Thursday, 2026-02-19

*** tkajinam_ is now known as tkajinam10:13
*** rosmaita1 is now known as rosmaita12:56
opendevreviewSebastian Krott proposed openstack/glance_store stable/2025.2: Fix final chunk's content_length for Swift image uploads.  https://review.opendev.org/c/openstack/glance_store/+/97731913:28
opendevreviewSebastian Krott proposed openstack/glance_store stable/2025.2: Fix final chunk's content_length for Swift uploads  https://review.opendev.org/c/openstack/glance_store/+/97731913:29
croeland1#startmeeting glance14:00
opendevmeetMeeting started Thu Feb 19 14:00:07 2026 UTC and is due to finish in 60 minutes.  The chair is croeland1. Information about MeetBot at http://wiki.debian.org/MeetBot.14:00
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.14:00
opendevmeetThe meeting name has been set to 'glance'14:00
croeland1#link https://etherpad.openstack.org/p/glance-team-meeting-agenda14:00
croeland1#topic roll call14:00
croeland1o/14:00
pdeoreo/14:01
croeland1is it only us? :)14:01
pdeoreI thought it's canceled, did see anything in the agenda :P14:01
pdeoreyeah14:01
rajiv_hi14:01
croeland1pdeore: yeah i'm lazy and write the agenda at the very last minute14:01
croeland1no rajiv is here!14:01
croeland1I think rosmaita is in another call14:01
croeland1there is not muc hto the agenda anyway14:02
pdeoreyeha14:02
croeland1The main thing this week was merging a ton of store patches and releasing M214:02
croeland1so congrats to us on that \o/14:02
croeland1#topic Release/periodic job updates14:02
pdeore\o/14:02
croeland1All good \o/14:02
croeland1#topic  Important stable patches14:02
croeland1so we have a few stable patches in the queue, ready to be merged in Glance14:02
croeland1but there is a failure in the CI that will be resovled by14:03
croeland1#link https://review.opendev.org/c/openstack/requirements/+/976227 14:03
croeland1I'm keeping an eye on this and once it merges I'll recheck other patches14:03
croeland1#topic eventlet patches14:03
croeland1we merged all the eventlet patches under review, we have very few uses of eventlet left, so we're almost there14:03
croeland1I'll try to talk to hberaud about this again and see what we can do14:04
croeland1#topic PTG14:04
croeland1#link https://etherpad.opendev.org/p/2026.2-ptg-glance-planning14:04
croeland1You can add topics there ^14:04
croeland1I see Abhishek has been working on this14:04
croeland1#topic Open Discussion14:04
croeland1And do we have anything else?14:04
rajiv_yes14:05
pdeorenothing from me 14:05
rajiv_i upgraded to flamingo with swift backend14:05
rajiv_the last chunk fails to update, i receive a 49914:05
croeland1Interesting14:05
rajiv_https://review.opendev.org/c/openstack/glance_store/+/977319 could this patch help fix this ? it works locally14:05
rajiv_we never had this issue in dalmatian14:06
mheno/14:06
croeland1so yes, but it should be submitted in master first14:06
rajiv_hey abhishekk, based on yesterdays issue discussion, we tested this locally and it works : https://review.opendev.org/c/openstack/glance_store/+/97731914:07
rajiv_ah ok ok, i will change the branch14:07
mhenregarding image encryption, I saw dansmith and melwitt reacting to my latest comments on the glance spec yesterday and I will review and answer as soon as possible14:07
croeland1$ git ll --no-merges origin/stable/2024.2..origin/stable/2025.2 glance_store/_drivers/swift/14:08
croeland197cbdf3 (tag: 5.1.0) Replace oslo_utils.encodeutils.exception_to_unicode14:08
croeland1d71ba18 swift: Drop support for v1/v2 auth14:08
croeland17b5817a [swift] Add size validation for image uploads14:08
croeland1rajiv_: ^ these are the patches between D (2024.2) and F (2025.2) that directly affect swift14:08
croeland1I wonder if anything here caused the issue14:08
croeland1oh yeah14:08
croeland1you mentionned 7b5817a in the commit log14:08
croeland1ok so if you can confirm that reverting 7b5817a helps, yeah, we should do something about it14:09
rajiv_yes, considering swift service never been updated in the past, we were looking into previous commits14:09
rajiv_it works locally, strange thing is the image upload works well with glancclient and fails with openstackclient14:09
croeland1ouch this is interesting14:10
croeland1so is the store at fault or the client? :)14:10
tkajinammaybe chunk size is handled differently in client side14:10
* tkajinam is glad that he is not blamed for his changes, seeing the other two are his. :-P14:10
tkajinamit seems the problem is caused by chunk alignment (causing failure if the last chunk size does not match the fixed one)14:11
rajiv_i was initially confused if it was the clients, considering a similar commit from Dan for openstackclient wasnt merged.14:11
rajiv_tkajinam: yes14:11
croeland1the glanceclient is going away, so...14:11
croeland1so yeah please submit your fix to master, than Abhishek and I will take a look14:11
rajiv_sure, thanks14:12
croeland1and make sure we fix your bug and fix the bug 7b5817a was supposed to fix14:12
croeland1this kind of code where we loop over chunks and handle multiple weird cases is headache-inducing14:12
* tkajinam changed the status of the bug because its fix has not yet merged to master14:13
tkajinam(fix committed indicates that the fix is merged but is not yet released14:13
croeland1thaks14:13
croeland1mhen: also thanks for your work on the spec14:13
croeland1anything else?14:13
croeland1well, thanks everyone for joining14:18
croeland1see you next week14:18
croeland1#endmeeting14:18
opendevmeetMeeting ended Thu Feb 19 14:18:06 2026 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)14:18
opendevmeetMinutes:        https://meetings.opendev.org/meetings/glance/2026/glance.2026-02-19-14.00.html14:18
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/glance/2026/glance.2026-02-19-14.00.txt14:18
opendevmeetLog:            https://meetings.opendev.org/meetings/glance/2026/glance.2026-02-19-14.00.log.html14:18
*** croeland1 is now known as croelandt14:18
tkajinamoh, one thing which might help you. py310 job is broken now (you may have seen POST_FAILURE) and we are waiting for new images built and deployed in CI infra14:19
* tkajinam runs away14:19
opendevreviewSebastian Krott proposed openstack/glance_store master: Fix final chunk's content_length for Swift uploads  https://review.opendev.org/c/openstack/glance_store/+/97732314:28
opendevreviewRajiv Mucheli proposed openstack/glance_store master: Fix final chunk's content_length for Swift image uploads.  https://review.opendev.org/c/openstack/glance_store/+/97732314:34
*** vhari_ is now known as vhari14:59
dansmithmhen: for a bootable LUKSv1-encrypted disk created by cinder, what does the actual disk layout look like? LUKS at the beginning of the disk15:56
dansmith?15:56
dansmithjust refreshing myself on the LUKS header format and.. it doesn't leave room for a bootloader that I know of, so surely a bootable disk needs to be LUKS only in a partition15:58
dansmithbut if cinder is wrapping the disk as seen by the guest in LUKS, that can never be directly bootable (like a real machine would) and only the _contents_ are bootable if the outer layer is being stripped - correct?15:59
mhendansmith: here is one example of Cinder's implementation (Ceph driver): https://github.com/openstack/cinder/blob/master/cinder/volume/drivers/rbd.py#L1127-L113516:04
mhenThe output file of that qemu-img command when queried via `file ...` is: "LUKS encrypted file, ver 1 [aes, xts-plain64:sha256, sha256] ..."16:05
mhenLUKS is directly at the beginning16:05
dansmithack, okay16:05
mhenThe guest never sees the LUKS stuff, only its payload16:06
dansmithyeah, I understand that part in this scheme16:06
mhenNova either uses QEMU's way to directly unwrap the LUKS during attachment (via LibVirt Secrets) in case of KVM or as a fallback, dm-crypt with /dev/mapper endpoint and passing that decrypted endpoint into the VM iirc16:07
dansmithyeah, I understand16:08
dansmithI just wanted to confirm that the layout of things is sort of the opposite of what would work on a real system.. i.e. that we don't have gpt->luks (in partition 1)->gpt16:09
mhenright, usually you would not use LUKS on the whole raw block device in real systems, but have partitions, raid/LVM etc. before that first16:10
dansmithyou would/could on a data-only volume, just not on something you expected to be bootable16:11

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!