21:00:10 #startmeeting swift 21:00:10 Meeting started Wed Feb 2 21:00:10 2022 UTC and is due to finish in 60 minutes. The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:10 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:10 The meeting name has been set to 'swift' 21:00:17 who's here for the swift meeting? 21:00:22 o/ 21:00:26 o/ 21:01:33 o/ 21:02:10 as usual, thhe agenda's at https://wiki.openstack.org/wiki/Meetings/Swift 21:02:20 first up 21:02:24 #topic PTG 21:02:41 registration is open! 21:02:44 #link https://openinfra-ptg.eventbrite.com/ 21:03:07 with luck, this'll be our last virtual one! 21:03:29 dates are April 4-8 21:03:43 Time to work on all the last ptg items before the next :p 21:04:40 ...yeah, i should go through those aaction items again... 21:04:52 but first 21:05:01 #topic 2.29.0 release 21:05:12 i want to get a release out! 21:05:31 changelog updates are up at 21:05:34 #link https://review.opendev.org/c/openstack/swift/+/826947https://review.opendev.org/c/openstack/swift/+/826947 21:05:43 #undo 21:05:43 Removing item from minutes: #link https://review.opendev.org/c/openstack/swift/+/826947https://review.opendev.org/c/openstack/swift/+/826947 21:05:47 #link https://review.opendev.org/c/openstack/swift/+/826947 21:06:09 it's going to be a great release! 21:06:28 Nice 21:06:55 🎉 21:07:26 if you've got any other work in flight, let's try to get it landed; i'll update the changelog as needed 21:07:34 timburke: do we have any outstanding patches we'd like to include before release? 21:07:45 oh, ok :) 21:08:07 there were a few i was thinking of; i put them on the priority reviews page 21:08:10 #link https://wiki.openstack.org/wiki/Swift/PriorityReviews 21:08:14 Is everything on the priority list done? I should just go check 21:08:21 https://review.opendev.org/c/openstack/swift/+/822585 - Trim sensitive information in the logs (CVE-2017-8761) 21:08:29 https://review.opendev.org/c/openstack/swift/+/813715 - s3api: Fix multipart upload listings 21:08:36 https://review.opendev.org/c/openstack/swift/+/826326 - proxy-server: add stats for backend shard_listing requests 21:08:49 though they're all a *little* hairy 21:09:16 maybe i'll rearrange my agenda a little ;-) 21:09:28 #topic tempurl signatures in logs 21:10:26 so i think the solution at https://review.opendev.org/c/openstack/swift/+/822585 is coming along nicely -- though some functions are probably in the wrong place 21:10:53 mattoliver took a stab at pulling them out to a new module in https://review.opendev.org/c/openstack/swift/+/825694 21:11:28 it creates a decent bit of churn, though, since it's re-homing register_swift_info as well 21:12:01 If people like the move then it can get squashed down into the first. 21:12:01 i wanted to get people's opinion on that -- should we go ahead and squash that down? 21:12:46 I have a preference to keep the behavioural change and the code relocation in separate patches, to make review easier 21:13:48 So leave them as 2 and as they are? Or the other way around? Move then behaviour change? 21:14:29 The move was also because the new sensitive registers stuff was in the wrong place 21:14:54 I don't want to create work for anyone but maybe move existing first then make change and add new functions makes most sense??? 21:15:59 Kk, if everyone else agrees I'll swap em :) 21:16:06 mattoliver, do you want to do that, or should i? 21:16:23 I'll do it 21:16:37 thanks mattoliver 21:16:38 👍 thanks 21:17:17 #topic multipart upload listings 21:17:32 #link https://review.opendev.org/c/openstack/swift/+/813715 21:18:08 i know we've been running with this in prod for a while, but i'm pretty sure i spotted a bug i'd like to fix before landing 21:18:39 i'll work on re-spinning that this afternoon (tests will probably be the time-consuming part) 21:19:00 #topic proxy-server stats 21:19:53 #link https://review.opendev.org/c/openstack/swift/+/826326 adds some stats for listing shard ranges that mirror stuff we've already got for updating 21:20:22 but when i put it on the reviews page i think i didn't realize it had some pre-reqs 21:21:10 in particular, https://review.opendev.org/c/openstack/swift/+/824608 - Fix statsd prefix mutation in proxy controllers 21:23:07 the more i dig into it, the more it seems ripe for a yak-shave 21:23:29 and i'm a little worried we're currently adding more hair :-/ 21:26:04 Is the yak changing the metric from what the patch is doing? Because if not maybe we land it to fix then yak-shave follow up? 21:26:09 acoles, should i just grin and bear it? can you help me feel better about the logger class hierarchies? 21:27:08 looking 21:28:09 i think what i *really* want is to pull the statsd client off the logger. start passing around both a statsd client *and* a logger everywhere we currently just have a logger, and have some separate statsd-client wrapper to handle the prefixing of stats 21:29:56 having both a PrefixLoggerAdapter and a MetricsPrefixLoggerAdapter feels weird. and i'm a little worried about the thread_locals stuff (though maybe it's unwarranted) 21:31:00 re. the conflation of logging and stasd, I also scratched my head at that but I don't think I can justify spending time on unravelling it 21:32:37 re. the LogAdapter inheritances, at first glance I'm not sure I see the problem - each adapter wraps another so doesn't need to inherit the other's behaviours 21:33:39 there is an assumption that the wrapped logger is a utils.LogAdapter (i.e. has stasd functions and thread locals) 21:34:08 the fact that we've got a LogAdapter that's a sibling of our SwiftLogAdapter felt weird to me. pre-existing behavior, but still... 21:36:45 on patchset 5 I got feedback that nudged me towards the separate MetricsPrefixLoggerAdapter 21:39:26 Passing around a statsd client too does sound a bunch of work. I kinda like we can statsd and log from the same client, but we do have adapters all over the place.. but I guess that's what the adapter pattern does, add functionality. 21:39:41 maybe utils.LogAdapter could inherit from utils.SwiftLoggerAdapter and lose the stasd delegate stuff?? IDK, I'd need to study with fresh eyes 21:39:44 ok, one other thing that bugged me: the idea of the patch seems to be that `set_statsd_prefix` (or maybe it's the statsd client's `set_prefix`?) is inherently undesireable -- if you're going to set a prefix, it ought to be via a wrapper, it'll be set in __init__, and you shouldn't need to touch it again. but then we still have (and use!) the other interfaces -- we're just more careful about how/when we do it 21:40:28 i kinda want to just get rid of the dangerous interface 21:40:41 I agree, but not in this patch :) 21:41:07 grin and bear it, got it! 👍 21:41:31 and, note that elsewhere we are considering introducing the same pattern to set proxy-server backend user agent (i.e. mutate attribute after init) 21:41:43 looking for that patch... 21:42:05 https://review.opendev.org/c/openstack/swift/+/819387 21:42:20 Until we can trick(convince) someone to go fix it all ;) 21:42:21 all right, enough of me whining about code purity 21:42:29 Lol 21:42:35 #topic updater fairness 21:42:59 we've landed a bunch of patches related to this; i think the only outstanding item is the deferral queue 21:43:07 #link https://review.opendev.org/c/openstack/swift/+/821736 21:43:15 timburke: I'm with you on the purity, just feeling more pragmatic :D 21:43:36 We should write it down as a bug/tech debt to fix up, so 1, we remember and 2, others can find it to work on it 21:45:04 acoles, i just wanted to check in on how the deferal queue is going -- i think i saw some fix ups recently? 21:46:40 eh, i should just go review the patch ;-) 21:46:51 #topic memcache refreshing 21:47:02 there was a little churn triggered by your query about the stats - actually, I've not had chance to catch up with you on that. I left a long comment, and fixed a related thing, but otherwise left it as is ... pending you pointing out my folly 21:47:32 I should add some comment in the patch perhaps about how to count things 21:47:45 cool -- i'll be sure to review it 21:48:04 the shard-range patch has landed now, and the reemoval of pipeline_property is making its way through the gate 21:48:42 i wanted to check if anyone thinkgs they'll have time to look at the account/container info patch: https://review.opendev.org/c/openstack/swift/+/821921 21:52:08 I'll put it on my todo 21:52:17 thanks mattoliver 21:52:32 that's all i had for this week 21:52:36 #topic open discussion 21:52:44 anything else we ought to bring up? 21:53:29 I pushed up a ptg version of serialisation v2. But gzip seek might not be supported in py2 21:53:29 thanks for shepherding everything timburke, there's been lots going on ! 21:53:46 So I left some questions in the review. 21:54:17 Maybe make it py3 only... anyway, maybe something to bring up next meeting. 21:54:48 oh yeah, thanks mattoliver! i'd been meaning to loop back to that 21:55:00 No stress, lots going on 21:55:23 Would like to progress that hair the so we can benefit from last primaries ;) 21:55:31 *chain 21:55:34 timburke said yoga is the end of the road for py2 🥳 21:56:01 No idea why phone change that to hair. 21:56:13 exactly! anything to make "emergency rebalance" mode less common will be *great* 21:56:17 Yeah, it'll make things much simpler. 21:56:19 I'd like to progress my hair ;) 21:56:32 Lol 21:56:58 acoles, i know the feeling ;-) 21:57:08 all right 21:57:19 thank you all for coming, annd thank you for working on swift! 21:57:22 #endmeeting