07:00:50 <mattoliver> #startmeeting swift
07:00:50 <opendevmeet> Meeting started Wed Oct  8 07:00:50 2025 UTC and is due to finish in 60 minutes.  The chair is mattoliver. Information about MeetBot at http://wiki.debian.org/MeetBot.
07:00:50 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
07:00:50 <opendevmeet> The meeting name has been set to 'swift'
07:01:03 <mattoliver> whose here for the swift meeting?
07:01:05 <cschwede> o/
07:01:24 <mattoliver> \o/ not talking to myself :P
07:01:49 <cschwede> :D
07:01:51 <mattoliver> And this time I took day light savings into account
07:02:21 <mattoliver> Still didn't update the agenda, so just hit some highlights, probably more informational today
07:02:31 <kota> thx mattoliver
07:02:32 <mattoliver> #topic eventlet removal POC update
07:02:43 <mattoliver> #link https://review.opendev.org/q/status:open+project:openstack/swift+branch:feature/threaded
07:03:05 <cschwede> Working on it, there will be a few updates before the next mtg.
07:03:16 <mattoliver> Cool
07:03:17 <cschwede> And I need to work on the "getting started" doc
07:03:41 <cschwede> I noticed Tim merged a few initial patches, which is quite nice :)
07:03:57 <mattoliver> next meeting is in 2 weeks and I'll be in the US that week.. so let's see how that goes.. might have to skip it. So you probably get 4 weeks :P
07:04:25 <cschwede> 3 then, because PTG :D
07:04:53 <mattoliver> oh yeah! PTG.. that's much better! and a much better target on the state of the feature branch :)
07:05:02 <mattoliver> speaking of whic..
07:05:14 <mattoliver> #topic October vPTG
07:05:30 <mattoliver> As cschwede says, its only 3 weeks away!
07:05:37 <mattoliver> Oct 27-31
07:06:03 <mattoliver> Register if your coming (its free) at...
07:06:07 <mattoliver> #link https://ptg.openinfra.org/
07:06:17 <mattoliver> And full out the etherpad of topics here..
07:06:24 <mattoliver> #link https://etherpad.opendev.org/p/swift-ptg-gazpacho
07:07:47 <mattoliver> Tim is unwell atm.. so I'll take a look at things and maybe setup a time poll or pick times that suite us like we usually go in his absence
07:08:24 <mattoliver> But please fill out topics and show interest, as that might help
07:08:28 <mattoliver> Moving on..
07:08:39 <mattoliver> #topic ringv2 next steps
07:08:51 <mattoliver> It's work I've been doing so will talk about it ;)
07:09:16 <mattoliver> chains starts:...
07:09:21 <mattoliver> #link https://review.opendev.org/c/openstack/swift/+/955263
07:09:54 <mattoliver> It's coming along nicely. I have reworked an old patch that added READ support using the history tables to the proxy.
07:10:51 <mattoliver> so HEADs and GETs will check an old primary as the first handoff.. but only for a certain confiugred time (for clusters that don't rebalance often)
07:11:02 <mattoliver> Still writing some tests for it tho.
07:11:32 <mattoliver> If anyone wants to cast an eye or even just the first few "clean up" patches because the meat of it starts, that'll be nice.
07:12:16 <mattoliver> #topic unique proxy offsets
07:13:17 <mattoliver> This too has moved on a bit. more digging into an alt approach that acoles come up with. So I'm kinda chasing both to see how they both work
07:13:25 <mattoliver> mine:..
07:13:33 <mattoliver> #link https://review.opendev.org/c/openstack/swift/+/959009
07:14:09 <cschwede> so the alt approach from Alistair is now in your patch or is that a different one?
07:14:35 <cschwede> nevermind, Alistair linked it in the review comments
07:14:36 <mattoliver> relates to using the offset field which is 16^16 to also store a unique number for each proxy.
07:14:50 <mattoliver> and acoles one:
07:15:21 <mattoliver> #link  https://review.opendev.org/c/openstack/swift/+/962275
07:15:52 <mattoliver> which is to have to proxy use a uid, that is passed, in his patch also via the timestamp class.
07:16:26 <mattoliver> his might have more code churn as we make sure that things like ssync all behave with new possible TS fields.
07:17:40 <mattoliver> I like the simplicity of mine and think it's what offsets should be used for.. but Al's is basically doing the same thing but just putting it as another part of the timestamp string. But the big advantage to his is we cold even just use the txid as the uid
07:18:41 <mattoliver> so although it isn't a unique proxy number, it is unique to the request and therefore stop collisions AND would mean you could tie an object directly to the req that placed it.. which is pretty freakin awesome.
07:19:46 <mattoliver> Eitherway, still work to do on both. mostly to push on his to see how much code churn it'll cause.. if minimal it might be a winner.
07:19:48 <cschwede> yes, that sounds pretty nice!
07:21:23 <mattoliver> Thats all I really have for today .. because I'm a little un organised today
07:21:30 <mattoliver> #topic open discussion
07:21:51 <mattoliver> I did see this patch come through today:..
07:22:03 <mattoliver> #link https://review.opendev.org/c/openstack/swift/+/963325
07:22:09 <mattoliver> and look who it's from!
07:22:39 <cschwede> Yes, noticed the IRC msg earlier - awesome :D
07:23:12 <mattoliver> I've known he's at nvidia for a while now. But Sam is now working at nvidia, has been for quite a while.. but not on swift, in a different area. But awesome to see a patch from him!
07:24:06 <mattoliver> Anything else from anyone else today?
07:25:46 <cschwede> Not from my side. Thanks mattoliver for running this mtg, have a nice evening!
07:26:03 <mattoliver> kk, then let's call it early.
07:26:23 <mattoliver> Don'f forget, we'll skip the next meeting, so see everyone at the PTG in 3 weeks!
07:26:31 <cschwede> +1
07:26:36 <mattoliver> Thanks for coming and thanks for working on swift!
07:26:38 <mattoliver> #endmeeting