Thursday, 2026-01-08

*** mhen_ is now known as mhen03:02
opendevreviewOpenStack Proposal Bot proposed openstack/glance master: Imported Translations from Zanata  https://review.opendev.org/c/openstack/glance/+/97261703:04
opendevreviewOpenStack Proposal Bot proposed openstack/glance_store master: Imported Translations from Zanata  https://review.opendev.org/c/openstack/glance_store/+/97262003:28
opendevreviewAbhishek Kekane proposed openstack/glance-specs master: Enhanced filesystem driver with timeout protection  https://review.opendev.org/c/openstack/glance-specs/+/96257005:43
opendevreviewAbhishek Kekane proposed openstack/glance master: Migrate functional tests to SynchronousAPIBase  https://review.opendev.org/c/openstack/glance/+/97264309:07
opendevreviewMathieu Tortuyaux proposed openstack/glance master: glance/utils: override urllib 'User-agent' header  https://review.opendev.org/c/openstack/glance/+/97250710:50
opendevreviewFernando Ferraz proposed openstack/glance_store master: Cinder: Fix race conditions on mount/umount operations  https://review.opendev.org/c/openstack/glance_store/+/96762612:56
opendevreviewMathieu Tortuyaux proposed openstack/glance master: glance/utils: override urllib 'User-agent' header  https://review.opendev.org/c/openstack/glance/+/97250713:15
opendevreviewAbhishek Kekane proposed openstack/glance master: DNM Test v3applicationcredentiali without keystone fix  https://review.opendev.org/c/openstack/glance/+/97267713:17
croelandt#startmeeting glance14:00
opendevmeetMeeting started Thu Jan  8 14:00:47 2026 UTC and is due to finish in 60 minutes.  The chair is croelandt. 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
croelandtHappy New Year!14:00
croelandt#topic roll call14:00
abhishekko/ 14:00
mheno/14:00
abhishekkHappy New Year!!14:01
croelandtI think dansmith won't be here, and rosmaita is probably in another meeting14:01
pdeore__Happy New Year ! \o/14:01
*** pdeore__ is now known as pdeore14:01
croelandtoh and pdeore__ is back!14:01
croelandt#topic Release/periodic job updates14:01
croelandtso openstack-tox-py312-with-oslo-master is failing14:02
rosmaitaindeed i am, happy new year, and welcome back pdeore__14:02
croelandtI see some changes in olso.concurrency that mght be related to this, so I'm going to try and reproduce it locally14:02
pdeoreThanks :)14:02
croelandt#topic Important Zuul patches14:02
croelandttkajinam sent us a lot of patches 14:03
abhishekkDo you have link for openstack-tox-py312-with-oslo-master failures?14:03
croelandtoh yeah14:03
croelandt#link https://zuul.opendev.org/t/openstack/build/b4e16ec722e748de826f1e1a3a79cddc/logs14:03
abhishekkthanks14:03
croelandtbut yeah I see a commit changing args to execute() for prlimits14:03
croelandtso anyway, zuul patches14:03
croelandt#link https://review.opendev.org/c/openstack/glance/+/970053/14:04
croelandt#link https://review.opendev.org/c/openstack/glance/+/970054/ 14:04
croelandt#link https://review.opendev.org/c/openstack/glance/+/970059/14:04
croelandt#link https://review.opendev.org/c/openstack/glance/+/970060/ 14:04
croelandtI will probably merge these sooner or later if there are no objections14:04
croelandtthese patches keep us up to date with recent versions of Python14:04
abhishekk+114:04
croelandtremove old jobs that are not that useful anymore14:05
croelandtetc.14:05
abhishekkwait14:05
croelandt#topic M3 will release today14:05
croelandtok go on :)14:05
abhishekkI think we need fix for oslo failure else if we merge this https://review.opendev.org/c/openstack/glance/+/970053 then our gate might be broken?14:05
croelandtwhy is that?14:06
croelandti think the issue comes from oslo master14:07
croelandtoh yeah14:07
abhishekkits not failing on the patch because we don't have py312 unit in our check jobs?14:07
croelandtok there is an oslo-tips here :)14:07
croelandtoh taht is interesting, it is not failing with oslo master14:07
croelandthm14:07
croelandtok I'll hold off before merging these then14:07
abhishekkack14:07
croelandtfix the periodic job issue first 14:07
abhishekkyep14:08
croelandt#topic M3 will release today14:08
croelandt#link https://review.opendev.org/c/openstack/releases/+/97258414:08
croelandtI might also release M2 for glance_store since it has one bugfix14:08
abhishekkM2? you mean?14:08
abhishekkin topic14:09
croelandtyeah M2 sorry14:09
* croelandt cannot type on a keyboard14:09
croelandtthe keys are all so close to one another!14:09
abhishekknp :)14:10
croelandt#topic Open Discussion14:11
croelandtOK, anything else for today?14:11
mhensmall update on image encryption14:11
mhenbefore the holidays dansmith left an extensive review on the glance patchset14:11
mhenwhich I addressed as best as I could14:12
mhenalso a few days ago there was a review on the scenario tests for the image encryption which are in the barbican-tempest-plugin14:12
mhenI fixed all the things mentioned in the tests14:12
croelandtdo you need more reviews?14:13
abhishekkI have updated the spec which we discussed yesterday, https://review.opendev.org/c/openstack/glance-specs/+/96257014:15
mhencroelandt: would be great as there are no applicable votes on the current iteration yet14:16
croelandtmhen: ok14:17
croelandtabhishekk: I'll take another look at the FS one, and also the lite specs14:17
abhishekkcool, thank you14:17
croelandtI think that's all for today?14:17
croelandtThanks for joining!14:18
croelandt#endmeeting14:18
opendevmeetMeeting ended Thu Jan  8 14:18:29 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.2026-01-08-14.00.html14:18
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/glance/2026/glance.2026-01-08-14.00.txt14:18
opendevmeetLog:            https://meetings.opendev.org/meetings/glance/2026/glance.2026-01-08-14.00.log.html14:18
tkajinamhm seems like https://review.opendev.org/c/openstack/oslo.concurrency/+/967408 is causing that test failure14:20
tkajinamwhich indicates it affects oslo-master job regardless of base python version14:20
croelandtso why are our checks not failing? :)14:21
tkajinambecause tips job is not run in check/gate14:22
tkajinamit's run only in experimental and periodic14:22
* tkajinam is re-reading the job definition14:22
croelandtwe do have something with oslo master though, don't we?14:22
croelandtoh no -tips is only for periodic?14:23
croelandtok ok14:23
tkajinamhttps://opendev.org/openstack/glance/src/branch/master/.zuul.yaml#L484-L48514:23
tkajinamso updating tips job definition should have no effect14:23
tkajinamjust fyi14:24
opendevreviewTakashi Kajinami proposed openstack/glance master: Avoid asserting internal processutils.execute call  https://review.opendev.org/c/openstack/glance/+/97271314:34
opendevreviewTakashi Kajinami proposed openstack/glance master: DNM: Testing oslo tips job  https://review.opendev.org/c/openstack/glance/+/97271414:39
opendevreviewTakashi Kajinami proposed openstack/glance master: DNM: Testing oslo tips job  https://review.opendev.org/c/openstack/glance/+/97271414:40
dansmithabhishekk: around?14:52
abhishekkdansmith: yes14:52
dansmithabhishekk: I meant to ask yesterday... "get() - wraps file open() and initial read" I assumed means "open the file and read the first chunk, if that works, assume we can read the rest of the file in the request/main thread ?14:53
abhishekkyes14:55
dansmithokay so if the mount goes away while we're reading, we'll still hang without recovery14:56
abhishekkI need to verify whether to wrap entire get call or not, but I assumed if file is opened then it will be accessible14:56
abhishekkOK, i will rephrase it to cover entire get call14:57
dansmithwell, hang on14:57
dansmithwhat's your plan for wrapping the whole call?14:57
abhishekksomething like threadpool.execute(call) ?14:59
dansmithwell, what I mean is.. you can't read the whole image into memory at once and then return it from the thread15:00
abhishekkif timeout is greater than 0 then ^^^ else process as it is normmal flow15:01
opendevreviewTakashi Kajinami proposed openstack/glance master: DNM: Testing oslo tips job  https://review.opendev.org/c/openstack/glance/+/97271415:01
abhishekkright15:01
dansmithso I think either you do what you have here, which is "test" the first read of a chunk and then let the main thread do the rest,15:02
dansmithor you need to create an object that we can read from that delegates the reads to the thread and queues the blocks so the other thread can get/return them15:02
dansmithright?15:02
abhishekkI didn't undestood other thread can get/return them15:04
dansmithglance_store returns a generator that produces the data chunks that the request thread iterates and itself yields or writes to the response right?15:06
abhishekkyes15:07
dansmithhttps://github.com/openstack/glance/blob/master/glance/api/v2/image_data.py#L57615:07
abhishekkSo if we go with 1st "test" the first read of a chunk and then let the main thread do the rest we still have a limitation15:08
dansmiththis ^ is how we get data returned from downloading the image right? we have to provide an iterator/generator that the request thread (which called us) will iterate over to get the chunks15:08
abhishekkyes15:08
dansmiththe only way to make that work but have the data be read in another thread is to create a special wrapper that will read from a memory queue (with a timeout) and that the producing thread (worker) will push chunks into as there is space15:09
dansmithwell, the only way _I_ know to make it work, I should say15:09
dansmithwhich is doable I think, I just wondered what your plan was there15:10
dansmith"test read and return the rest in the main thread" still requires a wrapper I think, but is much simpler.. but yes, will still hang if the mount hangs15:10
dansmithbased on "initial read" I assumed the latter is what you were planning to do, and just take the risk that you're not in the middle of reading when the mount goes down15:11
abhishekkhonestly I was thinking https://github.com/openstack/glance_store/blob/master/glance_store/_drivers/filesystem.py#L616 wrapping this function in threadpool.execute15:11
dansmithassuming you don't consume all available api workers, one request will get stuck but others will timeout and so mostly the correct behavior15:11
abhishekkack15:12
dansmithabhishekk: that's just the open and getsize but not any read right?15:12
dansmithactually not even the open I think15:12
abhishekkno, not open in this case15:14
dansmith...okay15:15
dansmithokay well, perhaps we should just go with what you have knowing that we need to figure out how to at least do the "open and initial read"?15:16
dansmithor we could just say that it's limited to metadata operations at the moment (getsize, stat, delete) but not the actual data?15:17
abhishekkI think later makes more sense15:18
dansmithokay, let me commit my nit comments15:18
abhishekkack15:19
opendevreviewAbhishek Kekane proposed openstack/glance-specs master: Enhanced filesystem driver with timeout protection  https://review.opendev.org/c/openstack/glance-specs/+/96257015:45
abhishekkcroelandt: I am wondering why our specs does not catch trailing white spaces or change in template issues 15:52
dansmithabhishekk: looks good thanks!16:01
dansmithI can help try to come up with a creative threaded iterator for the data stuff16:01
dansmith(when the time comes)16:03
abhishekk\o/16:04
croelandtabhishekk: in the spec lites, does "AC credentials" mean "Application Credentials credentials"?16:09
croelandtabhishekk: we could run doc8 if you'd like :)16:20
croelandtdoc8 fails with symlinks though, that is annoying, had to work around that16:20
croelandtnope, we have broken symlinks, nice16:39
JayFHow difficult is it to get a new image property added? I'm looking at an Ironic feature which would allow us, when presented with an image with proper metadata, to boot it as a ramdisk instead of as a disk image. That requires Ironic reconfiguration. Triggering that reconfiguration based on some property of the image seemed ideal. (see https://bugs.launchpad.net/ironic/+bug/2130646 for a similar, non-glance RFE)16:44
JayF(and given we already have driver-specific properties, e.g. vmware_*, it seems like it'd be potentially OK to add ironic_[something] to perform this effect)16:46
JayFWe don't pass on full image metadata in the virt driver patcher; only really image_source16:54
JayFhttps://opendev.org/openstack/nova/src/commit/4b90fdf9af8de53fb6536d27b8fc654a0c011e2f/nova/virt/ironic/patcher.py#L6116:54
JayFWhich means we must be roundtripping glance ourself for metadata16:54
JayFwhich is good news for not having to modify nova (if we're OK to fudge the instance_info ourselves at deploy time)16:54
JayFoh, ECHAN, those last 4 lines were for #openstack-ironic16:55
abhishek_croelandt: yeah :/16:55
opendevreviewCyril Roelandt proposed openstack/glance-specs master: Fix symlinks for Xena templates  https://review.opendev.org/c/openstack/glance-specs/+/97275916:56
opendevreviewElod Illes proposed openstack/glance_store master: DNM: gate health test  https://review.opendev.org/c/openstack/glance_store/+/97276016:57
croelandtabhishek_: "yeah we have broken symlinks" or "yeah ac credentials means application credentials credentials"?17:00
abhishek_for the later :D17:00
dansmithJayF: image properties are mostly unconstrained in glance.. are you actually asking about nova?17:03
JayFI was reading https://docs.openstack.org/glance/latest/admin/useful-image-properties.html and ensuring that I didn't need Glance approval or spec to create/use one :)17:04
JayFtrying to be a good citizen and not just backdoor something Ironic into the world17:04
croelandtabhishek_: ok that's weird but I'll pretend I did not see it :D17:05
dansmithJayF: glance has/holds image property metadefs that tries to be the source of truth for things like horizon (only, AFAIK)17:05
abhishek_me neither17:05
JayFnice, so I would expect I can just create/use the property, and get a small glance doc update merged to useful-image-properties (assuming folks agree it's useful) 17:05
dansmithJayF: but my point is, if you want something plumbed through image properties to virt/ironic then you'll need it to be in nova17:06
JayFIronic does it's own roundtrip to Glance already for properties17:06
JayFthe only thing Nova sends over is the image uuid17:06
dansmithJayF: okay, so then not nova (which is good because we'll want it to be generic if possible)17:06
JayF++ yep exactly this is the ideal case, all the complexity is inside ironic and the user knob is (after admin has configured Ironic) "upload image with ironic_ramdisk=True in properties"17:07
opendevreviewMerged openstack/glance master: Functional tests migration - test_gzip_middleware.py  https://review.opendev.org/c/openstack/glance/+/96820317:51
opendevreviewMerged openstack/glance master: Avoid asserting internal processutils.execute call  https://review.opendev.org/c/openstack/glance/+/97271317:51
dansmithcroelandt: I assume you're going to +W this tomorrow right? https://review.opendev.org/c/openstack/glance-specs/+/96257022:12
-opendevstatus- NOTICE: Zuul will be shutdown for maintenance work. See https://lists.opendev.org/archives/list/service-announce@lists.opendev.org/thread/WBBLBI6ZS6FA6Q5ZMH4C2MWPL3WG3H24/ for more details.23:43

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