21:00:29 <timburke> #startmeeting swift
21:00:29 <opendevmeet> Meeting started Wed Feb 12 21:00:29 2025 UTC and is due to finish in 60 minutes.  The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:00:29 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
21:00:29 <opendevmeet> The meeting name has been set to 'swift'
21:00:36 <timburke> who's here for the swift meeting?
21:01:17 <boosungkim[m]> Hello
21:02:47 <mattoliver> o/
21:04:06 <timburke> i remembered to update the agenda!
21:04:08 <timburke> #link https://wiki.openstack.org/wiki/Meetings/Swift
21:04:17 <timburke> first up
21:04:22 <timburke> #topic foundation news
21:04:33 <timburke> looks like OpenInfra may join the Linux Foundation!
21:04:38 <timburke> #link https://lists.openinfra.org/archives/list/foundation@lists.openinfra.org/thread/3B7OWPRXB4KD2DVX7SYYSHYYRNCKVV46/
21:05:01 <timburke> i'm not sure it'll have much impact on us, but thought it worth highlighting
21:05:47 <timburke> (especially since the second community meeting for it will be happening in just a few hours)
21:06:09 <mattoliver> yeah, that's really interesting. I wonder how it may change things.
21:06:18 <mattoliver> Oh yeah.. maybe I should listen in.
21:06:33 <timburke> next up
21:06:42 <timburke> #topic 2025.2 cycle name
21:06:55 <timburke> it's official, it'll be openstack flamingo
21:07:03 <timburke> #link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/message/IMHDQ3VVI46HHNO7BTWOTT5VH53TXZCR/
21:07:38 <mattoliver> flamingo, ok. why not :)
21:07:55 <timburke> aside from epoxy, i feel like they've been on an animal kick lately
21:08:38 <timburke> again, not much to really discuss, more of an announcement in case you missed it
21:08:52 <timburke> #topic boto3 breakage
21:09:11 <timburke> so we already knew about the aws-chunked trouble
21:09:21 <timburke> and the patch to address it has been making decent progress
21:09:27 <timburke> #link https://review.opendev.org/c/openstack/swift/+/836755
21:09:44 <timburke> but i think i forgot to mention the *other* bit of breakage!
21:10:30 <rcmgleite> key folks, are we meeting today?
21:10:44 <timburke> rcmgleite, welcome! yep
21:11:17 <timburke> recently, multi-deletes started failing against s3api because boto's no longer sending a content-md5 header (presumably because they're sending one of the checksum headers instead?)
21:11:21 <mattoliver> yeah, I'll take another look today, and maybe we can merge it soon
21:11:22 <mattoliver> maybe today?
21:11:31 <mattoliver> depending on how others feel of course
21:12:21 <timburke> this first manifested with our ceph-s3-tests job, but also would break our internal cross-compat tests once boto3 gets upgraded
21:12:41 <timburke> fortunately, we've got another fix for that already in review:
21:12:43 <timburke> #link https://review.opendev.org/c/openstack/swift/+/940172
21:13:28 <timburke> and all these s3api patches leads me to...
21:13:32 <timburke> #topic improving the maintainability of s3api
21:13:55 <timburke> (this also ties in with what you were saying, mattoliver ;-)
21:14:42 <timburke> while reviewing the aws-chunked patch, acoles was frustrated by the convoluted state of the s3request.py module in particular
21:15:35 <timburke> i don't blame him; it's a bit of a mess. one that we keep just throwing more hacks into as issues are found or new features requested
21:16:08 <timburke> i'm not entirely sure how to go about making it better, though
21:16:55 <timburke> a while back, i tried to pull out just the auth-y bits with https://review.opendev.org/c/openstack/swift/+/613452 -- maybe i should make another attempt
21:19:15 <timburke> i also tried to pull out some of the error-response handling in some earlier iteration of https://review.opendev.org/c/openstack/swift/+/751555 (though now it looks like it's all back in the one module again)
21:20:17 <mattoliver> Yeah, it's hard because one needs to load up much of the s3api in order to do some nice refactors and seperation.
21:20:51 <timburke> but i wanted to throw this out there as a hairy problem that we might want to face -- it's easily the largest module in s3api, and it's only getting larger
21:20:54 <mattoliver> And working on tech debt like this is super important, but also hard to justify to employers.
21:22:30 <mattoliver> thanks, we should write it down somewhere. Maybe a new catergory of the ideas page. Then we keep looking at them and maybe convince someone to help out (or if we get time)
21:24:04 <timburke> anyway, that's about all i've got this week
21:24:10 <timburke> #topic open discussion
21:24:19 <timburke> what else should we bring up?
21:25:27 <mattoliver> I've squashed 2 more things into the head of ringv2
21:26:39 <mattoliver> clayg would think it would be easier to review if we split the head patch into some structural changes but still just support v1, then all the v2 stuff to make it easier to review.. but having it sqaushed in is easier to carry (for us).
21:27:01 <mattoliver> (or maybe we just land it) ;)
21:29:07 <mattoliver> But we (nvidia) plan to carry the patch into our next downstream release.. so maybe it'll give everyone more confidence.
21:29:23 <mattoliver> or it'll go wrong, which it at least some more data :P
21:30:11 <timburke> either way, at least the patch seems unstuck -- i was disappointed that it languished so long. thanks for freshening it up, mattoliver!
21:31:24 <mattoliver> nps. That's really all I've been digging into. I have started looking at some of the expirer improvement patches, too. But really want to get ringv2 landed to fix device ids width. And then it opens doors for the rest of the chain which will make reconstructor rebalances better too!
21:31:52 <timburke> i think there's still at least one thing i ought to fix with it -- some bleeding together of GzipWriter and RingWriter
21:32:59 <mattoliver> yeah, can't hurt, but I think the ondisk format is decided, so don't mind if thats now or a follow up. eitherway, so long as it doesn't slow down the critical path :)
21:33:43 <timburke> oh! also, Callum027's policy-info in account listings patch landed!
21:34:06 <timburke> i pushed up a follow-up to address a few nits, but the bones of it looked good to me
21:34:19 <timburke> #link https://review.opendev.org/c/openstack/swift/+/941301
21:34:22 <timburke> for the follow-up
21:37:22 <mattoliver> nice! thanks Callum027 !!
21:37:51 <Callum027> No worries, thanks for reviewing it so quickly and for the feedback timburke and mattoliver
21:38:08 <timburke> 👍
21:40:22 <timburke> all right, seems like we can wrap a little early
21:40:34 <timburke> thank you all for coming, and thank you for working on swift!
21:40:38 <timburke> #endmeeting