21:00:01 #startmeeting swift 21:00:02 Meeting started Wed Apr 15 21:00:01 2020 UTC and is due to finish in 60 minutes. The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:03 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:05 The meeting name has been set to 'swift' 21:00:09 who's here for the swift meeting? 21:00:13 o/ 21:00:14 o/ 21:00:18 o/ 21:00:19 hello o/ 21:00:35 o/ 21:01:18 i think tdasilva and clayg may be busy with another meeting... 21:01:32 as usual, agenda's at 21:01:35 #link https://wiki.openstack.org/wiki/Meetings/Swift 21:01:51 so oh i was planning on skipping that other meeting too 🤔 21:02:00 haha 21:02:02 #topic virtual PTG 21:02:16 quick reminder to start gathering topics 21:02:25 #link https://etherpad.opendev.org/p/swift-ptg-victoria 21:02:35 i see clayg has already started :D 21:03:11 i still need to catch up on the mailing list a bit to see what kind of logistical issues i need to pay attention to 21:03:52 #topic tc election 21:04:05 just highlighting that the election results are in 21:04:12 #link http://lists.openstack.org/pipermail/openstack-discuss/2020-April/014135.html 21:04:46 congrats to the new (and returning) tc members! 21:05:04 #topic swiftclient 21:05:12 the gate's fixed! 21:05:24 🎉 21:06:12 also, i threw up a bunch of patches -- https://review.opendev.org/#/c/719686/ seems like it'd probably warrant further discussion 21:06:12 patch 719686 - python-swiftclient - Rip out SERVICENET support - 2 patch sets 21:06:49 I want to know how you came up with the idea for that patch? 21:07:01 servicenet was a dumb idea for AGES - why rip it out now 😆 21:07:54 (for context, i was looking at trying to replace all of our auth stuff in client.py with some plumbing into keystoneauth1, and the crap at https://review.opendev.org/#/c/719686/2/swiftclient/client.py@a557 seemed to present a problem) 21:07:55 patch 719686 - python-swiftclient - Rip out SERVICENET support - 2 patch sets 21:08:48 maybe i could work around it with an authv1 plugin variant... but OTOH... less code is good 21:09:09 less code IS good - i'm on board with ripping it out 👍 21:09:39 it was super auth v1 specific - and I think getting onto the keystone plumbing would help make some other things simpler so I encourage this effrot now that I understand the context 21:10:02 clayg, i think you were the one i was most worried about objecting, so... \o/ 21:11:36 and finally, on the release front... i'd meant to get a 3.10.0 client out ahead of the ussuri cutoff, but didn't quite make it 21:11:49 but versionign support will be that much better when it comes in victoria ;-) 21:12:28 that's about all i had by way of announcements -- any questions or comments before moving on to updates? 21:13:39 #topic waterfall-ec 21:13:45 clayg, how's it going? 21:15:48 we'll check in on him again later ;-) 21:15:53 not working on it right now 21:16:10 still thinking about it tho! I might have to put "fixing swift_zipkin" in front of going back to it 21:16:16 👍 21:16:19 I feel like the visualizations will be useful 21:16:26 #topic lots of small files 21:16:41 alecuyer, rledisez: how's it going? 21:17:08 I made good progress with the first "non cleanup" patch I'd like to send, but I have been pulled to something else. I hope to return to it soon 21:17:14 So I'm afraid no further news 21:17:49 I saw you send some patches to feature/losf timburke, thx for that, i'll have a look into them 21:18:40 yeah, they're not ready yet -- i was trying to fix up the losf func test job 21:19:32 i'm guessing there was a change in the default python interpreter for the unittest job that it's inheriting from, but i'm not sure how to switch it back yet 21:20:13 of course, eventually we'll need a py3 version of it anyway... 21:20:24 idk. still thinking about how to proceed there 21:20:42 #topic cors 21:21:01 so i reworked the test harness a bunch 21:21:36 unfortunately, i seem to have broken my once-working gate job in the process, so i still need to sort that out 21:22:50 you can see some of the new style in https://review.opendev.org/#/c/533028/13/test/cors/test-object.js 21:22:50 patch 533028 - swift - Add some functional CORS tests - 13 patch sets 21:23:46 the big benefit was that it let me do request chaining, like in https://review.opendev.org/#/c/710355/12/test/cors/test-s3-obj.js@160 21:23:47 patch 710355 - swift - s3api: Allow CORS preflight requests - 12 patch sets 21:24:18 (in the previous patchsets, that test would just do the write and not clean up after itself) 21:25:19 this *also* let me do test the workflow my users are actually interested in, which was initiating, uploading, and completing S3 MPUs 21:25:58 Nice 21:26:39 that needed a bunch more work than i was originally expecting to get it to work; p 720098 had to touch most of the functions in s3api's multipart upload controller, plus slo 21:26:40 https://review.opendev.org/#/c/720098/ - swift - WIP: s3api: Allow MPUs via CORS requests - 2 patch sets 21:27:24 makes me wonder if p 528106 might be ther better way to go 21:27:25 https://review.opendev.org/#/c/528106/ - swift - Move CORS to middleware. - 5 patch sets 21:29:31 anyway, long story short: there's going to be some more work needed on it, though i still think those preliminary patches will be useful even before we get to the point that my users are happy ;-) 21:29:59 #topic open discussion 21:30:08 anything else we should bring up today? 21:30:39 ah, can i ask quick question on py3 work? 21:30:46 yeah! 21:31:32 so I think Swift still wants to keep py2 on Ussuri (ref: https://governance.openstack.org/tc/goals/selected/ussuri/drop-py27.html#projects-keeping-py2-7-support) but I'm not sure if there are known issues yet. 21:32:01 known issues with py2 or py3? 21:32:07 so if some links (e.g. etherpad) to know issues let me know. 21:32:09 ah, py3 21:32:21 the context is from dropping py2 support for storlets 21:32:25 osa brought https://bugs.launchpad.net/swift/+bug/1872553 to my attention just the other day 21:32:25 Launchpad bug 1872553 in OpenStack Object Storage (swift) "fallocate_reserve is still broken on py3" [Undecided,New] 21:32:57 openstack infra team actively pushed some patches to drop py2-compatible code and I'd like to check the status of Swift itself. 21:33:19 drop py2-compatible code for Storlets 21:33:19 yeah, makes sense. fwiw, i still don't have any clusters on py3 yet beyond my dev saio 21:34:03 oic https://bugs.launchpad.net/swift/+bug/1872553 is absolutely one of big problems. 21:34:03 Launchpad bug 1872553 in OpenStack Object Storage (swift) "fallocate_reserve is still broken on py3" [Undecided,New] 21:34:04 i really want to change that, but... it's hard to squeeze in 21:35:23 https://bugs.launchpad.net/swift/+bug/1863053 was another fairly recent one that caught me by surprise, but it has a fix on master and backported to train 21:35:23 Launchpad bug 1863053 in OpenStack Object Storage (swift) "Swift on py3 doesn't like message/rfc822 data" [High,Fix released] - Assigned to Tim Burke (1-tim-z) 21:35:40 good 21:37:13 there was also a formpost issue, and s3api header casing -- i should probably do a 2.23.2 with more py3 fixes as i get 2.25.0 together... 21:37:29 was sort of the communication between proxy and object fixed? I think some issues exist on MIME protocol on eventlet... 21:39:06 yeah, https://bugs.launchpad.net/swift/+bug/1840507 was fixed ahead of the initial train release 21:39:06 Launchpad bug 1840507 in OpenStack Object Storage (swift) "Mixed py2/py3 environment allows authed users to write arbitrary data to the cluster" [Undecided,Fix released] 21:39:53 nice 21:39:53 (unless there's some other issue i wasn't aware of) 21:40:57 ok. thx. anything else you remember later, please let me know. 21:41:16 and the situation on Swift upstream fits my idea. 21:41:40 we still needs more production clusters on py3 and make them stable. 21:41:58 before dropping py2. 21:42:25 kota_, do you have a feeling for how storlets users usually deploy their swift these days? i'd say that should be your main input in deciding whether to continue supporting py2 21:43:56 i don't have enough input from users but some users like hosanai was not actively upgrade the swift IIRC 21:44:06 arguably, even if rledisez and clayg were running all py3-only, if there were significant numbers of storlets users still running py2, that should be enough reason for swift to continue supporting it 21:44:41 I'm not sure he (they) wants to upgrade storlets independently but keep py2-code would be ok but the new code should be developed as py3 21:44:42 mm. fair enough -- if they don't upgrade swift, do they still upgrade storlets? 21:46:40 fortunately (or unfortunately) Storlets is now not having drastic change on codebase so we might accept to drop testing step by step. 21:47:20 I don't think they upgrade storlets yet. 21:48:47 IMHO, storlets can follow the Swift way so you don't have to think of "keeping py2 for storlets" 21:50:08 and otoh, storlets should absolutely feel free to say they don't want to keep trying to support py2 -- i think you guys can do that any time between now and when swift finally does drop py2 21:51:12 i see thanks 21:51:58 that is the reason why I ask the status. before dropping py2 completely, I'd like to check py3 works on Swift :P 21:52:28 all right, i'm gonna call it 21:52:37 thank you all for coming, and thank you for working on swift! 21:52:41 #endmeeting