19:00:58 #startmeeting swift 19:01:00 Meeting started Wed Nov 19 19:00:58 2014 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:01:01 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:01:03 The meeting name has been set to 'swift' 19:01:04 who's here for the swift meeting? 19:01:08 o/ 19:01:11 o7 19:01:14 \o 19:01:14 hi 19:01:16 o/ 19:01:19 here 19:01:41 yo 19:01:50 welcome :-) 19:01:52 #link https://wiki.openstack.org/wiki/Meetings/Swift 19:02:14 several general things to go over this week and a few bigger ones 19:02:40 first up, meeting time. note (for those of us in the US) the meeting time is based on UTC, so it's an hour earlier now 19:02:53 still 1900UTC 19:03:16 #topic paris feedback 19:03:18 I was able to enter the meeting into Zimbra in UTC, so I'm good with reminders 19:03:38 what did you think of the paris summit? what was good, what was bad? 19:03:52 so maybe just me but I think the design session have gotten to big to be run as design sessions 19:04:04 peluse: agree 19:04:14 maybe 'tech talks' would be better, have fewer of them and more time alotted for open discussion/hackathon style work like Fri morning was 19:04:29 peluse: that's an interesting idea. I like it 19:04:59 peluse: i like that also 19:06:10 * notmyname is taking notes 19:06:12 what else? 19:06:32 scheduling? location? topics? content? conference part? 19:07:05 I think it felt a lot like the last summit wrt logistics, worked OK for me. Conf floor was a bit crowded but I suppose thats a good thing :) 19:07:47 i thought all in all it went well 19:07:53 I kinda expected more from the dev ops session, not sure if it was a timing thing where we left it for the last session of the day 19:08:00 I liked the venue better than atlanta (and many of the previous) since the summit sessions were in the same hotel and not blocks away 19:08:01 notmyname: i was kinda sad the wide-open Fri. morning thing was at the very end (also having so many different groups in one large room sucked) 19:08:26 swifterdarrell: yes, Cinder + Swift in the same room sucked 19:08:27 swifterdarrell: +1 19:08:33 swifterdarrell: aggree 19:09:00 tdasilva: so maybe getting more involved n the official ops tracks? my struggle there is that those have a lot of emphasis on nova/neutron 19:09:24 yeah...not sure...i think we basically only heard from briancline 19:09:34 I went to the "user survey feedback" session, and it was just "make nova and neutron not break" 19:09:57 tdasilva: yeah, I think we had a better ops session in the past 19:10:00 :/ 19:10:20 any ideas on how to make that better (the ops session)? 19:11:08 a structured set of questions to ask? commitments to get people in the room? 19:11:18 notmyname: create and maintain a list of swift operators and badger them to be there and participate? 19:11:20 humm...i was wondering if it was a timing issue where we only had developers left 19:11:23 it it even needed? 19:11:37 tdasilva: (seems quite possible) 19:11:48 maybe it needs to be earlier in the week 19:12:06 but that's a good question: is it even needed? helpful? can we learn anything from them? 19:12:44 I do think that ops session is needed and at least those I attended (didn't go to Paris) were useful for me as a guidance. I was this close to trying and making a whole new logger direct to files and ops told me to can it. That was guidance. 19:12:49 tdasilva: notmyname: if there aren't operators with feedback they want to give back to the devs, then it's not needed... but I find it hard to believe that's the case 19:13:30 zaitcev: +1 19:13:50 yeah, I totally think ops feedback is crucial. and I've liked in the past to have a forum. but there are other forums now (with arguable benefit). 19:14:19 IMO we should keep scheduling a swift-specific one, even if not specifically in the "swift track". and we attend and get others there 19:14:48 notmyname: +1 19:14:50 +1 19:14:56 what other feedback from the summit/conference? 19:15:06 food was great! :-) 19:15:12 not the conf. food btw 19:15:40 heh 19:16:06 https://www.youtube.com/playlist?list=PLIr7I80Leee7CSbRGL7q_3ftFD8a4i4dI <-- conference videos about swift. I think there might have been a few others that I need to add 19:16:16 s/think/know/ 19:16:30 tdasilva: ya, parisian food was very nice 19:16:49 next summit is in vancouver and the one after is in tokyo 19:17:00 anything else before we move on? 19:17:03 notmyname: you can add there my Spark thing, it also about Swift 19:17:20 gvernik: ya, there are at least 3 more that need to be added 19:17:36 notmyname: I also had a vbrownbag tech talk about swift + swiftonfile in case you want to add 19:17:56 https://gist.github.com/notmyname/d710d3775ad0d585c851 19:18:06 ^ raw list of youtube links 19:18:17 notmyname: some feedback on the openstack food. I was told that the Vegeterian selection was extremly lacking, again. Seems that OpenStack summit needs to put more effort around this.. I am not sure if you can talk to the "powers that be" about this 19:18:26 lpabon: ack 19:18:29 and noted 19:20:03 there have been some conversations (mailing list and meetings) about the summit. I'll try to jump in with your feedback. thanks! 19:20:28 please feel free to email me or say something in IRC if you have other comments 19:20:37 me@not.mn 19:20:45 moving on then 19:20:52 #topic OPW project 19:21:06 so there's this thing called the "Gnome Outreach Program for Women" 19:21:20 it's an internship style thing 19:21:32 where people work on a project in an open-source project 19:21:50 mahatic applied and was accepted to work on swift for the December-March OPW term 19:22:04 cool 19:22:07 mahatic: welcome and congrats! 19:22:11 and from this round its only "outreach program", everyone other than men are welcome to apply :) 19:22:21 o/ 19:22:22 mahatic: ah. thansk :-) 19:22:42 mahatic: congrats 19:22:42 thank you! a big one, for taking time out for the whole process! 19:23:07 mahatic will be initially working on a checker tool that a few of us have talked about. to probe the cluster to verify that things are good (eg read the ring and to OPTIONS requests to the servers) 19:23:14 thanks acoles! 19:23:57 mattoliverau, hello! have fun at your vacation 19:24:27 mahatic: any questions or other things that need to be brought up? 19:25:08 not really. I'll be hanging around quite a bit and will try to ask only sensible questions :) Please do correct me 19:25:19 me in any case* 19:25:35 mahatic: still on holidays, so not quite so good at timezones :p 19:25:56 #topic new specs guidelines 19:26:10 mattoliverau, yeah, got that :D 19:26:40 the swift-specs repo has been an experiment, and we put together some guidelines (actually written down now) at the summit 19:26:51 I hope you've seen them, but if not here's the summary 19:27:16 merge specs early and often. when you and a reviewer agree on something merge that part of it 19:27:51 if you +1/+2 a spec, you're giving a promise to review the patch when it comes around (or at least we can all bug you about it) 19:28:27 and, importantly, specs aren't to be the final docs for a feature. specs are a place to collaborate on design and are living documents 19:29:01 related to that, we wrote down that docs must land before or with a patch. you can't land a big patch with a promise to write the docs later (thanks swifterdarrell) 19:29:17 heh 19:29:39 that's the summary, and there is one small change to the organization of the specs repo 19:29:47 now there are 2 high-level folders 19:29:49 all sounds great 19:29:55 in-progress and done 19:30:28 "done" is for stuff that has had the related patch merged. think of it as a way to "archive" (pejoratively, trash) the document to get it out of the way for in progress stuff 19:30:53 so eg once EC lands (with its docs!) then the spec can move to the done folder 19:31:05 make sense to everyone? 19:31:14 yep 19:31:27 Yup, sounds great 19:31:53 thats what i remember we said in paris 19:32:00 importantly, keep in mind that the tool is to help us be better. don't worry about how other people are doing it, other than to learn what works and doesn't 19:32:16 so if something works, let's do more of it. if it doesn't work, let's change 19:32:20 very goodpoint 19:32:36 sounds great to me! 19:32:48 any questions about any of that? 19:32:52 Okay, I avoided specs until now, but I'll give it a try by putting the Account/Object-in-Kinetic into a spec. 19:33:17 zaitcev: cool. that'd be great 19:33:29 * notmyname didn't know you were working on that 19:33:43 what is that? 19:33:52 tdasilva: read the spec ;-) 19:33:55 curious to see the spec now 19:33:58 lol 19:33:59 (when zaitcev writes it) 19:34:08 uh-oh 19:34:22 lol 19:34:27 :) 19:34:45 ok, next up 19:34:52 #topic docs change 19:34:56 #link https://review.openstack.org/#/c/135102/ 19:35:27 the docs team (ping annegentle) has made some changes in how docs are organized in order to deal with the scale of the number of projects 19:35:51 summary is that some of the api docs are moving into project repos instead of being in the docs repos 19:36:11 that patch above is a copy/paste of the existing stuff into our swift repo 19:36:46 which means that (1) we'll be responsible for it and (2) it will be available from swift.openstack.org along with all the other dev docs we have 19:36:50 looking for a rubber stamp or want one of to double check the copy n paste? 19:37:27 peluse: mostly rubber stamp to get it moved. it does need to be reviewed for accuracy 19:37:35 rgr 19:38:04 ie if we're directly responsible for maintaining it now, we need to ensure it reflects what we believe to be true 19:38:24 thunk... (done). I'll read it while driving home later today 19:38:36 peluse: ! 19:38:44 hopefully not while behind the wheel! 19:39:00 :) Seriously though, I will buid and read and post anything that looks off as a new patch 19:39:05 thanks 19:39:18 * tdasilva was wondering if peluse has a chauffeur 19:39:31 ...swanky intel jobs... 19:39:41 #topic EC status update 19:39:44 :) free beer in the back of my limo 19:39:54 a flurry of activity around EC lately (yay!) 19:39:55 notmyname: did/does annegentle's team have tech writers that generate/maintain these docs or do we now get to write our best prose 19:40:20 peluse: can you give an update on current EC status and key patches to review? 19:40:34 I'm going to focus on not letting any lies in, the prose is for someone else 19:40:47 sure... trello is up to date https://trello.com/b/LlvIFIQs/swift-erasure-codes for those interested 19:40:51 acoles: I don't think they did. and we always get to do our best with the prose 19:40:55 zaitcev: ;-) 19:40:55 and the priority review page is also up to date 19:41:01 peluse: good 19:41:02 thanks 19:41:42 I'm in the middle of implementing some proto work for the portions of the reconstructor that we didn't anticpate being so complex - basically rebalance 19:41:55 clayg and torgomatic are both now working on it here (swiftstack) as a top priority 19:42:00 sounds like clayg is getting a jumpstart on GET which is awesome 19:42:24 ok 19:42:32 timing should be good because the reconstructor will depend on the GET patch for the next phase (actual reconstruction) 19:43:37 we've been doing EC a little differently that we did storage policies. SP had a "demoable" version pretty quickly and we had a tough slog at the end. EC is doing a lot of the very hard groundwork up front, but it feels slower because we don't even have read/write done yet 19:43:50 I hear ya 19:44:07 I'm not sure if that's better or worse, and I'm not looking for an answer right here right now, but it's somehting to think about 19:44:42 there's a shitload of behind the scenes stuff for sure. The big win right now is the dependency, pyeclib , the guys did a great job having that ready in time for us to start putting meat behind PUT/GET and the reconstructor 19:44:58 peluse: beyond "review EC patches", what's the current blocker? 19:44:58 notmyname: I posted an EC policy patch for review yesterday and have the .durable + HEAD stuff coming up (WIP posted yesterday) 19:45:06 tsg_: nice! 19:45:30 ah yes, tsg, I spaced that but that is also critical as the PUT path really isn't done w/o it 19:46:16 i hate "big bang" stuff like this but honestly the real functionality is going to happen in sorta that fashion.... 19:46:22 ok 19:46:25 unlike SP by its nature 19:46:54 I think the important thing is to properly set expectations. a big bang isn't bad if people know that 19:47:17 peluse: anything else on EC for today? 19:47:28 don't think so, no. any questions from anyone? 19:48:10 notmyname, just saw your question wrt current blocker... just reviews and coding right now... 19:48:14 ok 19:48:23 churn and burn baby 19:48:28 #topic priority reviews 19:48:31 #link https://wiki.openstack.org/wiki/Swift/PriorityReviews 19:48:42 peluse: i will try to carve out some EC review time over next few days 19:48:58 I updated this yesterday (monday?) with some of the active things that have been proposed since the summit 19:49:02 Kota and Gil are still on huh 19:49:12 the fsync() on dirs and the splice one are interesting 19:49:20 ya, kota's has one +2 19:49:22 acoles, thanks! concurrent GET would be a good/fairly quick one (on the prioirty review page) 19:49:35 peluse: k 19:50:00 and the migration middleware ( gvernik) tends to cause passions to rise 19:50:07 i am working to implement changes we discussed with Clay in Paris. Will submit another patch very soon 19:50:14 ok 19:50:42 so that being said, take a look at stuff on that page :-) 19:50:47 #topic open discussions 19:50:58 anything else to bring up and discuss here in the meeting this week? 19:51:00 peluse: thanks for working on the concurrent reads while I've been away 19:51:14 mattoliverau, no prob, I didn't do much :) 19:52:05 if there's nothing else, let's adjourn. 19:52:13 rock n roll... later 19:52:15 got here just in time 19:52:18 thanks everyone for coming. and thanks for working on Swift! 19:52:20 heh 19:52:26 #endmeeting