21:00:14 <timburke> #startmeeting swift
21:00:15 <openstack> Meeting started Wed Dec  4 21:00:14 2019 UTC and is due to finish in 60 minutes.  The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:00:16 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
21:00:18 <openstack> The meeting name has been set to 'swift'
21:00:26 <timburke> who's here for the swift meeting?
21:00:51 <kota_> hi
21:01:05 <seongsoocho> o/
21:01:44 <mattoliverau> o/
21:01:52 <tdasilva> o/
21:03:03 <timburke> i just got back last night from a week away, so... i'll be interested in hearing what everyone's been up to :-)
21:03:12 <timburke> so, updates!
21:03:25 <timburke> tdasilva, clayg: how's versioning going?
21:03:32 <timburke> #topic object versioning
21:04:02 <clayg> versioning is GREAT!
21:04:09 <timburke> i saw https://review.opendev.org/#/c/682138/ landed :D
21:04:10 <patchbot> patch 682138 - swift - Allow internal clients to use reserved namespace (MERGED) - 38 patch sets
21:04:11 <kota_> lol. good
21:04:17 <clayg> we landed the new null namespace primitive - which is pretty cool
21:04:23 <mattoliverau> nice
21:04:32 <clayg> the next patch in the change will actually *use* it
21:04:45 <clayg> the next patch in the chain is also the BEST patch in the chain
21:04:47 <clayg> because...
21:04:51 <clayg> 1) tdasilva wrote most of it
21:04:56 <clayg> 2) it's a new swift feature!
21:05:15 <mattoliverau> both great reasons!
21:05:18 <clayg> mattoliverau: was looking at it and playing with the API - and i'm sure he agrees it's pretty freaking sweet
21:05:41 <clayg> honestly my favorite way to consume the versioning api is using the swiftclient patch
21:05:44 <mattoliverau> it is pretty damn awesome, I'll continue to play and review today
21:05:56 <clayg> I'm hoping they can all come together at the same time
21:06:05 <tdasilva> lol
21:06:41 <clayg> one issue that somehow we didn't spent a lot of time functionally testing is what happens when you try to use the new version'd api requests against containers with legacy versioning mode
21:07:22 <clayg> we don't allow uses to set both modes simoultaneously on purpose - the metadata is orthogonal
21:07:45 <clayg> I'd assume if they do both get set because of eventual consistency we prefer the new versioning mode?
21:07:59 <clayg> somewhere at the top of versioned writes we must pick which paath to go down 🤔
21:08:04 <clayg> tdasilva: how does that work?
21:08:18 <clayg> anyway - we're gunna spend some time polishing that
21:08:48 <clayg> versioned api requests against *unversioned* containers are very intuative - so hopefully we can make versioned api requests against legacy versioned containers equally obvious
21:08:54 <tdasilva> yeah, i need to look that up again
21:09:11 <clayg> the finally last in the chain is the s3api patch that puts aws object versioning on swift's versioning
21:09:26 <clayg> it works great as long as swift's object versioning is enabled 🤣
21:09:39 <tdasilva> and symlinks ;)
21:09:55 <clayg> so I think our fearless leader has volunteered to take a stab at making it degrade more gracefully
21:10:16 <clayg> @tdasilva well, i don't think s3api cares so much about symlinks - as swift object versioning requires symlinks
21:10:40 <clayg> object versioning itself already blows up very loudly if you don't have symlinks 👍
21:10:47 <tdasilva> true true
21:10:49 <timburke> i did! http://paste.openstack.org/show/787123/ makes it a bunch better -- need to look at multi-deletes, though
21:10:56 <clayg> so it's a CHAIN of dependencies - but the last link (s3api) wasn't as annoying as it should be
21:11:15 <mattoliverau> nice work guys
21:11:17 <clayg> and look at that timburke is all over it
21:11:28 <clayg> mattoliverau: it's been a team effort - this last mile stuff feels good tho
21:11:39 <clayg> when you start USING the new version api - it's like butter
21:11:41 <timburke> i knew if i didn't do it quick, i'd hardly write *any* code today ;-)
21:11:54 <clayg> timburke: yeah welcome back!  😁
21:12:52 <clayg> i would recommend anyone that wants to "poke at the edges" to try and spin up the s3api patch - and make sure you can deploy a saio with everything turned on (symlinks + allow_versioned_writes)
21:13:07 <clayg> then maybe use `aws s3api` to put-object-versioning
21:13:55 <clayg> then if you're feeling REALLY fancy - do some curl requests against a noauth proxy with allow_+allow_reserved_names_header
21:14:22 <clayg> the format of the null namespace container is probably worth a grock if you can get down to it
21:17:03 <timburke> does anyone have any comments or questions on those patches?
21:17:45 <kota_> sounds awesome progress on the work
21:19:41 <clayg> kota_: thanks!  yes we're super exicted - so close we can taste it!
21:19:44 <clayg> FINALLY
21:20:17 <timburke> all right. looks like neither rledisez nor alecuyer are around right now to talk about their lots of files investigations (either continuing to play with feature/losf or trying to split off FS metadata to a separate, faster device)... so...
21:20:23 <timburke> #topic open discussion
21:20:48 <timburke> sorry, i'm really kinda under-prepared this week ;-)
21:21:01 <timburke> does anyone have anything else they'd like to bring up?
21:21:29 <mattoliverau> I don't have too much to say, been a little busy ramping up in new role. Also we have bush fires close. I'm safe, friends have been evacuated and this morning our streets are filled with smoke.. so exciting times
21:21:42 <clayg> timburke: since you just got back from Hawaii and Thiago is in Oz - I'm feeling jelous and wanting to plan the hackathon at mattoliverau 's beach house
21:21:55 <timburke> eep! stay safe, mattoliverau!
21:22:00 <clayg> bush fires!!!???
21:22:03 <mattoliverau> yeah, come for the smoke.
21:22:09 <clayg> nm, we can plan the hackathon later 😬
21:22:11 <mattoliverau> bush fires is australian for wildfires
21:22:22 <clayg> mattoliverau: I used to live in CA - i know all about the bush fires
21:22:30 <mattoliverau> clayg: I bet you do :)
21:22:33 <clayg> or... or wild fires
21:22:49 <mattoliverau> I'm safe, on the opposite side of a bay and a bridge
21:22:58 <kota_> wow, stay safe please.
21:23:02 <clayg> tdasilva: how close are you to mattoliverau ???   like 6-8 hours 🙄
21:23:19 <mattoliverau> if tdasilva is in Syd, then about 5 hours.
21:23:23 <tdasilva> not sure, i'm in sydney, i think it's close
21:23:30 <mattoliverau> because of some mountains in the way :)
21:23:32 <tdasilva> well...not so close
21:23:48 <timburke> "close" by australian standards :-)
21:23:54 <mattoliverau> ^ ++
21:23:54 <tdasilva> heh
21:24:01 <kota_> lol
21:24:11 <timburke> it's a bit worse than "texas close", i think
21:24:19 <mattoliverau> tdasilva: if you feel like a drive, you definitely welcome to come visit!!
21:24:32 <tdasilva> getting on the car right after this :)
21:25:01 <kota_> 5 hours... maybe I can go to Fukuoka by bullet train :/
21:25:06 <mattoliverau> if you decide to visit my hometown of Canberra, then you're getting much closer :)
21:25:14 <clayg> https://www.texasmonthly.com/wp-content/uploads/2015/01/australia.png #fakenews
21:25:17 <clayg> texas is biggest
21:25:29 <mattoliverau> kota_: they've talked about putting a fast train down the east coast of Oz
21:25:47 <seongsoocho> 5 hours is 'close" in australia?
21:25:52 <mattoliverau> yup
21:25:57 <clayg> seongsoocho: !!!
21:26:00 <seongsoocho> wow...
21:26:08 <timburke> could be worse, could be perth!
21:26:13 <clayg> seongsoocho: what are you working on?
21:26:17 <mattoliverau> I used to go visit my wife grandmother in Mel for the weekend, that's 9-10 hours drive.
21:26:34 <tdasilva> mattoliverau: from sydney or from your house?
21:26:53 <mattoliverau> ours, or about 7 hours from Canberra.
21:27:17 <seongsoocho> clayg:  Hi ~ I try to do some fix easy issue ( you told me on PTG)
21:27:47 <clayg> seongsoocho: please let me know if you have some patches I need to look at - it's very easy to get distracted and forget what other people are working on
21:28:01 <clayg> anytime you're stuck or waiting on reviews this meeting is best time to come poke and remind folks
21:28:22 <seongsoocho> Ok ! (
21:28:27 <mattoliverau> tdasilva: but we're going to catch up in Mel over Jan aren't we? :)
21:28:47 <timburke> also, you're always welcome to add items to the agenda if there's anything you'd like to discuss or bring attention to
21:29:01 <tdasilva> mattoliverau: yeah, sounds great!
21:30:53 <timburke> oh yeah, so i was thinking about the new allow_reserved_names_header option for gatekeeper (https://review.opendev.org/#/c/682138/38/swift/common/middleware/gatekeeper.py) ...
21:30:54 <patchbot> patch 682138 - swift - Allow internal clients to use reserved namespace (MERGED) - 38 patch sets
21:31:19 <timburke> and i wrote a thing: https://github.com/tipabu/no_op_gatekeeper
21:31:49 <timburke> basically, an alternate gatekeeper that doesn't do any gate-keeping ;-)
21:32:30 <timburke> thought it might be useful for anyone else who likes a no-auth proxy for their dev environment
21:33:05 <mattoliverau> interesting
21:34:49 <timburke> if you look at https://github.com/tipabu/no_op_gatekeeper/blob/master/no_op_gatekeeper.py, it's really not ;-)
21:35:48 <mattoliverau> lol
21:36:05 <kota_> lol
21:37:07 <timburke> i suppose i should go sign up for pypi or something, but fwiw, `pip install git+https://github.com/tipabu/no_op_gatekeeper@1.0` should work just fine
21:38:07 <kota_> or just call `python setup.py devel (or install?)` maybe?
21:38:31 <kota_> ah, after cloning, i mean.
21:38:40 <timburke> yeah -- i was mainly thinking if you didn't want to have to clone it directly :-)
21:38:51 <kota_> timburke: good
21:40:10 <timburke> anyway, that's about all i've got. clayg, let's let all these good folks get breakfast ;-)
21:40:30 <clayg> 👍
21:40:42 <timburke> thank you all for coming today, and thank you for working on swift!
21:40:47 <timburke> #endmeeting