19:00:38 #startmeeting swift 19:00:39 Meeting started Wed Feb 18 19:00:38 2015 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:00:40 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:00:42 The meeting name has been set to 'swift' 19:00:47 who's here for the swift meeting? 19:00:49 o/ 19:00:53 me :) 19:01:00 me too 19:01:00 hola 19:01:17 👈 19:01:26 here 19:02:13 welcome everyone 19:02:24 #topic hackathon report 19:02:33 last week was great! 19:02:40 +1 19:02:49 thanks for coming 19:02:52 #link https://swiftstack.com/blog/2015/02/13/openstack-swift-hackathon/ 19:02:56 #link https://thiagodasilvablog.wordpress.com/2015/02/16/openstack-swift-mid-cycle-report/ 19:03:05 hello 19:03:05 tdasilva: ^^ had a really great writeup 19:03:27 I hope everyone made it back ok. for you on the US east coast, the weekend after you left it was about 70-75 here 19:03:36 :P 19:03:45 tdasilva: that's _above_ zero 19:03:48 ;-) 19:03:54 lol 19:04:00 80+ here 19:04:19 we are at 27 today! that's balmy 19:04:33 no wind, feels great 19:04:58 so a lot of great stuff done. a lot of patches landed. a lot of great conversations continued and started about new stuff 19:05:05 maybe CA will trade you some warmth for some water 19:05:56 anyway, just wanted to say thanks for coming and participating. and if you weren't there, you were missed 19:06:13 I want to cover one big thing this week: EC 19:06:18 #topic erasure codes 19:06:27 we're getting close to the openstack kilo release 19:06:39 it's about 7-8 weeks away (until we need an RC for it) 19:06:46 ie early/mid may 19:06:55 err, sorry 19:06:59 ie early/mid april 19:07:03 damn it 19:07:16 peluse: no rest for you! :P 19:07:26 so, here's what I'm thinking 19:08:07 first, at a high level, push to release an EC beta in the kilo release. this is something that works, at least from a functional and probe test sense, and would be ok in a lab environment 19:08:41 and has great value for performance characterization 19:08:46 of course 19:08:57 so what's needed to get there. first let's look at timelines 19:09:08 whoa, nice work tdasilva 19:09:17 if we allow 2 weeks for a merge to master, that pushes us back to march 27 19:09:21 that's 5 weeks from now 19:09:41 so in those 5 weeks, there's a few major things to get done 19:09:50 PUT path with the .durable files 19:09:53 GET path 19:10:07 some refactoring on those to make them supportable in a code sense 19:10:09 clayg: thx 19:10:14 how about I add a new column to trello (or color on remaining items, something to make them stand out) as "must haves" for beta? 19:10:17 (eg separate ECObjectControllers) 19:10:23 peluse: ya, that's good 19:10:33 then there's the reconstructor 19:10:43 oh yeah, that 19:11:13 if we have functional tests passing against an EC policy and probe tests passing against the reconstructor, I think we'll be in really good shape 19:11:28 tdasilva: s/week point/oppertunity to improve/ 19:11:46 it's a bunch of work, but I think it's possible to get done 19:12:08 I've redone the priority reviews page to add EC up at the top 19:12:09 #link https://wiki.openstack.org/wiki/Swift/PriorityReviews 19:12:14 agree, I'd like to schedule some online meeting times for reviews as well to make sure the cirical ones stay moving if that's OK 19:12:24 peluse: ok 19:12:33 also, note, I updated the EC section this morning so its current 19:12:36 thanks 19:12:52 I'll announce the dates/times in the channel and on trello, probably Fridays mid morning 19:13:14 peluse: are you thinking in IRC or phone? 19:13:20 phone 19:13:32 ok. yes please announce that in IRC 19:13:56 also looking for a small face to face in Santa Clara in Mar for at least those working directly on the required elements for beta 19:14:07 probably just 1 day 19:14:27 I tink the next 2 weeks are critical. if we have good momentum, then we have a good chance of getting EC done(ish) by kilo. if we don't make good progress, then it will be nearly impossible 19:14:50 so I expect to be able to make a call (as a group) in 2 weeks on an EC beta in kilo 19:14:54 I tink so too :) 19:15:04 which is really important to all of the non-devs involved in it as soon as it's released 19:15:14 so.... 19:15:20 what questions do you have? what concerns? 19:16:29 I'm all in. 19:16:33 (this is where you start talking and not me ;-) 19:16:45 I think its a sane way to move forward, lets try and get as far as we can in the next 2 weeks 19:17:09 mattoliverau: I think we have like 5 weeks to get proposed to master, 2 weeks to get merged, and then rc 19:17:33 yes 19:17:42 notmyname: will the propose to master be a chain, like with SP, or one big patch? 19:18:09 and everyone is good with a 'beta' tag where we are rock solid on the repl code paths that are touched by EC changes and clearly label EC policies as beta right? 19:18:19 acoles: we can do whatever is needed. but I was thinking a chain like last time 19:18:29 my big hope is that we get the on-disk data format right 19:18:33 acoles: tbd, but I think overall the chain idea "worked" but really only from a conceptual level - most people just reviewed the last patch and everything merged at once 19:18:36 we can always fix the code later 19:19:10 one important part, for people writing and reviewing, is that anything we write down for EC needs to be versioned 19:19:10 acoles: one thing well do different is put the doc patch at the end so people can just push up grammer fixes without worrying about making the rest of the change need a rebase 19:19:13 clayg: yeah, but 2 weeks we make the call 19:19:21 wrt chain vs one big patch, it may very well be just 2-3 patches (small chain) so I wouldn't worry about that too much 19:19:23 mattoliverau: oh god 19:19:27 eg new hashes.pkl? version it. new MIME doc/trailers? version it 19:19:44 version it 19:20:03 clayg: yeah! the rebasing up the chain a pain to keep track of what had changed where 19:20:04 there's bound to be a meme for "stick a XXX on it" 19:20:25 acoles: ya, we'll take the lesson from last time and put the docs at the end of the chain! 19:20:26 you could use birds for version designators, I suppose 19:20:38 version 'swallow', version 'kestrel', version 'albatross' 19:20:49 torgomatic: swift, crow release with the bluebird protocol? 19:20:54 heh 19:21:14 * torgomatic is especially fond of "albatross" for things that change the on-disk data format 19:21:14 torgomatic: also, that sounds like an amusing and terrible idea ;-) 19:21:16 the eagle has landed 19:21:31 acoles: in the openstack liberty release?! 'MURICA! 19:21:35 ... and ... next topic :P 19:21:40 ya :-) 19:21:59 #topic service tokens 19:22:08 acoles: you just added this to the agenda (after the meeting started! 19:22:19 #link https://review.openstack.org/#/c/137086/ 19:22:21 acoles is sneaky 19:22:24 acoles: ready for review? 19:22:27 ok, i just wanted to flag up that imho its ready for review 19:22:29 the agenda is eventually consistent 19:22:34 great 19:22:36 notmyname: took words out my mouth 19:22:43 review EC first. then service tokens ;-) 19:22:50 lol 19:23:10 but as usual with stuff that relates to keystone there's some keystone setup to do to run the func tests 19:23:11 acoles: thanks for working on that. it's a cool feature. and tell donagh thanks too 19:23:20 acoles: thanks for providing a script, that'll help heaps for us non keystone guys :) 19:23:23 so i have put copiou notes on a gist 19:23:30 copious 19:23:43 acoles: nice. we need to find all those keystone notes you have and get them on the wiki 19:23:45 * peluse needs to stwp away or just a few minutes... 19:23:54 and if anyone is up for reviewing but needs help then ping me 19:24:15 I slightly updated https://wiki.openstack.org/wiki/Swift so it auto links to any other page on the wiki that starts with "Swift" 19:24:25 * acoles realises he is up against EC for review-time ;) 19:24:53 * acoles should say "donagh he is up against EC for review-time ;)" 19:24:55 acoles: anything else on service tokens for today? 19:25:25 notmyname: no, just flagging it up - can i add it to priority review or is that presumptious ;) 19:25:32 ok, thanks 19:25:37 acoles: do it! 19:25:39 acoles: ya, under the EC section 19:25:47 also... it is presumptious - but I like that 19:25:52 +1 19:26:03 notmyname: under EC? 19:26:12 below, I mean 19:26:18 in the "swift general" section 19:26:26 notmyname: ah, that makes sense 19:26:43 i'm not so presumptious as to put it *above* :D 19:26:47 I don't think we'll likely have a release between now and kilo, so as we get closer, I'll start a section for the next release 19:27:17 #topic opther 19:27:20 #topic other 19:27:32 nothing else was added to the meeting agenda 19:27:38 do you have anything else to bring up this week? 19:27:58 one thing... let me grab the link 19:28:05 ok 19:28:34 as noted in #openstack-swift earlier, https://www.openstack.org/vote-vancouver/Presentation/openstack-is-doomed-and-it-is-your-fault is a potential presentation 19:28:50 go vote appropriately (I voted 3) 19:28:51 :-) 19:28:53 that's all I have 19:29:07 (no, I have no affiliation with the author of that presentation) 19:29:12 (it just looks fun) 19:29:18 and I think redbo will also really like it 19:29:28 notmyname: is "won't fix" the correct status for "not really a bug" -> https://bugs.launchpad.net/swift/+bug/1133069 19:29:29 Launchpad bug 1133069 in OpenStack Object Storage (swift) "Concurrent uploads to same object" [Wishlist,Won't fix] 19:30:00 in that case cschwede (correctly) noted that 'this can't be "fixed" in the general sense' 19:30:26 ya 19:30:29 but redbo had previously marked it as "meh, doesn't seem so unreasonable for the object server to do this best effort, but it's not a big priority" 19:31:05 clayg: there's also "invalid" as a status 19:31:23 and I agree that while *I'm* not planning on working on it, if someone wanted to figure out how to make hashdir cleanup return a boolean or something so the server can respond with 4xx if a more new timestamp slipped in while uploading to /tmp - it'd *review* it? 19:31:27 but won't fix is probably better IMO. "invalid" almost seems like it's not a thing 19:32:15 hrmm.... i hate "won't fix" it's like the most asinine wording - like gerrit used to be "do not merge this" 19:32:27 clayg: that sounds like a great patch for after EC ;-) 19:32:39 hell, throw it in the EC bucket :) 19:32:48 lol 19:32:59 peluse: no! cut scope for a beta! 19:33:13 anything else to bring up? 19:33:21 lol, it's not a priority to me at all - but mattoliverau was talking about cleaning up the bug list and christian was griping he never knows what's the right status - so I was wonder if there was a same page for us to get on? 19:33:37 notmyname: I don't want the spec's repo to be an "ideas board" -> https://review.openstack.org/#/c/151335/ 19:33:41 as torgomatic the summit session voting is ongoing. there's some cool swift stuff 19:33:55 clayg: ya I agree with that 19:34:13 i don't care for the new interface, i keep loosing my place when I'm trying to work my way down a list of topics from a search term 19:34:23 as discussed last week, we need to be better about bug triage. I'm working on that 19:34:29 clayg: new tabs are the only answer 19:34:44 and we'll get to the cleanup and status stuff 19:36:32 if there's anything else, please mention it soon. otherwise I'm closing the meeting in about 30 seconds 19:37:06 short meetings are best meetings 19:37:38 thanks everyone for coming. thanks for working on swift 19:37:40 ya 19:37:42 #endmeeting