Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2025-03-25
16:16:47 keerthivasan86[m] Just quick question why https://bugs.launchpad.net/nova/+bug/1766387 ( why tx_queue_size is not getting implemented in vhost , any specific reason )
16:16:52 gibi just to double check
16:17:14 gibi we do not track anything for RC2
16:17:17 sean-k-mooney keerthivasan86[m]: if i recall corectly its only supproted in vhost-user but not kernel vhost
16:17:19 gibi is it corr4ect?
16:18:02 sean-k-mooney gibi: not that im aware of
16:19:04 keerthivasan86[m] sean-k-mooney: As part of vhost-net able to see rx_queue_size is implemented at vm level as well. It works for vhost-net as well
16:19:10 Uggla gibi, not yet. Do you think about FUP for vfio pci migration ?
16:19:43 sean-k-mooney keerthivasan86[m]: when it was implented there was a specific reaosn why we didnt enabled it
16:20:12 gibi Uggla: it is going to land on master probably today. I'm not advocating for pusing for an RC2 tomorrow just for that single thing. (it is not an Epoxy regression)
16:20:25 sean-k-mooney keerthivasan86[m]: that may haved chnted but the current code was becasue of how qemu/libivrt worked at the time we implemtned https://specs.openstack.org/openstack/nova-specs/specs/rocky/implemented/libvirt-virtio-set-queue-sizes.html
16:21:30 keerthivasan86[m] sean-k-mooney: if you are ok , can i work on to implement tx_queue_size for vhost-net ?
16:22:11 sean-k-mooney sure but we need to know when it started to be supported, and we need to take accound of upgrades
16:22:13 dansmith the only PCI FUP that should be backported is doc stuff right?
16:22:19 dansmith Uggla: gibi ^
16:22:31 dansmith the test and cleanliness things should not be backported, IMHO, but definitely not hold up rc2
16:22:44 Uggla gibi, this is what I have in mind as well. Waiting for GA as this is the only thing.
16:22:48 sean-k-mooney keerthivasan86[m]: so im not sure fi it needs a spec or if a specless blueprint woudl be enough.
16:23:06 sean-k-mooney dansmith: the doc is the one that really should be
16:23:15 sean-k-mooney dansmith: some of the others are nice to have but not required
16:23:39 dansmith backporting test cleanliness doesn't make sense to me unless it's needed for an actual fix
16:23:57 gibi dansmith: for RC2 prespective only https://review.opendev.org/c/openstack/nova/+/944105/5 in interesting
16:24:04 sean-k-mooney right the doc i could see for an RC2 but at this point i woudl just wait and do that after the release
16:24:27 sean-k-mooney gibi: oh right to make numa work
16:24:40 gibi dansmith: but as I wrote above it is not an E regression and will only land today
16:24:43 dansmith gibi: ah I missed that one.. that's not really a regression i t seems, but..
16:24:49 dansmith ack yeah okay
16:25:00 gibi yeah I would not spin an RC2 just for thish
16:25:10 gibi but if there are other things forcing an RC2 ...
16:25:11 sean-k-mooney so we can do a stable release basifcally right after the actual release
16:25:17 dansmith right
16:25:23 Uggla +1
16:25:27 sean-k-mooney so i think RC2 is overkill for any of those
16:25:27 gibi yeah doing stable backports is the way to go if nothing else forces RC2
16:25:44 sean-k-mooney sounds like we are all in violent agreement
16:26:38 gibi yepp
16:26:40 bauzas yup
16:26:46 bauzas (sorry, meeting was over)
16:27:01 Uggla bauzas, no still in the meeting
16:27:17 bauzas as a reminder, any new release candidate impacts our distros
16:27:34 bauzas so we definitely only need to provide a RC2 if this is a regression
16:27:47 elodilles (with my release core hat on) anyway, let me know if an RC2 is needed (or not needed) as this week is when we can release a final RC as i said o:)
16:28:53 elodilles (i'll generate final RC patches (where there are merged and unreleased stuff on stable/2025.1) somewhere around 2nd half of the week fwiw)
16:29:18 bauzas ack thanks
16:29:22 elodilles np
16:30:36 dansmith are we done?
16:30:56 Uggla keerthivasan86[m], are the answer of sean ok for you ?
16:31:29 Uggla any other topic ?
16:31:56 masahito question for the ptg topic selection and timeslot
16:32:29 Uggla masahito, the current topic are in the etherpad above.
16:32:49 masahito thank. ah, yes. I added some.
16:33:06 Uggla I think not all of them are available. I know new ones will come.
16:33:22 Uggla Then we will organize the timeslot to discuss all of them.
16:33:29 sean-k-mooney Uggla: i assume masahito wanted to ask about when there topics will be scheduled
16:33:39 sean-k-mooney or perhaps suggest when tehy are aviable
16:33:51 masahito my question is the ptg time slot is limited so i don't think the ptg can't cover all topics. so i want to know when and how it's decided.
16:34:13 sean-k-mooney masahito: we dont have many topics currently
16:34:20 masahito s/ptg can't/ptg can/
16:34:22 sean-k-mooney i woudl be surrprised if we didnt cover them all
16:34:52 Uggla I think for the time slot it will be almost ok next week.
16:35:04 keerthivasan86[m] yes Uggla As sean-k-mooney suggested specless blueprint or fresh blueprint to implement tx_queue_size
16:35:27 opendevreview sean mooney proposed openstack/placement master: improve test logging and replace psycopg2 with psycopg2-binary https://review.opendev.org/c/openstack/placement/+/945487
16:35:35 sean-k-mooney dansmith: ^
16:35:50 keerthivasan86[m] can i create fresh blueprint to implement tx_queue_size for vhost-net feature ?
16:35:55 dansmith sean-k-mooney: ack
16:36:00 Uggla Masahito, we’ll try to accommodate your schedule so you can join.
16:36:13 masahito i got it for the next week meeting.
16:36:54 Uggla any other topic ?
16:36:55 masahito thanks. I prefer 14:00UTC or later slot because of family reason as it's remote session :)
16:37:28 Uggla masahito, sure I'll keep that in mind.
16:38:31 Uggla looks we can close the meeting
16:39:22 Uggla thanks all
16:39:27 Uggla #endmeeting
16:39:27 opendevmeet Meeting ended Tue Mar 25 16:39:27 2025 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)
16:39:27 opendevmeet Minutes: https://meetings.opendev.org/meetings/nova/2025/nova.2025-03-25-16.00.html
16:39:27 opendevmeet Minutes (text): https://meetings.opendev.org/meetings/nova/2025/nova.2025-03-25-16.00.txt
16:39:27 opendevmeet Log: https://meetings.opendev.org/meetings/nova/2025/nova.2025-03-25-16.00.log.html
16:39:41 elodilles thanks Uggla o/
16:39:48 masahito thank you o/
16:39:50 Uggla elodilles, ty
16:42:10 bauzas Uggla: bravo for your first meeting :)
16:44:28 Uggla bauzas, ty
16:46:23 gibi Uggla: thanks
17:33:24 simondodsley elodilles: not sure if you saw the conversation yesterday, but are the stable cores open to backport these 3 XS patches to resolve the SCSI page 83 passthrough issue: https://review.opendev.org/q/topic:%22lun-device-serial%22
18:10:10 gboutry Hello, I'm currently splitting my hair around https://bugs.launchpad.net/ubuntu/+source/nova/+bug/2103413
18:10:13 gboutry I'm currently running nova 31.0.0RC1 tag on python3.13, and I can't create a new instance.
18:10:17 gboutry When I'm trying to, I'm getting the error described in the ticket.
18:10:22 gboutry At around nova/nova/network/neutron.py:1233 (allocate_for_instance), admin_client.httpclient.dict seems to get totally wiped
18:10:26 gboutry going from {'session': <keystoneauth1.session.Session object at 0x7e1a2c601950>,... 'endpoint_override': None, ...} to {} (empty dict)
18:10:31 gboutry Thus failing with the error in the lp bug. the line number it happens varies but always in the function, so this looks like a finalizer going off the rails?
18:10:35 gboutry This seems python 3.13 related, but I'm not sure
18:11:30 gboutry I would appreciate any advice :)
18:13:16 sean-k-mooney gboutry: 3.13 is not offically supproted yet
18:13:24 sean-k-mooney it might work but its not actully tested
18:13:50 sean-k-mooney https://github.com/openstack/governance/blob/master/reference/runtimes/2025.1.rst is the supproted python verisons
18:14:26 sean-k-mooney gboutry: initall testing for 3.13 will be part of 2025.2 and it will only become requried in 2026.1
18:16:22 gboutry sean-k-mooney: thanks for the info, Epoxy is targeted to be released in Ubuntu Plucky (25.04) which is using Python 3.13, we'll most likely need to carry a patch downstream if I can find the root cause of this issue
18:16:32 sean-k-mooney gboutry: i wonder if this could be eventlet related. teh error is coming form keystone authg via neutron client
18:17:26 gboutry I am getting some eventlet assertion errors in the background: https://pastebin.ubuntu.com/p/WhKxDrvQHt/
18:18:30 gboutry https://github.com/eventlet/eventlet/commit/4d48d10b990910cc87ec6bbeeaea63b295fe314d python 3.13 has been added in this commit

Earlier   Later