07:02:16 <mattoliver> #startmeeting swift 07:02:16 <opendevmeet> Meeting started Wed May 21 07:02:16 2025 UTC and is due to finish in 60 minutes. The chair is mattoliver. Information about MeetBot at http://wiki.debian.org/MeetBot. 07:02:16 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 07:02:16 <opendevmeet> The meeting name has been set to 'swift' 07:02:29 <mattoliver> Who's here for the swift meeting? 07:02:34 <kota> o/ 07:02:41 <cschwede> o/ 07:03:09 <mattoliver> \o/ first 0700 meeting in years 07:03:20 <mattoliver> Agenda is at: 07:03:23 <mattoliver> #link https://wiki.openstack.org/wiki/Meetings/Swift 07:03:26 <kota> good. it's healthy time for APAC 07:03:43 <mattoliver> kota: yeah, much easier for us too ;) 07:03:49 <cschwede> Same for EMEA, which is nice as well :) 07:04:10 <mattoliver> I've seeded some topics, and stole some from Tim more as an FYI 07:04:34 <mattoliver> First up one I made for cschwede :P 07:04:44 <mattoliver> #topic Bug triage 07:05:13 <mattoliver> Hows the triage going cschwede ? we probably should all be helping you tho. 07:05:42 <cschwede> On it. A bit slower than expected, but I'm running over bugs in NEW state atm. Updated ~10 earlier today, and will update our wiki page at https://wiki.openstack.org/wiki/Swift/BugTriage later 07:06:05 <mattoliver> nice one! 07:06:16 <mattoliver> #link https://wiki.openstack.org/wiki/Swift/BugTriage 07:06:25 <mattoliver> (so it links properly in logs) 07:06:44 <cschwede> Noticed a few very easy ones (like one-liner doc changes) that I might simply propose. Tagged as low-hanging fruit for now 07:07:38 <mattoliver> cool, yeah, if there too easy then even newbies might not be that interested in them. 07:08:03 <mattoliver> Thanks so much for working through these each fortnight! 07:08:05 <cschwede> Right, at the same time i don't want to close them simply without ack'ing the time of the reporter 07:08:51 <mattoliver> yeah fair enough 07:10:10 <mattoliver> I see all the extra emails in my inbox. Anything we need to do on our end? Can I help by taking a look. Though sounds like your marking them as expected 07:11:14 <cschwede> I run over a couple of bugs that need more investigation. I think I will create a list of things that need a second look and share that as well 07:12:22 <mattoliver> cool, sounds like a plan. Happy to double check anything your need. 07:12:27 <mattoliver> next topic then 07:12:31 <cschwede> Thanks Matt! 07:12:42 <mattoliver> #topic eventlet removal POC update 07:13:02 <mattoliver> This is just a place holder topic, no stress on how it's going.. but curious 07:13:26 <cschwede> Working on it, and also looked into cheroot mentioned earlier by timburke_. As expected it's a rabbit hole, but I hope to have a POC by the end of June 07:13:31 <cschwede> s/hope/plan 07:14:05 <mattoliver> lol, yeah huge rabbit hole 07:14:37 <mattoliver> Nice tho, I haven't looked at cheroot, so sounds like I might get nerdsniped :P 07:15:27 <mattoliver> I've let my manager know eventlet removal will be something that will have to happen in the future for us too, so thanks for giving at least a POC a crack so we can start moving in the right direction! 07:16:23 <cschwede> That's great news, very much appreciated! 07:16:24 <mattoliver> Want me to leave it in the meeting agenda to check in, or should I just add in every month or so? 07:16:54 <cschwede> I won't have an update in 2 weeks, but let's keep it on the agenda and we check in 4 weeks on the state. 07:17:21 <mattoliver> kk 07:17:23 <mattoliver> #topic SO_TIMESTAMP/SO_TIMESTAMPING patch 07:17:32 <mattoliver> #link https://review.opendev.org/c/openstack/swift/+/944103 07:17:42 <mattoliver> I bring this up because its one of mine :P 07:18:16 <mattoliver> I think its coming along nicely, timburke_ has been reviewing it, so starting to look good. 07:19:01 <mattoliver> This patch will allow us to track the time the request hits the server, not just when we accept it. Also the time it waited on the accept queue can also be logged. 07:19:05 <cschwede> That's the one we discussed at the PTG, right? Is there any issue with swift-bench that needs a fix (unrelated to this patch), I see Tim mentioned that in the comments? 07:19:14 <mattoliver> yup 07:19:48 <mattoliver> if the client is doing pipelined requests, then the fields in the logs (if defined) will just be '-' because we already have the connection open. 07:20:00 <mattoliver> Although, I guess I better make a test for that or something :) 07:20:51 <mattoliver> We want to alert of time spent in accept queues because this is a sign of overloading. Maybe it's interesting to others too. 07:21:18 <mattoliver> especally when a client's time a request took and differ then our logs because of the wait. 07:21:28 <cschwede> Yes, definetely useful 07:21:58 <mattoliver> Otherwise, not too much to report there. If anyone wants to take it for a spin or review great. :) 07:22:20 <mattoliver> next topic 07:22:26 <mattoliver> #topic ringv2 07:22:34 <mattoliver> We talked about this at the PTG too.. 07:23:02 <mattoliver> in fact alot of the serialisation design work happened with everyone at PTGs past 07:23:20 <cschwede> Looks like it got some updates very recently 07:23:20 <mattoliver> But it still hasn't landed :( 07:23:43 <mattoliver> But good news is, we've been running it prod (defaulting to v1). 07:24:09 <cschwede> That's the case for quite some time already, isn't it? 07:24:23 <mattoliver> Soak tested a few racks by running 4-byte devid rings in our biggest cluster. 07:24:26 <mattoliver> yup 07:24:33 <mattoliver> which went well. 07:24:49 <mattoliver> And just today we've enabled ringv2 in some of our smaller clusters 07:25:11 <mattoliver> *small prod clusters 07:25:35 <cschwede> That's great news, sounds like it is working as expected then 07:25:57 <mattoliver> timburke_: responded and cleaned up some stuff, so I hope we can finally land it. 07:27:10 <mattoliver> Once it lands I want to blow the dust of my last_part_table patch, so reconstructors can find missing frags without rebuilding after a rebalance (hopefully remove the need for handoffs_only option) 07:27:25 <mattoliver> Well remove the need after a rebalance at least 07:28:41 <mattoliver> The last few topics are really just FYIs, as they're topics in the other timed meeting. 07:28:49 <mattoliver> #topic aws-chunked 07:29:15 <mattoliver> So I went on the landing rampage a few days ago, and went and landed the first 2 patches in the chain. 07:29:42 <mattoliver> They work great and figured people can do any more polishing in follow up patches. 07:30:26 <mattoliver> This means we finally have support for aws-chunked so latest boto should now work. 07:30:42 <mattoliver> The next patch in the chain is important tho 07:30:46 <mattoliver> #link https://review.opendev.org/c/openstack/swift/+/944073 07:31:05 <mattoliver> As it actually means we actually start checking the crcs users give us :P 07:31:22 <mattoliver> So want it to land before the next actual swift release 07:31:45 <cschwede> Supporting latest boto release and crc checking sound like very useful fixes 07:31:50 <mattoliver> There are some test requirements causing issues, so couldn't land that one yet. 07:32:07 <mattoliver> yeah! 07:32:30 <mattoliver> And I hope we can start using more of the crcs in swift too. 07:32:47 <mattoliver> Moving on 07:33:00 <mattoliver> #topic swiftclient release 07:33:35 <mattoliver> timburke_ needs to cut a new swiftclient release soon. 07:33:44 <cschwede> Nice, ptetera and his fix now mentioned as author in the 4.8.0 release 07:33:59 <mattoliver> \o/ 07:34:39 <mattoliver> Final topic I have 07:34:50 <mattoliver> #topic conditional PUTs for s3api 07:35:11 <mattoliver> #link https://review.opendev.org/c/openstack/swift/+/949680 07:35:36 <mattoliver> This one we found by a user using some s3 client (as normal). 07:36:38 <mattoliver> AWS has finally added If-Match and If-None-Match support now. We've always had If-None-Match: in swift 07:37:08 <mattoliver> So that patch starting using it in s3api 07:37:22 <mattoliver> oh and it's already merged, so that makes life eaier 07:37:27 <cschwede> :) 07:38:12 <mattoliver> There is more work to do on it, but at least we could easily support part of it. 07:38:43 <mattoliver> Anyway, that's all I got for the agenda so.. 07:38:49 <mattoliver> #topic open discussion 07:39:08 <mattoliver> Anything else anyone wants to bring up the meeting? 07:39:26 <mattoliver> hows ptetera going? 07:40:01 <cschwede> He's working on more patches - will meet with him later today 07:40:17 <mattoliver> cool! 07:41:58 <cschwede> Nothing else to bring up today on my side, thx 07:43:10 <mattoliver> We're working on base OS migration stuff. Which means tracing is finally coming onto our upcoming roadmap. So tracing might finally land soon! I've been cleaning it up. Will push the rebased version (which is always a pain) tomorrow probably. 07:43:37 <mattoliver> I should probably go a rebase the vsaio branch too. 07:43:46 <mattoliver> Anyway, that's all I have 07:43:48 <cschwede> Are you using tracing to detect differences between OS releases? 07:44:35 <mattoliver> I could. but more the fact so I can actually build modern python packages :P 07:44:46 <mattoliver> and therefore start actaully using otel :P 07:44:47 <cschwede> :) 07:45:35 <mattoliver> Cool, well I'll call it early. I might go start making dinner for the fam :) 07:45:44 <mattoliver> Thanks for coming and thanks for working on swift! 07:45:49 <mattoliver> #endmeeting