21:02:41 <timburke> #startmeeting swift 21:02:42 <openstack> Meeting started Wed Mar 17 21:02:41 2021 UTC and is due to finish in 60 minutes. The chair is timburke. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:02:43 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:02:46 <openstack> The meeting name has been set to 'swift' 21:02:48 <timburke> who's here for the swift meeting? 21:02:53 <mattoliverau> o/ 21:02:54 <acoles> o/ 21:03:12 <kota_> o/ 21:04:53 <timburke> as usual, the agenda's at https://wiki.openstack.org/wiki/Meetings/Swift 21:05:11 <timburke> first up 21:05:15 <timburke> #topic PTG 21:05:41 <timburke> virtual PTG is scheduled for April 19-23 21:06:04 <timburke> and i made a doodle to figure out when we want to do meeting times 21:06:09 <timburke> #link https://doodle.com/poll/cnr3qb7y2afiad9v 21:06:24 <timburke> as well as an etherpad to collect discussion topics 21:06:28 <timburke> #link https://etherpad.opendev.org/p/swift-ptg-xena 21:07:02 <timburke> if you haven't already, please register (iirc it helps the foundation with planning) 21:07:05 <timburke> #link https://april2021-ptg.eventbrite.com/ 21:08:14 <timburke> there's a chance i might not be around to lead discussions -- would anyone like to step up if i'm out? 21:08:51 <mattoliverau> for the whole thing or at certain times? 21:08:56 <acoles> sure, depending on time of the meeting 21:09:20 <timburke> whole thing -- and you'd be able to have more control over the meeting times ;-) 21:09:30 <timburke> or, you could split it up between you 21:09:37 <mattoliverau> either way, we can figure something out, but your input input would be missed :) 21:09:52 <mattoliverau> yup both inputs :P 21:10:15 <timburke> i accidentally scheduled a camping trip at the same time -- though now it looks like my daughter might be going back to in-person schooling the week before. suffice to say, schedules are in flux 21:10:44 <mattoliverau> vptg camping trip, great idea! :P 21:11:23 <timburke> #topic 2.27.0 release 21:11:35 <acoles> we'll work something out, sharing around might be good and help cover timezones 21:11:40 <timburke> we've got a changelog! https://review.opendev.org/c/openstack/swift/+/777708 21:12:00 <timburke> and there should be a tag once https://review.opendev.org/c/openstack/releases/+/780959 merges 21:13:28 <timburke> not sure there's much to say beyond that, just want to keep people in the loop ;-) 21:13:28 <timburke> #topic relinker 21:14:15 <timburke> we've had a bunch of great stuff land, including a bit of a rewrite of the cleanup phase so it can make a final attempt at relinking 21:15:01 <timburke> acoles currently has a patch chain up to take that refactor even further: 21:15:09 <timburke> #link https://review.opendev.org/c/openstack/swift/+/779564 21:15:40 <timburke> which makes relink and cleanup use much the same code paths 21:16:03 <timburke> #link https://review.opendev.org/c/openstack/swift/+/779566/4 21:16:40 <timburke> which makes relink just a cleanup with an early exit 21:17:21 <timburke> meanwhile mattoliverau has been working on improving progress visibility 21:17:39 <timburke> #link https://review.opendev.org/c/openstack/swift/+/780440 21:17:58 <timburke> adds policy info to end-of-partition logging 21:18:32 <timburke> #link https://review.opendev.org/c/openstack/swift/+/779655 adds a recon endpoint to monitor relinking progress 21:19:00 <mattoliverau> ^ I should add policy to recon I guess too. 21:20:06 <timburke> i think you already do -- i see a `get_policy_string("progress", policy)` in there ;-) 21:20:35 <acoles> more visibility is great 21:21:09 <mattoliverau> oh good :P past me was smarter the present me :P 21:21:22 <timburke> and i'm still working on getting parallelization somewhere that we could actually merge it 21:21:28 <timburke> do we have any strong feeling about what we want to land first? i'm assuming the refactor will conflict a decent bit with anything else 21:21:41 <zaitcev> mattoliverau: heh 21:23:10 <acoles> mattoliverau: there's an age in life where that becomes inevitable ;) 21:23:20 <mattoliverau> lol 21:23:24 <mattoliverau> yeah, I dont mind the order. there will be rebases which ever way we go. 21:23:40 <mattoliverau> Depends how "done" we think both chains are I guess 21:23:41 <acoles> timburke: I don't have strong feelings. TBH I haven't studied mattoliverau 's patches much so can't imagine the conflicts 21:24:15 <acoles> we could agree to hold off +A's and see which get to +2's first then agree on on order?? 21:24:50 <timburke> well at the moment, everything's in merge conflict anyway; everybody needs a rebase and conflict-resolution :P 21:24:53 <mattoliverau> We could also dump some of the extra stats that acoles collects in later patches into recon, but that again doesn't change when things land :) 21:25:20 <acoles> I can look at mattoliverau 's patches tomorrow (sorry, today was a work on shard repair day) 21:25:22 <mattoliverau> acoles: races to +2 :) 21:25:31 <timburke> i think i might be inclined to do the refactor sooner rather than later, given that both mattoliverau and i have +2s on it currently 21:25:59 <acoles> ok - I need to chat to you about the small concern I had, but maybe that can merge soon 21:26:31 <timburke> yeah, i was going to bring that up ;-) 21:27:20 <timburke> i can work on pushing up a rebase after the meeting, if that'd help. or you can do your investigation and rebase in the morning. either way 21:27:35 <timburke> but we can have that discussion later 21:27:44 <timburke> #topic sharding / shrinking 21:28:08 <timburke> how's it going? 21:29:27 <mattoliverau> Like always, progressing. Current blocker for me is adding an active_age to shrinking so containers that shard too fast because they're mostly empty replicas don't become candidates for shrinking which the other bigger ones shard. 21:30:12 <mattoliverau> but it's not as simple as you expect because of how the state machine works and how shrinking is root driven. 21:30:13 <acoles> since last meeting mattoliverau 's patch to have swift-manage-shard-ranges load config from sharder conf section has merged 21:31:43 <mattoliverau> Some info I placed in https://review.opendev.org/c/openstack/swift/+/777927 and playing with a WIP to set state_timestamp in https://review.opendev.org/c/openstack/swift/+/780705 21:32:42 <mattoliverau> but the latter is far from stable or the solution. in short, more discussions are needed. acoles has another interesting idea. if anyone is curious join in the conversations in those patches ;) 21:35:41 <mattoliverau> I'll make sure the latest info has been added to the active_age one if it isn't completely up to date.. I can't remember :P 21:36:31 <timburke> acoles, looks like you recently rebased the chain starting with repair and analyze commands -- is that good to go, as far as you know? anything we ought to be watching for? 21:36:47 <acoles> mattoliverau: I did wonder if an etherpad might be useful for that topic? to capture the various ideas and discussion? 21:36:55 <mattoliverau> oh great idea! 21:37:07 <timburke> +1 21:37:09 <acoles> timburke: yes ready for review (I know mattoliverau already started review) 21:37:20 <mattoliverau> I'll create one and post it in channel and in the changes 21:37:29 <timburke> excellent! 21:37:51 <mattoliverau> yeah, I've been working thorugh it and loading it back into my head. I want to get progress on that. And acoles has done an awesome job on repair! 21:38:21 <timburke> #topic CORS func test failures 21:38:32 <acoles> fun fun fun 21:38:37 <acoles> func 21:38:38 <timburke> so last week i promised to look into those failures more 21:38:47 <timburke> and wrote up https://bugs.launchpad.net/swift/+bug/1918864 21:38:49 <openstack> Launchpad bug 1918864 in OpenStack Object Storage (swift) "CORS functional tests sometimes fail with `invalid argument: can't kill an exited process`" [Undecided,New] 21:39:29 <timburke> trouble seems to be geckodriver complaining "Error: cannot open display: :99" 21:40:34 <timburke> looks like acoles put up https://review.opendev.org/c/openstack/swift/+/781084 to try to help -- is that mostly speculative/hopeful? i've still not managed to repro the original bug locally 21:41:10 <acoles> yep, hopeful ! I also have not managed to repro 21:41:53 <acoles> but maybe running headless avoids trying and failing to open a display 21:42:11 <timburke> what do you think about always trying to use the headless option? any reason not to? 21:43:05 <acoles> fine by me 21:43:28 <acoles> I was just being cautious not to tread on toes by making it optional, happy to hard wire it 21:43:59 <timburke> all right. maybe i'll respin and merge that then 21:44:12 <acoles> obviously it doesn't apply to all browsers but for the zuul job we're pinned to firefox right? so that's ok 21:44:26 <timburke> if it doesn't seem to resolve the bug, i'll probably follow-up by making the job non-voting 21:44:35 <acoles> timburke: go for it 21:44:51 <timburke> all right 21:44:57 <timburke> #topic open discussion 21:45:07 <timburke> what else should we discuss this week? 21:48:59 <timburke> ok. we'll let everyone get on with their day 21:49:10 <timburke> thank you all for coming, and thank you for working on swift! 21:49:13 <timburke> #endmeeting