21:00:11 #startmeeting swift 21:00:12 Meeting started Wed May 27 21:00:11 2015 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:14 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:17 The meeting name has been set to 'swift' 21:00:17 who's here for the swift meeting? 21:00:20 o/ 21:00:21 o/ 21:00:22 o/ 21:00:23 rock-n-rolla 21:00:24 o/ 21:00:28 hi 21:00:29 hi 21:00:41 hey 21:00:44 o/ 21:00:50 welcome :-) 21:00:51 Yay tdasilva and mrda, 2 of my favorite peeps :) 21:01:00 :) 21:01:05 he 21:01:08 o/ 21:01:10 mattoliverau: hi! 21:01:16 * notmyname doesn't have a face with mrda 21:01:19 mr davies? 21:01:26 notmyname: that's me 21:01:37 my 4th grade math teacher? 21:01:43 he 21:01:46 lol 21:01:47 that's scary 21:01:51 notmyname: he ate in your house in SFO :) he had a face then 21:02:00 I hope 21:02:06 hi 21:02:09 It was required. No face, no mexican :) 21:02:19 o/ 21:02:26 welcome to our first meeting at 2100UTC 21:02:37 lol 21:02:52 I'm happy to see everyone here, and I hope the new meeting time works better for those of you far away from me 21:02:55 yay, first sleep in meeting! 21:03:02 thanks cschwede and acoles for staying up late 21:03:09 +100 21:03:13 o/ 21:03:13 :) 21:03:17 no problem 21:03:21 np! 21:03:30 wow, big crew today - awesome! 21:03:45 agenda for today: 21:03:48 https://wiki.openstack.org/wiki/Meetings/Swift 21:03:50 #link https://wiki.openstack.org/wiki/Meetings/Swift 21:03:57 #topic summit follow-up 21:04:13 thanks everyone who was at the summit. you helped make it great! 21:04:15 hi 21:04:22 and if you weren't there, you were missed 21:04:26 * notmyname looks at tdasilva 21:04:39 * cschwede nods 21:04:42 to start with, I want to get some feedback on last wekk 21:04:43 :) hopefully next time 21:04:45 lol 21:04:47 what did you like? 21:04:50 I really liked the new format on the design side (muchbetter than Tokyo). Favorite sessions was genomics+swift with Joe 21:05:11 soup dumplings were good ;) 21:05:17 peluse: you've already been to the tokyo summit? 21:05:20 peluse: you mean Paris? or time-travelling, is it getting worse? 21:05:22 as were the margaritas 21:05:23 peluse: tokyo? back from the futrue 21:05:24 ? 21:05:32 I liked meeting several of you - y'all were so friendly! 21:05:34 tokyo, hong kong. whatever 21:05:36 better meeting spaces 21:05:42 paris I meant, ahh never mind 21:05:48 yeah, I liked the general format better too 21:05:57 the work sessions and the friday open time were very good 21:06:07 Yeah the workrooms were great, all they needed was a projector 21:06:12 right 21:06:16 mattoliverau: +1 21:06:20 was close to perfect. more chairs on the friday room would be great, and a beamer 21:06:25 mattoliverau: he didnt ask for negatives yet ;) 21:06:27 what did you not like and what do you want to see changed for tokyo? 21:06:41 lol 21:06:50 projectors in tokyo please ! 21:06:58 I guess fishbowl vs work session wasn't much different for us other than whether the rom had a projector or not... 21:06:59 more tap on the table 21:07:04 what acoles said :) 21:07:04 wishlist: a swift exabyte deployment keynote on monday morning :) 21:07:22 cschwede: oh can you talk about some prod clusters we don't know of yet? ;-) 21:07:32 torgomatic: man you didn't waste no time with simplejson 21:07:37 better network connectivity (wifi) 21:07:54 yeah so on the projector thing. clarity on whether all rooms will have it or just fishbowls so we can plan accordingly (for tokyo) 21:07:55 clayg: nope :) 21:07:56 ho: ah? I didn't have any problems 21:08:25 notmyname: sometime it's very difficult to connect network. 21:08:43 ok, good to know 21:08:51 Aside from the venue, I was impressed with how many swift talks there were, a great sign for the health of our project. 21:08:56 I agree about the need for a projector 21:09:03 mattoliverau: +1 keeps going up :-) 21:09:26 I thought the Ops session went much better then last year, we have a bunch of ops releated feed back to work on. 21:09:28 so I thought that the work sessions were too short. too fast of context switching I think I heard from someone 21:09:34 yeah, the ops session was great 21:09:35 19 swift talks by my count 21:09:42 8 cinder, 8 ceph 21:09:56 notmyname: we did seem to run over into most breaks 21:10:13 (9 cinder) 21:10:18 yeah, having double work sessions (and potentially fewer fishbowls) might be good 21:10:19 yeah, were obviously a talkative bunch :P 21:10:20 notmyname: yeah, we seemed to have to stop just as we got to the meat 21:10:29 joeljwright: yup, I felt that way too 21:10:38 * clayg assumes mrda's comment about friendlyness applies to the *rest* ya'll 21:10:48 lol 21:10:54 *double* work sessions! 21:11:01 you too clayg, your grumpy in a nice way :P 21:11:06 like we do container sharding twice as hard!? 21:11:08 *triple* work sessions!! 21:11:25 lol 21:11:28 I found friday discussion very useful, have more of that format 21:11:34 notmyname: if we need a projector it's not a work session 21:11:39 we just need to apply EC to work sessions… 21:11:45 notmyname: I just think some of the discussions benifited from the smaller rooms 21:11:46 clayg: that's not always true. eg code reviews 21:11:56 notmyname: good point, i was wrong 21:11:59 or ones where a picture helps everyone get on the same page 21:12:01 especially for us blind guys 21:12:15 clayg: we could always have done with the etherpads on a screen 21:12:48 joeljwright: that's only helpful is someone doesn't know how to eatherpad (not that *I* know anyone like that) 21:12:54 ok, so what was your favorite session (tech or conference) during the week? 21:12:55 :D 21:13:00 i would like to see work sessions earlier in the week, some people already left Friday afternoon due to flights, and everyone seemed to be a little bit tired (me included) after a packed week 21:13:01 ya'll are right - I take back everything I said about sreens in working sessions 21:13:09 but I do still think the fishbowls were too big 21:13:10 hey joeljwright is up late too :D 21:13:27 cschwede: oh yeah. I was super tired by friday and just enjoyed sitting and chatting with you 21:13:28 :) 21:13:38 genomics: https://www.youtube.com/watch?v=_u91VbQyltU 21:13:45 peluse: yeah, that one was fun 21:14:05 clayg and i were the last ones standing on friday at 5pm 21:14:09 on the tech side I really enjoyed the ops sessions and the other ops one (from softlayer) 21:14:15 I think there's like some sort of perhaps historic reason to do the work/hack days at the end of the week? 21:14:22 acoles: clayg: somehow that doesn't surprise me 21:14:31 heh 21:14:40 i like this https://libertydesignsummit.sched.org/event/51f4e0a00d70a063dfd5d99cd6cf5ccd#.VWYzoc_5fIU 21:14:42 it was quite intense doing 3 solid days of design/work sessions though 21:14:45 Talk wise, I actaully enjoyed all the metric talks (donagh, cschwede, etc) nice work all 21:14:49 in tokyo, there is 100% overlap between the conference and summit. IIRC the whole thing is 4 days long 21:15:01 yipes 21:15:05 mattoliverau: thx! 21:15:38 any thoughts from anyone who was there for the first time? 21:15:39 that's going to be 4 intense days 21:15:52 mrda: no kidding 21:16:01 anythign we can do better to prepare for the next one? 21:16:01 I guess we just drink on the fifth day then? :) 21:16:28 drink a fifth of what now? 21:16:41 I was there first time, thursday sessions were short ('fishbowl'?), friday discussion I found very useful 21:16:50 i think we were pretty well prepared 21:16:53 yay friday! 21:16:59 notmyname: my storage architect has already quoted you from the hearding cats talk several times 21:17:01 slavisa: +1 21:17:06 jrichli: uh oh :-) 21:17:19 we have cats? 21:17:21 jrichli: everything notmyname says is *pure gold* 21:17:22 jrichli: +1 21:17:24 thats a good thing : encouraging lots of communication with community 21:17:27 notmyname: not thought of any better preparation ideas, i think the agenda prep on etherpad went ok 21:17:30 I liked that people were open when approached to discuss ad hoc topics as well 21:17:39 jrichli: basically most of the time the SwiftStack office only communicates in notmyname quotes 21:17:45 imkarrer: you need cats to take pictures of cats to store pictures of cats. Or at least I thought so. 21:17:59 ok, I feel we should move on.... 21:18:03 what did you guys think of the panel you held on monday? 21:18:09 lol 21:18:14 notmyname: you could just /kick me 21:18:17 clayg: ic :-) 21:18:23 tdasilva: I thought it went well. I was really happy with how many people were there 21:18:34 actually surprised it was so crowded 21:18:39 140+ 21:18:41 tdasilva: shameless self-praising: i think it went well, i liked it 21:18:49 tdasilva: the loud guy in the middle was annoying 21:18:59 LOL 21:19:03 lol 21:19:07 clayg: you always stand out :P 21:19:15 cschwede: did you get anything abotu a recording of it? I haven't seen a link anywhere? 21:19:15 I think it went well, nice to put names to faces for newbies like myself 21:19:15 we should have all wore cat tshirts as a segway into notmyname's talk 21:19:18 i liked the dudes with the accents 21:19:24 you all were great! 21:19:24 notmyname: one sec 21:19:34 clayg had a youtube link? 21:19:35 peluse: yes. I loved that we have people from everywhere 21:19:45 notmyname: https://www.youtube.com/watch?v=z0pF69yMNAI 21:19:51 yeah once cschwede told me it was on youtube a search found it for me np 21:19:54 cschwede: thanks! 21:20:21 ok, anything else for summit follow-up? 21:20:30 but really - it wasn't nearly as interesting as joe or donagh's stuff 21:20:44 but the accent award has to go to Duncan on Cinder, have you hear him say "sucks"... "sooooks" 21:21:11 peluse: heh 21:21:22 #topic summit decisions and TODOs 21:21:30 so there were quite a few of these 21:21:36 first up, no more py26 21:21:43 * peluse has to bolt... catch up y'all later 21:21:44 thanks again to RAX for supporting that 21:21:55 \o/ 21:21:58 ding dong, 2.6 is dead 21:22:03 peluse: o/ 21:22:11 so torgomatic has already submitted patches for it 21:22:19 heh - talk about chompin' at the bit 21:22:21 I had a patch merge today that drops the py26 tox env 21:22:31 so that's the "official" end of it 21:22:48 #info py26 is no longer supported in swift 21:23:03 next up, and as you can see at this very moment, meeting time has changed 21:23:09 well, I have a hard problem that's giving me fits, so I'm doing this other stuff instead to produce the illusion of productivity ;) 21:23:12 I suppose if you're here now, you already know that 21:23:13 yay 21:23:22 notmyname: what's up with py3? earlier I saw a bunch of patches re py3.... 21:23:33 yeah, I wanted to bring that up in a bit... 21:23:39 ok 21:23:49 do we still need py26 in swiftclient? 21:24:04 joeljwright: you tell me ;-) 21:24:34 joeljwright: actually, I think it's better to keep support there than in swift. but I'm also ok with dropping it 21:24:46 it might cause issues 21:24:48 I don't have any numbers on how many swiftclient users use py26 21:24:59 no way swiftclient is successfully making py3 progress while still supporting py26 21:25:19 yeah, that's a big issue (from the code side) 21:25:30 I'll have a quick look at linux distros python versions 21:25:34 thank 21:25:36 s 21:25:39 scientific linux/centos worries me 21:25:56 yeah, cent/rhel6 has py26 21:26:09 i really didn't think swiftclient still supported py26 - but oh kay 21:26:17 clayg: yeah, me either 21:26:30 joeljwright: can you look in to it and report back 21:26:38 yeah np 21:26:41 thanks 21:26:47 last I loked py26 was still a swiftclient gate 21:26:51 ah 21:27:09 notmyname: clayg: it was a year or two ago, but we had someone with CentOS 5.x systems needing to run python-swiftclient (where 2.6 is the *new* python for the distro version) 21:27:23 I'm hoping to have some midcycle/hackathon details shortly (in a matter of days), so stay alert for that 21:27:24 swifterdarrell: wow 21:27:40 notmyname: clayg: and RHEL/CentOS 6.x still have some legs on 'em, right? 21:27:46 joeljwright: yeah i guess i was wrong, i see py26, py27 and py33 checks on python-swiftclient - crazy 21:28:12 ok, next up for post-summit TODOs... 21:28:28 I've gone through the etherpads and consolidated some todo items from them 21:28:43 then I took those and put most of them onto the priority reviews page 21:28:44 #link https://wiki.openstack.org/wiki/Swift/PriorityReviews 21:29:02 so the high-level bullets are "stuff people are working on" 21:29:07 and that's a big list 21:29:53 my next step is to consolidate more of the TODOs and get them summarized onto a wiki page (or somewhere) and put some relative priority to them 21:30:06 this is awesome: "python-swiftclient improvements: lots" 21:30:14 lol, yup 21:30:24 there was a really good list put together in that working session 21:30:38 stuff that's annoying, stuff that needs improvement, stuff that's just broken 21:31:20 just little things like auth, docs, interface… 21:31:22 :S 21:31:42 so question for everyone, how do you want to see all the info? are you looking for whatever I come up with? or do you have strong feelings on how it should be done? 21:31:57 ie all the ideas/todos/etc with some priority 21:32:17 notmyname: i think that list is useful 21:32:23 +1 21:32:47 could be a frame for adding priority reveiws as they come and go? 21:32:48 I am open to whatever: I just want to be able to add a few things I never got to add to the etherpads 21:33:10 yeah, i'm thinking of consolidating it with the ideas wiki page. just need to communicate some difference between "wouldn't it be nice..." and "this is being done" 21:33:45 but yeah, definitely having the high-level picture of stuff going on and decided is a great way to frame the priority reviews. I agree acoles 21:33:56 I like how priority reviews in gerret are starred by you in the dashboard. and haveing a list with priorities and ideas in the wiki works well in conjunction 21:34:36 the last piece is the triaging and tagging of the launchpad bugs. after getting this list together, i'm going to start working on the LP bug list 21:34:46 notmyname: can I add fast-post and migrating keystoneauth to policy.json (ho) ? 21:34:56 acoles: yes 21:35:12 acoles: thanks! 21:35:32 my goal is to have a good picture of what people are talking about, working on, and needs to be done so that everyone can see it 21:35:33 acoles: so much things 21:36:01 yeah, I didn't get several things that were discussed on friday 21:36:08 notmyname: should that list be constantly updated by the people working on those things or are you going to update it? 21:36:13 goodness we left of part power too - so not going to get to all of this 21:36:30 tdasilva: everyone can, I'll mostly do it 21:36:36 tdasilva: kinda like it works now :-) 21:36:56 clayg: I left that off intentionally. the conclusion last week was "interesting but not really a problem anyone has today" 21:37:07 we should link to specs on the wiki when there is one too, cause then people can see ideas that have already have had some motion, even if stalled so they know how to start work (if there interested). 21:37:09 notmyname: wfm 21:37:23 we should land the spec, but nothing to be done yet, at least not before the other stuff 21:37:33 speaking of....we gotta get better about specs 21:37:51 if an author and reviewer agree on something in a spec, land it! 21:37:53 notmyname: how did we not brainstorm on this while we were in summit? 21:38:02 clayg: which part? 21:38:12 which means that specs should start small and then get bigger as they get fleshed out 21:38:15 clayg: they all left early remember? ;) 21:38:19 notmyname: how to make specs more lightweight 21:38:32 clayg: epic fail? 21:38:36 idk 21:38:58 I think the whole idea of specs may have been half-cocked 21:39:05 it's an expirament 21:39:18 on one hand I really like that people can say "what about this idea" before having to write some code 21:39:30 it's to facilitate communication. let's change it as needed and use it as helpful. it's not a thing that must or must not be done 21:40:04 notmyname: yeah, i'm just curious if we can point to spec that was like "wow, a++ would buy from again" 21:40:32 "spec contained live bobcat" 21:41:37 for now, I'd like to continue specs. let's give it another month and see where we are 21:41:48 and if I'm not bugging people about them, then bug me abotu that :-) 21:41:55 (clayg is good about bugging me) 21:42:22 notmyname: thanks for working on the priority reviews/ideas page, more visibility will be awesome. 21:42:22 i think they are useful to see what people are thinking and what new features are on radar but agree we need to get away from them being never-landed 21:42:40 acoles: yeah, that's my thought too 21:43:01 maybe land a small spec with a rough idea first, and then work on the details? 21:43:03 * acoles knows thats partly my fault 21:43:23 Do we need a swift spec rule page on the wiki? how our process works, in_progress vs. complete and land fast? 21:43:36 cschwede: yeah, I think that's the idea we've got written down. 21:43:57 mattoliverau: I think there's something like that *IN* the specs repo? 21:43:57 https://github.com/openstack/swift-specs/blob/master/README.rst 21:44:08 mattoliverau: could probably be updated refined and elaborated on 21:44:23 yeah, but put it somewhere where new contributors will look. 21:44:37 mattoliverau: the truth was when we started with specs a) we didn't know what we were doing b) no one seemed to be doing anything we liked that we could take hints from 21:44:41 notmyname: i thought of something even more simple thant currently stated in the readme 21:44:58 I think the issue is that incompleteness should be addressed, but should not prevent acceptance of what is there 21:44:58 I think it's because we're all generally _really_ picky about patches that go in to swift, so we dont' like landing patches anywhere that feel "not finished" 21:45:08 so no technical implementation details in the beginning to speed up accept/deny 21:45:31 clayg: I agree, I just want to make it clear so when we tell someone to write a spec there is something to point too. but yeah, lets start with an updated README 21:45:34 notmyname: I do think we could clean up the spec template to get rid of a *bunch* of junk and focus the conversation more on the use case - operations implications - interation with other features - stuff that matters to us 21:45:38 cschwede: ah interesting 21:45:56 clayg: liek what cschwede is saying? don't do technical design in the spec 21:46:12 at least not in the first commit 21:46:20 I like that idea. avoids the waterfall design trap of specs (/me feels other projects' specs are tending that way) 21:46:39 it's not like a spec makes us legally bound to implement it or anything... if the spec looks good but it turns out that any possible implementation is too horrible to contemplate, then we just won't do it 21:46:45 notmyname, cschwede: just not sure that is the main issue... 21:47:04 so specs do make nice progressions, it's just that they sit in review for a loooong time 21:47:19 tdasilva: well, i think sometimes the idea of a proposed spec is ok, but the technical details take a huge amount of time 21:47:20 most specs do start pretty small, no? 21:47:25 torgomatic: yeah we should write something down on the process to back away from in-progress spec slowly just so we have an escape hatch clearly marked 21:47:30 ... even if we never use it 21:47:36 I think noone lands them cause people look for specs in review rather then in there landed location 21:48:10 cschwede: yeah I think avoiding implementation minutia is a *great* way get some agreement 21:48:24 the symlinks spec is a good example of some but not complete implementation detail (no offence to torgomatic because thats fine) but we can say 'nice idea lets do it' 21:49:40 ok, so to summarize what i'm hearing, specs take way too long to land, tech details up front cause it to slow down, discoverability of landed specs may be an issue 21:50:14 yup, so all that are communication issues.. which we can solve :) 21:51:29 so where there is a landed spec, link to it from priority review page section 21:51:32 #info mattoliverau to solve specs 21:51:35 did someone already say that? 21:51:49 acoles: I think that's a good idea. and mattoliverau implied it, but you said it 21:52:05 if it's in review, it should probably be linked to, right? 21:52:13 clayg: me and my big mouth :P 21:52:18 tdasilva: we do have specs right at the top of the review dashboards 21:52:21 mattoliverau: turn your abandoner bot on its head and give us an auto-lander-bot ;D 21:52:49 acoles: this spec will automatically be merged in 5... 4... 21:52:59 acoles: lol, auto-lander.. I could do that :P 21:53:27 can I somehow sneak in swift patches? 21:53:28 mattoliverau: can you add the landed specs (in the in_progress) dir to the top of the priority reviews wiki? could you do that today? 21:53:29 if there is a +1 and sitting there for > 1 week 21:53:40 no auto-landing yet :-) 21:53:46 notmyname: sure 21:53:49 mattoliverau: thanks 21:53:58 ok, anything else to bring up in the meeting this week? 21:54:32 ok, then let's call it 21:54:39 Thanks everyone for agreeing to move the meeting forward.. I'm much more awake 21:54:44 :-) 21:54:46 :) 21:54:47 :) 21:54:51 I can tell. you're much more talkative 21:54:52 clayg: mattoliverau : it would keep us reviewers on our toes :) 21:54:55 and cschwede is quieter ;-) 21:54:55 and apparently volenteer for much more :P 21:55:07 mattoliverau: right! 21:55:14 thanks for coming today, and thanks for working on swift 21:55:18 #endmeeting