Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2026-03-30
16:58:39 sean-k-mooney if i cant tell that after 90s of lookign at the spec it means you need to motivate the problem statemenadn user case better in the spec
16:58:41 womax_ sean-k-mooney need to make a few modifications to use the feature it is implemented in glance
16:58:59 sean-k-mooney but why
16:59:09 sean-k-mooney nova does not need ot use every glance fature that is added
16:59:27 sean-k-mooney you need to explin in the spec what benift this ahs to users and adminstiartor of nova
16:59:38 womax_ oh ok in this sense
16:59:57 womax_ to be short, dessaturate glance traffic by bypassing it when download image
17:00:09 womax_ I'll add it
17:00:36 sean-k-mooney yep so in the specific case that we are using swift/s3 or ceph rados gateway
17:00:47 womax_ yes
17:01:00 sean-k-mooney if glacne implenmt that feature we coudl use a direct temp url to download the iamge instead of streaming via glance
17:01:08 womax_ yes exactly
17:01:09 sean-k-mooney but it woudl be good to explain that up front
17:01:40 sean-k-mooney anyway we are at time but it woudl also be good to capture how nova woudl knwo to use this in the spec
17:02:00 sean-k-mooney i.e. if we have a glance multi store env how do we know which stores supprot it and which do nto
17:02:13 sean-k-mooney ectra so that the nova spec is mostly self contaiened
17:02:24 Uggla yep we need to close.
17:02:35 womax_ ok i'll think more about it
17:02:40 Uggla Discussion can be done later outside of the meeting.
17:02:41 womax_ and we can discuss later
17:03:14 Uggla Thanks for joining this meeting. Have a nice day/evening and see you in 2 weeks.
17:03:22 opendevmeet Log: https://meetings.opendev.org/meetings/nova/2026/nova.2026-03-30-16.00.log.html
17:03:22 opendevmeet Minutes (text): https://meetings.opendev.org/meetings/nova/2026/nova.2026-03-30-16.00.txt
17:03:22 opendevmeet Minutes: https://meetings.opendev.org/meetings/nova/2026/nova.2026-03-30-16.00.html
17:03:22 opendevmeet Meeting ended Mon Mar 30 17:03:22 2026 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)
17:03:22 Uggla #endmeeting
17:03:47 elodilles thanks o/
17:04:36 sean-k-mooney womax_: have you recived much feedback on the glance spec. is it a new proposal on that side?
17:05:12 opendevreview Merged openstack/placement stable/2024.2: [CI][stable-only] Workaround for missing pkg_resources https://review.opendev.org/c/openstack/placement/+/979642
17:05:54 womax_ sean-k-mooney i already have had feedback from glance, it is indeed a new proposal
17:06:48 sean-k-mooney my expecation is that either this new feature would return the tempurl in the same location as the current glance api path in the image locations api or there will be a new microversion and api endpoint to requst a temp url and some way for nova to discover if that is supprroted for a given glance store?
17:06:52 womax_ we had a bit of discutions around security concern, but consensus was to know if nova would be interested in such use case before having further discutions
17:07:22 womax_ feature spec can be found here https://review.opendev.org/c/openstack/glance-specs/+/970214
17:08:22 womax_ for the time i proposed a 2 new api path one for receiving tempURL (useful for ironic) and one that returns a redirection with nova in mind
17:08:56 womax_ and updating stores details to reflect if store has the capabilities and enabled
17:09:11 sean-k-mooney are these tempurls 1 time use? or time based
17:09:15 womax_ time based
17:09:28 sean-k-mooney ack im askign incase we need to rety the download
17:10:01 sean-k-mooney i think we allwo up to 3 tries today
17:10:30 womax_ retry could work, expirations time would juste had to be configured accordingly
17:10:34 womax_ to not expires too soon
17:12:09 sean-k-mooney womax_: out of interest what is your primary storage backend for images
17:12:24 womax_ s3
17:12:37 womax_ too be precise s3 compatible
17:12:47 sean-k-mooney well that what im trying to udnerstadn
17:12:56 sean-k-mooney its not ceph radows gateway right
17:13:01 womax_ no
17:13:19 sean-k-mooney its minio or some other not swrit not ceph s3 compatibale storage
17:13:34 womax_ yes
17:13:58 womax_ it's the s3 compatible storage we offered in public cloud at ovh
17:14:00 sean-k-mooney ack the reason im asking is that is not one fo the case were we can already optmize out the storage tansfer
17:14:58 sean-k-mooney womax_: to me this feels more like the type of optimisation we do for rbd where we use the image location api to provide a driect rbd url
17:15:10 sean-k-mooney but restict ti to admin/service users
17:15:14 sean-k-mooney not normal users
17:15:20 sean-k-mooney so normal user use the glance api
17:15:38 sean-k-mooney but service like ironic and nova cna get the direct backend storage url and if they supprot it use that directly
17:16:17 womax_ ok for nova i could dig around direct backend, it could solve our problems
17:17:13 sean-k-mooney so how the rbd direct download works is if you have it enabel and have the rbd credtial for that sotre we use the ceph client to do the download directly
17:17:36 sean-k-mooney skiping the streaming via the glance api when the host is using say qcow2 for its nova provisioned storage
17:17:44 sean-k-mooney that effectivly the same problem statemnt
17:18:34 womax_ yeah that would indeed be a solution to what we want, but i assume there is no current equivalent for backend like s3 / swift already done?
17:18:41 sean-k-mooney nova is suing qcow2, image is in s3 compaitbale storage have nova take an optimised path if it has been configured to supprot that adn galace can prove either a direct buket link or tempurl
17:18:58 sean-k-mooney correct that does not exist
17:19:40 sean-k-mooney you coudl do either either by provie nova the s3 credital used by glance or via the tempurl you are suggesting then with jsut a http client
17:20:04 sean-k-mooney womax_: https://specs.openstack.org/openstack/nova-specs/specs/victoria/implemented/nova-image-download-via-rbd.html
17:20:14 sean-k-mooney that was the rbd oen
17:20:33 sean-k-mooney you could add a new direct_url_scheme in glance
17:20:34 womax_ providing s3 credentials to every hosts seems a bit dubious, it seems too large to distribute credentials
17:20:39 sean-k-mooney allowed_direct_url_schemes=tempurl
17:20:54 womax_ so temp url seems a bit more appropriate
17:21:09 sean-k-mooney womax_: i mean yes an no we od it for cpeh
17:21:36 womax_ ok didn't know about it
17:22:02 sean-k-mooney there isnt a tempurl schema today
17:22:23 sean-k-mooney im just sayign this migh be a way to achive the same goal folowing the established patterens
17:23:26 womax_ and, for context i started digging tempURL as I did some changes in ironic that currently generates temp url on it own, and would have like to centralize it in glance, and then i came back to nova as a potential user
17:23:48 womax_ but yeah i will dig into how it is done for rbd and see how i could have something similar
17:24:46 Uggla I have updated the topic for H cycle ^
17:39:56 opendevreview Merged openstack/placement stable/2025.2: [CI][stable-only] Fix docs job https://review.opendev.org/c/openstack/placement/+/979685
22:41:14 opendevreview Damian Dąbrowski proposed openstack/nova master: libvirt: refresh frontend QoS on volume extend https://review.opendev.org/c/openstack/nova/+/982733
22:42:31 opendevreview Damian Dąbrowski proposed openstack/nova master: libvirt: refresh frontend QoS on volume extend https://review.opendev.org/c/openstack/nova/+/982733
#openstack-nova - 2026-03-31
02:10:42 masahito Uggla: tkajinam: late reply. thanks for the info. yes, we have our own Flamingo base sev-snp support code and use it in our production. we're preparing the spec for the snp support now. once it's ready to commit, hopefully before the ptg, we will push it to the gerrit.
07:56:19 opendevreview Matt Crees proposed openstack/nova master: Don't drop pci_dev network caps when VF is in use https://review.opendev.org/c/openstack/nova/+/982660
12:21:27 opendevreview Kamil Sambor proposed openstack/nova master: Replace eventlet with standard threading in novncproxy https://review.opendev.org/c/openstack/nova/+/976089
16:09:18 opendevreview Juan Larriba proposed openstack/nova master: [WIP] Add OpenTelemetry tracing middleware for Nova https://review.opendev.org/c/openstack/nova/+/982871
16:22:13 gibi Uggla: tkajinam: I'm about to load context on the SEV refactor. I'm wondering if all the folks who are signing up to provider extra feature top of that are actually read the refactor or not
16:22:33 gibi they will be the prime users of the new interfaces so their oppinion probably matters
16:22:53 gibi I see tkajinam's comments in the series but I'm not sure who are the others I should ping
16:23:25 Uggla gibi, I know Anton (TDX) will rely on this. Hence the reason why I'm a bit pushy on that topic.
16:24:00 gibi do we have an IRC nick for Anton?
16:24:15 Uggla yes let me check
16:25:12 Uggla gibi, the nick is "antia"
16:25:12 gibi So Anton - TDX, but who is SEV-SNP, and who is armCC?
16:26:16 Uggla SEV-SNP are tkajinam and masahito
16:26:37 gibi OK
16:27:45 gibi taketani is for armCC then?
16:29:31 Uggla r-taketn is the author and armCC
16:30:20 gibi cool then I only need to involv masahito and antoni to the refactor series as r-taketn is the author of the refactor and tkajinam already saw the series
16:31:01 Uggla Note : gibi, Anton already has a spec https://review.opendev.org/c/openstack/nova-specs/+/979608
16:33:22 Uggla gibi, also Markus Hentsch might be interested. Not sure about the irc nick. He reported on the ML that he has a downstream implementation of SNP.
16:33:26 gibi OK I pinged them in the review https://review.opendev.org/c/openstack/nova-specs/+/979608

Earlier   Later