26:57 gibi So last week we discovered that the logic how the libvirt driver delays the STOPPED libvirt event to avoid VM state flip-flop in the DB during hard reboot does not work with the current native threading proposal.
16:27:13 gibi details are here https://review.opendev.org/c/openstack/nova/+/974445
16:27:32 gibi I have a solution that probably works but not fully tested yet
16:27:39 gibi this will be my focus during the week
16:28:29 gibi I have no other news
16:28:38 gibi I will check the vmware results as well and report back
16:28:48 gibi that is it from me
16:30:07 Uggla thanks gibi
16:31:52 Uggla #topic Open discussion
16:32:03 Uggla (erlon): https://bugs.launchpad.net/nova/+bug/1452641 Bug 1452641: Static Ceph mon IP addresses in connection_info can prevent VM startup
16:32:29 Uggla Not sure erlon is available.
16:33:43 Uggla seems not
16:33:56 Uggla Anything else you'd like to discuss ?
16:35:01 sean-k-mooney i have not looked at the bug but i assume this is related to the fact that if you chagne the mon ips it breaks the vms becuase the bdms are out of sync
16:35:18 sean-k-mooney we have already consier that to be operator error in the past rather then a bug
16:35:34 tkajinam sean-k-mooney, yeah I think that's a kind of long standing limitation
16:35:43 sean-k-mooney im not sayign we cant make it better just aht we also want to be able to start a vm if cidner if offline
16:35:49 tkajinam and vms should be migrated/shelved to get a new ips passed down from cinder
16:35:58 sean-k-mooney so we should not just rebind the voluem on every boot
16:36:22 sean-k-mooney tkajinam: ya that the normal workaround do a move op to update the connection info
16:36:32 sean-k-mooney we do have a nova manage command too
16:37:33 tkajinam I guess an annoying point is that we have to first make cinder to expose up-to-date ips in attachment records
16:37:38 tkajinam so that nova can pull the updated info
16:37:51 tkajinam but I can be wrong because I looked into it some time ago
16:38:05 tkajinam anyway I think we need erlon to understand what needs to be discussed further
16:38:34 sean-k-mooney well its also an issue for nova provisioned sotrage to a degree althogh if you update the cpeh.conf i think a hard reboot is enogh in that case
16:38:48 tkajinam ah yes. that's true
16:39:42 Uggla tkajinam, sean-k-mooney at least that first info that erlon look at.
16:39:54 tkajinam yup :-)
16:40:14 Uggla moving on ?
16:40:28 Uggla or you'd like to talk about something else ?
16:41:10 Uggla #topic Bug scrubbing
16:41:15 Uggla #info up to 204
16:41:20 Uggla #link: https://etherpad.opendev.org/p/nova-bug-triage-roster
16:41:27 Uggla #info: I will assign a new round of bug tomorrow or Wednesday.
16:41:52 Uggla In the mean time does someone need to discuss about a specific bug ?
16:42:12 Uggla *needs
16:43:03 gibi sorry I lost most of the context on the previously assigned bugs
16:43:29 Uggla yep sorry about that.
16:43:48 gibi I think https://bugs.launchpad.net/nova/+bug/2115726 needs discussion
16:45:04 gibi right now if it is not a doc bug but a code bug then it is incomplete in a sense that I'm not sure what is the expected behavior of nova here
16:45:59 gibi I can reject [pci]device_spec={} but there is a lot of partial matches that need to be decide if valid or not like {address="0000"}
16:48:07 Uggla hum anyone has opinion on that ?
16:53:08 Uggla gibi I think we have some kinfd of checks for PCI devices ?
16:53:15 tkajinam I got the impression that we may start with rejecting empty thing while leaving the others as these could be "valid partial match", I guess
16:53:42 gibi tkajinam: what happens that these fields are basically act as filters
16:53:53 gibi so the empty filter list means all devices
16:54:32 gibi so for me it is logical how nova behaves today
16:54:41 tkajinam hmm ok
16:54:50 gibi removing just the empty filter list case feels strange to me
16:55:06 gibi especially if we keep {address=""} doing the same thing, meaning all devcie
16:55:09 gibi device
16:56:50 gibi {managed=false} also means all devices today
16:57:15 gibi as managed is not a filter config but a result enhancing config here
16:57:37 gibi it means all matching devices get the managed flag set to false
16:57:44 Uggla yep, that's plenty of these exceptions I guess.
16:57:50 gibi yeah it is complex
16:58:01 gibi we can say 0 partial match but then we loose valid use case I feel
16:58:11 Uggla ack
16:58:27 gibi like { "vendor_id":"8086", "product_id":"10ca", "address": "0000:07:"}
16:58:33 Uggla can we improve the doc somewhere ?
16:58:43 gibi I have VFs under 07 and I want to match all
16:59:29 gibi Uggla: we can improve the doc be descrbing the current behavior but then the todays behavior becomes the expected behavior and Sean disagrees
16:59:37 gibi a/be/by/
17:00:13 Uggla if not I agree to close it at non fix, explaining that it is not "ideal" but hard to improve.
17:01:00 tkajinam maybe we could document the current behavior first and then discuss how it can be improved separately
17:01:28 gibi As other cores I silent I would say lets move on. It is not that important
17:01:50 gibi s/other cores I silent/ other cores are silent/
17:01:59 Uggla ok, we can discuss this later as well in async mode.
17:02:05 gibi sure
17:02:14 gibi my question is in the last comment of the bug
17:02:31 Uggla because we are on the top of the hour, time to close.
17:02:43 Uggla Thanks for joining this meeting. Have a nice day/evening and see you next week.
17:02:48 gibi thanks Uggla
17:02:51 Uggla #endmeeting
17:02:51 opendevmeet Meeting ended Mon Jan 26 17:02:51 2026 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)
17:02:51 opendevmeet Minutes: https://meetings.opendev.org/meetings/nova/2026/nova.2026-01-26-16.01.html
17:02:51 opendevmeet Minutes (text): https://meetings.opendev.org/meetings/nova/2026/nova.2026-01-26-16.01.txt
17:02:51 opendevmeet Log: https://meetings.opendev.org/meetings/nova/2026/nova.2026-01-26-16.01.log.html
17:03:15 elodilles thanks o/
17:15:50 opendevreview Merged openstack/nova stable/2024.2: libvirt: Skip unsupported firmware types https://review.opendev.org/c/openstack/nova/+/974409
17:20:44 opendevreview Nisha Brahmankar proposed openstack/nova master: Fix exception handling in nova cache https://review.opendev.org/c/openstack/nova/+/974695
19:12:59 opendevreview Balazs Gibizer proposed openstack/nova master: Libvirt event handling without eventlet https://review.opendev.org/c/openstack/nova/+/965949

Earlier   Later