21:01:32 <notmyname> #startmeeting swift 21:01:34 <openstack> Meeting started Wed Apr 4 21:01:32 2018 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:01:35 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:01:37 <openstack> The meeting name has been set to 'swift' 21:01:39 <notmyname> who's here for the swift team meeting? 21:01:43 <zaitcev> o7 21:01:43 <clayg> best topic 21:01:50 <m_kazuhiro> o/ 21:01:51 <timburke> o/ 21:01:52 <kota_> o/ 21:01:52 <acoles> here 21:01:55 <clayg> \o/ 21:02:08 <rledisez_> hi o/ 21:02:15 <notmyname> welcome everyone 21:02:16 <torgomatic_> . 21:02:21 <notmyname> agenda at 21:02:22 <notmyname> #link https://wiki.openstack.org/wiki/Meetings/Swift 21:02:33 <tdasilva> hello 21:02:49 <notmyname> there are a lot of things going on, so let's talk about it 21:02:57 <notmyname> #topic lots going on, not a lot of time (or reviewers) 21:03:24 <timburke> idk what you're talking about -- just hop on a feature branch and +A all the things :P 21:04:07 <notmyname> currently, AFAIK, we've got feature/deep (which is huge), feature/s3api (which may or may not be big), torgomatic_ is working on improving the consistency engine, rledisez_ is doing awesome stuff with LOSF (which we haven't seen yet *hint*hint*), there's py3 works, and probably a bunch of other stuff too 21:05:01 <notmyname> furthermore, we've all got employers who are looking for us to deliver things on "timetables" and "schedules" and with "marketing announcements" 21:05:02 <notmyname> ;-) 21:05:30 <notmyname> but seriously, we've got users in general who want the stuff we're writing 21:05:39 <notmyname> and that's really great 21:05:48 <notmyname> but hard (impossible) to all get done at the same time 21:06:04 <notmyname> especially given the fact that we've got fewer people around that we used to 21:06:23 <clayg> timburke: knows how it is 21:06:57 <notmyname> so I'd like to talk about 3 big things and how we think we can order them in order to not get ourselves more unhappy than necessary 21:07:12 <notmyname> feature/deep, feature/s3api, and improved consistency engine 21:07:44 <notmyname> (I realize that some people may not agree with those being the three biggest things right now. that's valid) 21:08:28 <notmyname> feature/deep is the biggest current ongoing work right now 21:08:49 <notmyname> ideally, there will be something to propose to master in a few weeks 21:09:09 <notmyname> and at that time, it will need a lot of eyes for review 21:09:48 <notmyname> so before moving on, I want to make sure we're generally in agreement. does anyone have any disagreements with anything I've said so far? 21:10:32 <notmyname> I take silence as agreement :-) 21:10:49 <acoles> agree 21:10:57 <timburke> big things, check. lots of eye, check. need a plan -- yup 21:10:59 <kota_> it's not different with mine. I estimated the review time for feature/deep in the last half of this month. 21:11:18 <acoles> kota_: +1 21:12:13 <notmyname> ok, if we all agree that feature/deep is the biggest thing currently going on and that when a merge to master is proposed we'll need a lot of eyes on it... 21:12:35 <notmyname> then that means we've already answered some hard questions: essentially, feature/deep comes first 21:13:16 <notmyname> for example, seeing that feature/s3api isnt' yet proposed to master, it's feeling to me ike it's getting really late to do that before feature/deep 21:14:08 <timburke> idk -- how much more do we need beyond the 3 outstanding patches? kota_? 21:14:20 <notmyname> that is, a big merge to master for s3 compatibility is a lot to ask right before we do an ever bigger merge to master for feature/deep 21:14:59 <notmyname> timburke: IMO it's not even the code as it is the cognative overhead knowing that we'll all need a lot of mental energy to review feature/deep 21:15:03 <kota_> if you don't have opposition, timburke, I'll add +2 +A for the three. then propose the branch to the master ;-) 21:16:03 <clayg> maybe reviewing s3api will take less energy and be a good warm up? It'd be all new to me so I really can't speak to it... 21:16:05 <kota_> they are the last pieces I'd resolve before I'll propose feature/s3api to the master. 21:16:17 <zaitcev> s3api is just a middleware, isn't it 21:16:28 <clayg> otherwise... AFAIK, after is fine too - except yeah... we might all be gun-shy/exhausted 21:16:56 <notmyname> "just" yes, but it's also a big new functionality to officially take on. 21:17:31 <notmyname> and TBH, not everyone is in 100% agreement with the outcome of the dublin discussions and decision to make s3api a simple import of swift3 21:18:17 <notmyname> which means if we don't actually merge it now, we've not done anything that prevents future us from working in a particular direction 21:18:39 <zaitcev> Gaah 21:18:41 <notmyname> sorry if that's vague. I just mean that doing nothing is perhaps a valid thing to do right now 21:19:03 <zaitcev> On both of these, I have a question 21:19:06 <notmyname> sure 21:19:43 <zaitcev> They are on a branch. Suppose I need to catch up. I didn't look at these for a while. But... there's no review on which to comment with -1 and questions. It's a separate branch, not a review. 21:19:52 <zaitcev> How am I supposed to interact with this 21:20:23 <notmyname> zaitcev: for feature/deep, the best thing to do is ask timburke or acoles or clayg in IRC or email 21:21:28 <notmyname> for feature/s3api, you could ask kota_, but what he's been doing so far is simply treating is as an import of swift3. specific question should wait for the proposed merge to master, but more general things could be asked directly 21:21:29 <timburke> i think -- pick a chunk of functionality you want to dig in on. ask questions in channel as necessary. for things that still don't make sense, hack on it until they start to, then propose the result as a new patch 21:21:38 <clayg> zaitcev: if you want to catch up *before* the merge to master - yes - checkout feature/deep - run tests (make sure you config works or use vsaio) - then read the docs - then ask questions in IRC or ML 21:21:38 <acoles> zaitcev: depending on the query, propose a patch to feature/deep if it is a fix, or ask here 21:21:51 <acoles> here == #openstack-swift of course 21:21:53 <zaitcev> oh 21:22:28 <zaitcev> acoles: thanks, that's what I missed 21:22:53 <clayg> notmyname: maybe at somepoint during (or at the begining?) of the merge to master tdasilva can do the bluejeans thing and we can have a virtual design recap (mattoliverau can tell the story for the... 10th time?) 21:23:06 <notmyname> clayg: that's a good idea 21:23:26 <tdasilva> +1 21:23:59 <clayg> I'm mostly assuming at this point a lot of folks know how it "works" - but either they forgot some specific (when the proxy request what, whos responsible for whatever) or they're generally curious out some nitty gritty detail (what happens to async pendings when while your sharding if you rebalance?) 21:24:10 <acoles> zaitcev: the nature of feature branch is that things are a little more fluid than master, so if you see something nonsense then ask before pondering too long - for example, there are chunks of code that need to be deleted which could be confusing (the delete will happen soon) 21:24:11 <notmyname> I'll talk to acoles (mostly) to figure out the right time for that. don't want to distract yet, but dont' want to wait too late either 21:25:07 <acoles> clayg: notmyname: +1 design recap is a great idea. but not this week :) 21:25:43 <notmyname> I added that to my todo list 21:25:53 <clayg> notmyname: my first thought was honestly like schedule it after the first meeting after the first patch is proposed to master 21:26:01 <notmyname> zaitcev: did you get a satisfactory answer to your question? 21:26:14 <zaitcev> notmyname: yes, thanks 21:26:43 <clayg> notmyname: like give folks a few days to see the changes coming, meeting time everyone gets told "yes, it's really happening, tell you managers" then at some point a couple days later we have "ok, I've been reading code for two days straight but i still don't understand XYZ" in a highbandwidth 21:27:13 <notmyname> kota_: I don't want you to feel like you were skipped over. I strongly implied that we should wait on merging feature/s3api for now until after feature/deep has landed. how do you feel about that? 21:28:05 <notmyname> is there anyone who would like to advocate for getting feature/s3api merged to master before feature/deep is proposed to master (in about 2 weeks) 21:28:30 <kota_> depending on how feature/deep takes a time... 21:29:23 <clayg> kota_: that's probably hard to estimate, do you have any idea how long it will take to merge feature/s3api? 21:29:51 <kota_> 2 points talking about the time. 1. How much time before feature/deep proposed to the master, 2. how much time before *landing* to the master. 21:30:01 <kota_> clayg: i kno it's hard tho. 21:30:05 <kota_> know 21:30:37 <clayg> notmyname: so we have some goals related to the timing of feature/deep correct? 21:31:04 <notmyname> kota_: the goal is to propose feature/deep to master no earlier than the week of april 16 21:31:21 <notmyname> (that's and and a half weeks from today) 21:32:41 <kota_> hmm 21:33:10 <notmyname> and hopefully see a merge to master around may 8 (those dates are on https://wiki.openstack.org/wiki/Swift/PriorityReviews) 21:34:25 <notmyname> kota_: hmm indeed :-) 21:35:23 <notmyname> kota_: I'm interested in what you're thinking about 21:37:22 <notmyname> it seems to have gone quiet in kota_'s world 21:37:42 <kota_> thanks to take care about, but for now, I'm not able to get summary of my thought... 21:37:48 <notmyname> ok 21:37:50 <kota_> notmyname: sorry 21:38:05 <notmyname> kota_: please feel free to share later, either publicly or to me privately 21:38:16 <notmyname> so aside from the big feature branches, torgomatic_ has been improving consistency engine pieces 21:38:55 <notmyname> specifically, he's making the object replicator multiprocess, improving the ec reconstructor, and adding some nice config options to the db replicators 21:39:01 <notmyname> his work starts at https://review.openstack.org/#/c/555563/ 21:39:02 <patchbot> patch 555563 - swift - Multiprocess object replicator 21:39:35 <notmyname> unfortunately, he's stuck because normally the people who would review it are all working on feature branches right now! 21:40:48 <notmyname> and that gets us back to the original problem. is it possible to land anything before these feature branches, and if so, how or who does it? 21:40:53 <clayg> feature branch? 21:40:58 <clayg> ^ that's a joke 21:41:14 <clayg> zaitcev: will review it - he loves db replicators 21:41:19 <notmyname> tdasilva: rledisez_: you two have been quiet so far in today's meetings 21:41:22 <notmyname> any thoughts? 21:42:37 <tdasilva> sorry...just multiple things going on here :( 21:43:20 <tdasilva> started looking at py3 work too, but i know that's not even on the high priority list atm... 21:43:34 <tdasilva> so yeah...too much going on... 21:43:41 <rledisez_> hum, not much to say honestly. pain points first, i agree feature/deep is a priority. consistency engine too :) 21:44:42 <notmyname> rledisez_: would love it if OVH could take a look at some of torgomatic_'s patches. I think your combined ops experience would be really good for it 21:44:54 <rledisez_> the object-updater would deserve some love too. i will submit something about prioritization of async_pending tasks 21:45:22 <rledisez_> notmyname: bookmarked ;) 21:45:27 <torgomatic_> that'd be handy :) 21:45:28 <notmyname> thanks :-) 21:46:27 <clayg> rledisez_: yes, that would be wonderful 21:47:00 <notmyname> ok, so to sum up, as I see the state of the world right now: feature/deep is big and the most important thing going on right now. we hope to propose a merge to master in about two weeks. rledisez_/ovh will look at torgomatic_'s patches. kota_ will land the last feature/s3api patches, but we won't prioritize merging it to master right now 21:47:24 <notmyname> is that an accurate summary of our last 45 minutes? 21:48:26 <notmyname> (again, silence is agreement! ;-) 21:48:30 <clayg> and also I want a pony 21:48:41 <notmyname> sorry, no ponies 21:49:11 <clayg> drats 21:50:13 <acoles> notmyname: yes it's accurate, although hard that we have so much good stuff vying for attention 21:50:24 <notmyname> the last comment I made in my summary about feature/s3api is something that kota_ is still collecting his thoughts on. I definitely want to make sure we account for his work 21:50:35 <notmyname> acoles: yes. very difficult 21:50:50 <kota_> notmyname: thanks 21:51:46 <notmyname> however, despite being very difficult, I'm glad it's this group of people working on it. you are what make swift amazing. thank you for all the hard work you put in 21:52:22 <notmyname> anything else to add from anyone right now? 21:53:36 <notmyname> all right. let's close the meeting thing. thanks for coming today 21:53:39 <notmyname> #endmeeting