19:00:42 #startmeeting swift 19:00:42 Meeting started Wed Mar 4 19:00:42 2015 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:00:43 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:00:45 The meeting name has been set to 'swift' 19:01:06 hello, who's here for the swift meeting? 19:01:11 o/ 19:01:12 hi 19:01:12 o/ 19:01:14 o7 19:01:28 ya 19:02:12 * notmyname gives people a moment to join 19:02:35 #link https://wiki.openstack.org/wiki/Meetings/Swift 19:02:39 agenda is there 19:03:01 the goal of this meeting to to review the EC status and make a call on our commitment to an EC beta in kilo 19:03:12 also, I've got a hard stop in 27 minutes :-) 19:05:08 I cannot answer the final question, sorry. Ask Paul/Sam/Clay, decide yes/no, done! Should be free way before the hard stop. 19:05:28 it's for all of us :-) 19:05:43 I should look how intrusive EC is and perhaps it's not bad wrt regressions 19:05:47 hi 19:05:51 acoles: hello 19:06:00 If it is, the sooner we pull it, the better 19:06:19 I have been talking to them, nearly every day. /me is trying to to be annoyed that they aren't here 19:06:23 *not to be 19:06:53 peluse mentioned something last night about traveling to Boston today 19:07:01 ah, right 19:07:02 not sure if that was after the meeting 19:07:28 and peluse just sent me a text message. wifi not working at the airport 19:07:41 ok, let's get started 19:07:49 #topic ec status 19:08:02 here's where we are, as I know it 19:08:18 the swift part of the code is going well 19:08:23 ok, I'm here 19:08:27 basice GET and PUT functionality has landed 19:08:29 peluse_: found some internets 19:08:29 peluse_: yay 19:08:40 found curt's PC :) 19:09:07 there will need to be some refactoring work on the read/write path, but that may not make it into a beta 19:09:14 so yeah the big review is the multiple fragment index one, thats important and needs to be next focus 19:09:16 there is some additional functionality that is being worked on 19:09:38 eg torgomatic is working on ranged gets and cleaning up some of the put path 19:09:46 agree on read/writre refactoring. we may want some tweaks before beta but beyond that and bug fix should be OK 19:09:57 peluse_: you mean https://review.openstack.org/#/c/159637 ? 19:10:17 not on my laptop so don't want to click away and get lost but the one you are working on with me, yes :) 19:10:38 peluse_: sure, thats the one then :) 19:10:41 clayg is looking at some of the bckground daemon processes. eg container-sync and the account-reaper have to be refactored to account for using internal proxy instead of direct calls 19:11:05 and the reconstructor will be back up and running this weekend. unfortunately I'm out of office til Sun 19:11:07 there will likely be somewhat of a mad dash at the end to get the background daemon jobs working 19:11:19 ok. so the reconstructor is what peluse_ is working on 19:11:33 peluse_: that's been split apart, right? 19:11:51 notmyname: great news on the rework of the daemons, glad clayg is on it 19:11:59 the multipl fragment index one (that acoles linked) 19:12:04 acoles: you're working with peluse_ on that? 19:12:14 yes and i can spend more time on it 19:12:20 thanks 19:12:29 had some comments from mattoliverau and clayg to follow up on 19:12:33 ok 19:12:35 tdasilva: I know you've been jumping in too. where are you looking right now? 19:12:52 notmyname: https://review.openstack.org/#/c/161030/3 19:13:10 ah, good. thanks 19:13:16 notmyname: container update of etag and size 19:13:17 but clayg let a comment in #openstack-swift about wanting to refactor into per-policy diskfiles, so i'm not sure whether to proceed or if he is going to do a big re-write 19:13:21 * notmyname stars that patch 19:13:22 s/let/left/ 19:13:40 acoles: ok 19:13:57 so we should talk more about policy diskfiles... make sure implications are clear to everyone 19:14:12 including me :) 19:14:36 I think clayg is looking at the b/g stuff first. but yes. I definitely want to understand that part 19:14:52 cool, maybe Mon we can talk 19:15:15 ok, what else is ongoing in ec-landia? mattoliverau? kota_? you've both been reviewing and writing code in there 19:15:29 zaitcev: are you looking at any ec stuff right now? 19:15:40 notmyname: not this moment, no 19:15:44 I've been reviewing and testing 19:15:49 ok 19:16:01 peluse_: notmyname : so should i hold off on multi frag patch until you've discussed per-policy diskfile? 19:16:04 zaitcev: if you have time to look at the multi FI patch (if someone can paste link) that would be rgeat. its diskfile guts changes 19:16:06 on Swift part, I'm not working, right now, but will start to review 19:16:14 * acoles doesnt wantto waste effort 19:16:14 kota_: thanks 19:16:23 on PyECLib part, almost done my work to add NTT plugin. 19:16:51 kota_: right. you've been working on pyeclib. good (/me will come back to that is just a moment) 19:16:58 multi FI == https://review.openstack.org/#/c/159637/4 19:17:36 peluse_: what do you think of acoles's question? per-policy diskfile discussion first or after? 19:18:07 sorry before or afer what? 19:18:19 before or after i do any more work on it :D 19:18:27 OK, I see it now. I think we press forward 19:18:36 until we change our minds :) 19:18:40 heh 19:19:04 peluse_: thanks for splitting up some of the reconciler stuff 19:19:04 lol, just what acoles wants to hear 19:19:18 it's a big piece, and the longer it stays there, the more worried I am about it 19:19:21 so seriously though, yeah I think we should keep with the current direction until we have good reaons to change and all agree 19:19:27 mattoliverau: right! 19:19:29 agreed 19:19:47 fwiw i was finding ways to hide policy specific stuff in the policy classes 19:19:53 and i think there is more scope 19:19:59 ok, last question about the swift code: has anyone been looking specifically at tests for ec? 19:20:16 its whether that is sufficient or we flip to multiple diskfile classes 19:20:24 acoles: yup, agree. little things like that can make full per policy diskfile less interesting 19:21:10 peluse_: ok i will press on with a philosophical attitude to the lifespan of anything i type :) 19:21:31 tdasilva: I moved your task on trello into "doing" 19:21:40 notmyname: cool, thanks! 19:21:44 ok, last moving piece is pyeclib 19:21:46 zaitcev: mattoliverau : ^^ i will spin another version of https://review.openstack.org/#/c/159637 tomorrow fyi 19:21:49 * tdasilva needs to check trello more often 19:22:01 * notmyname assumes nobody is yet looking specifically at test updates 19:22:09 thanks 19:22:13 I talked to tsg about pyeclib this morning 19:22:25 he's tracking down some dependency issues in ubuntu to get 1.0 available 19:22:33 acoles: cool 19:22:37 1.0 removes the "remote calls during install" issue 19:22:52 and basically will be needed for any packaging and a lot of testing 19:23:12 feature freeze for openstack kilo is in 2 weeks, so that's the deadline for getting that sorted out 19:23:42 so, all that being said.... 19:23:47 looking at dates 19:23:57 mar 27 is 4.5 weeks from today 19:24:07 that gives us 2 weeks for a merge-to-master 19:24:17 for an RC on april 10 19:24:35 cool... i think 19:24:58 can we commit to an ec beta (most features, but not "certified prod ready") in that timeframe? 19:25:04 thanks acoles for the huge help on the fi patch! 19:25:17 yes 19:25:18 notmyname: sorry, little confused. you said "feature freeze for openstack kilo is in 2 weeks" 19:25:27 tdasilva: sorry, dependency freeze 19:25:50 notmyname: got it, thx 19:27:07 #startvote Do we commit to an EC beta in Kilo? yes, no, delay 19:27:08 Begin voting on: Do we commit to an EC beta in Kilo? Valid vote options are yes, no, delay. 19:27:09 Vote using '#vote OPTION'. Only your last vote counts. 19:27:18 for delay, I mean, make that choice later 19:27:23 notmyname: you asked for commitment and it all went quiet ;) 19:27:26 yes to commit to kilo 19:27:32 no to have ec after kilo 19:27:32 lol 19:27:33 #vote yes 19:27:38 acoles: I see that 19:27:42 notmyname: now that is the question :) so long as its "beta" i say yes. i've been impressed in testingbasic put/pull. nice work everyone 19:27:57 #vote yes 19:27:58 it will definitely only be a beta 19:28:20 but if we commit to it, then we will cut scope to have something in kilo before we drop ec from kilo 19:28:21 #vote yes 19:28:43 and i think thats good we need real feedback on perf and usage even if were 100%done 19:28:45 I've been impressed with the momentum over the last two weeks and so I 19:28:46 #vote yes 19:28:48 to having the current best ec implementation we can muster up for a review to merge to master two weeks before the first RC get's cut 19:28:55 #vote yes 19:28:58 clayg: yes, that 19:28:58 #vote yes 19:29:02 #vote yes 19:29:13 * lpabon if my vote counts for anything :-) 19:29:20 acoles: cschwede: tdasilva ? 19:29:20 i'm trusting clayg peluse_ and other opinions, haven;t looked at state of tests 19:29:26 #vote yes 19:29:29 lpabon: you just volunteered to do reviews! ;-) 19:29:30 #vote yes 19:29:32 i think we can do it 19:29:35 lol 19:29:54 acoles: we desperately need to start working on tests - but there's still functionality we know is missing :'( 19:29:59 #endvote 19:30:00 Voted on "Do we commit to an EC beta in Kilo?" Results are 19:30:01 yes (9): notmyname, acoles, tdasilva, kota_, mattoliverau, lpabon, peluse_, clayg, zaitcev 19:30:13 no one opposed. that's great! 19:30:26 what was the diff between no and delay anyway? 19:30:37 delay = choose later 19:30:42 oh, delay the vote 19:31:09 tdasilva: delay always appeal to me when faced with a decision ;) 19:31:20 ok, I've got to run now. I'll be back in #openstack-swift in a while 19:31:32 acoles: understand you completely 19:31:34 :) 19:31:42 now the scary part starts with poepl outside of the devs start to get involved and plan around our commitment 19:31:51 :-) 19:32:11 thank you for coming and discussing. thanks for working on swift 19:32:15 #endmeeting