21:00:52 #startmeeting swift 21:00:53 Meeting started Wed Jun 8 21:00:52 2016 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:54 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:56 The meeting name has been set to 'swift' 21:01:00 hello. who's here for the swift team meeting? 21:01:06 o/ 21:01:07 o/ 21:01:09 yo 21:01:09 hi 21:01:11 o/ 21:01:11 \o/ 21:01:12 o/ 21:01:12 hi 21:01:13 o/ 21:01:17 o/ 21:01:27 o/ 21:01:34 evening 21:02:09 * onovy 21:02:46 welcome, everyone 21:02:53 o/ 21:03:00 agenda this week is 21:03:02 #link https://wiki.openstack.org/wiki/Meetings/Swift 21:03:04 here 21:03:19 most of these topics should go pretty quickly 21:03:29 #topic 2.8.0 release 21:03:40 release has been tagged and approved (https://review.openstack.org/#/c/326771/) 21:03:51 well, will be tagged whenever the bots get done with that patch 21:04:24 https://github.com/openstack/swift/blob/master/CHANGELOG has all the highlights in it 21:04:38 thanks, everyone, for all the great stuff in this release 21:04:53 any questions about the release? 21:05:16 ok 21:05:28 #topic bugs arising from transpations 21:05:35 #topic bugs arising from translations 21:05:49 patch 323950 is still in progress 21:05:49 notmyname: https://review.openstack.org/#/c/323950/ - swift - Refactor locale tests and fix unicode issue 21:05:54 I think that's the only update there 21:06:10 cschwede: anything else about that one? (or acoles or timburke) 21:06:35 acoles and timburke had a look at it; that was really helpful - thx! apart from that - needs reviews 21:06:41 ok 21:06:45 thanks cschwede for working on a fix and test improvement 21:07:12 my french vocabulary has increased :) 21:07:15 heh 21:07:18 #topic rolling upgrade testing 21:07:27 cschwede: same status here? waiting on -infra reviews? 21:07:35 anything we can do from our side? 21:07:41 nothing new, i rebased my patch, passes, and yes - just waiting 21:07:56 https://review.openstack.org/#/c/297311/ 21:07:56 notmyname: patch 297311 - openstack-infra/devstack-gate - Run swift services on subnode 21:08:01 that's the one, right? 21:08:10 no 21:08:23 patch 304465 21:08:23 cschwede: https://review.openstack.org/#/c/304465/ - openstack-infra/devstack-gate - Use subnodes for Swift storage nodes in a multinod... 21:08:28 I always click the wrong one 21:08:51 no worries, my comment on 297311 links to the other :) 21:09:17 ok 21:09:21 #topic hackathon 21:09:23 the important test is gate-tempest-dsvm-multinode-full 21:09:26 here's a fun topic 21:09:31 cschwede: ack 21:09:49 hurricanerix is hard at work setting up logistics for the hackathon next month 21:09:58 details are at 21:10:00 #link https://www.eventbrite.com/e/openstack-swift-july-2016-hackathon-tickets-25913773795 21:10:18 you can register there so we can properly plan for who will be there 21:10:39 hotel info should be added Real Soon Now. waiting on the bigwigs at rackspace to sign some papers, I think 21:10:43 hurricanerix: any updates/ 21:10:44 ? 21:11:07 also, my email filters have been kinda weird lately, so if you emailed me any questions and i haven't replied, just poke me in irc to make sure i saw it. 21:11:31 no updates yet, still trying to get the hotel finalized. should be soon. 21:12:11 ok. i'll add that info and email everyone (via eventbrite) as soon as the info is published 21:12:23 any questions for myself or hurricanerix about the hackathon? 21:13:01 are there any other hackathons planned for later this year? 21:13:42 mathiasb: likely not. the openstack foundation is organizing a big "joint hackathon/midcycle" thing between barcelona and the following summit 21:13:54 the Project Team Gathering, IIRC 21:14:02 in february, but details haven't been finalized 21:14:13 sounds like a Highlander movie 21:14:24 joeljwright: I'll bring my sword 21:15:11 is it for improving cross-collaboration and solving our problems with gevent/uvloop? 21:15:34 nadeem: of course ;-) 21:15:45 ;) 21:15:49 ok, moving on then :-) 21:15:54 #topic crypto update 21:16:07 I've been hearing great things from acoles and jrichli__ this week on this topic! 21:16:33 last critical patches to feature/crypto have landed? and acoles has a review branch all queued up? 21:16:44 I just removed the final must-have review form priority review list, meaning we should be ready to propose to a crypto-review branch 21:16:54 fantastic! 21:17:13 mahatic timburke and jrichli__ have been finding improvements but I think we are "stable" enough 21:18:04 next step is for acoles and myself to get the review branch proposed to master 21:18:08 of course I anticipate plenty of comment and change on crypto-review :) 21:18:27 here's how I think that's going to look 21:18:46 acoles will propose the patch chain to master as a set of patches with one merge commit into master 21:19:07 the individual patches will each be a "step in the journey" to help you understand the whole 21:19:35 you can either review each patch or just the whole thing at once (ie the merge commit diff), but it's the merge commit that will be approved and landed to master 21:20:10 notmyname: that sounds different, I thought we'd use another feature branch? not that I have any preference. 21:20:15 I'll likely put a -2 on the merge commit initially to "plug" it until we've got enough reviews on it 21:20:59 acoles: no, I wasn't meaning for it to sound different. I didn't mention "new feature branch" because I wanted to experiment with something first (ie do we actually need it), but the process should be the same regardless 21:21:19 so, yeah, it's very likely all this will be in a feature/crypto-review branch 21:21:47 what is the purpoae of a second feature branch? 21:21:53 notmyname: ok. you'll have to teach me the "merge commit" part. 21:22:23 clarkb: it's what we've done in the past for storage policies and erasure codes. it's the simplified versions of the work that's been done 21:22:33 I see 21:22:41 clarkb: didn't we have this same conversation last year about ec and the year before about polices? ;-) 21:22:44 basically the reviewable representation 21:22:48 maybe 21:22:54 right 21:23:09 clarkb: but, we *might* not need to do that. I want to check something first and see what the gerrit experience is 21:23:32 if there isnt' an upstream branch and it's proposed as a merge commit 21:23:41 ok, moatly just curious. either way should work 21:23:42 clarkb: yes, a short chain of patches that comprise the final state of the original feature branch, that is easier for reviewers to digest. 21:23:58 acoles: actually, the process should be exactly the same as the merge-to-master stuff. just swapping the source and dest 21:24:20 notmyname: ah, i wondered if that might be the case. 21:24:30 * acoles has learnt to remember '-R' :) 21:24:43 clarkb: if we need it, I'll likely be coming to -infra with that request in the next day or two 21:25:03 acoles: are the patches ready to go locally for you? 21:26:08 notmyname: I need to add the recent tweaks, basically timburke's recent patch that is in flight to merge. But that shouldn't take long. So tomorrow I will have them ready. 21:26:15 ok 21:26:29 so for everyone.. 21:26:43 like with the previous big merges, we'll all need to review this one 21:26:52 if you weren't aroudn for EC or storage policies, that's ok 21:26:56 here's how it works 21:27:33 we have a soft-freeze on master. unless there's something critical and/or everyone (especially acoles) knows about it, don't review land stuff on master until the crypto branch lands 21:27:45 everyone review the crypto branch 21:28:07 leave your reviews as normal, and acoles will be managing the patch chain in gerrit 21:28:48 a feature branch like this should take more than the normal 2 +2s. we want to make sure we've got a lot of eyes on it and a lot of people understand what's going on 21:29:03 +1 to that 21:29:17 notmyname: the soft-freeze date is not known yet, right? 21:29:28 cschwede: getting to that :-) 21:29:51 of course, since a lot of people have worked on this feature, we'll get the "I can't approve my own code" viewpoint. but leave your review anyway. nobody did 100% of the crypto work 21:30:27 and if you're not a core reviewer, your reviews still matter a lot. please leave lots of comments 21:30:52 basically, before the crypto branch lands, we want a lot of people to have looked at it, understand it, and be happy with the changes it's making 21:31:16 I'm expecting that the review process will take about 2 weeks. at least that's the initial budget we'll allow 21:31:36 I don't expect that we'll have enough good-enough reviews sooner than that 21:31:49 and I hope we don't have to extend that too much (or at all) 21:31:57 but in two weeks we can re-evaluate 21:32:33 I looked at what's outstanding right now (open patches). the good news is that there doesn't seem to be any big, critical thing, and we just did a release 21:32:48 so it seems that now's a really good time to start the crypto reviews 21:33:07 the timer starts, though, when the patch chain is proposed 21:33:20 does that make sense to everyone? is everyone ok with that? 21:33:38 (your affirmative answer instead of silently agreeing is helpful here) 21:33:54 makes sense 21:33:54 seems good 21:33:56 sounds good, so we’re probably done before the hackathon which is great! 21:34:01 agreed 21:34:02 seems like a good approach :) 21:34:03 sounds good to me 21:34:18 y 21:34:30 how much does a reviewer need to know about (understand) about the specifics of the encryption (crypto details)? 21:34:33 can I add a couple of things... 21:34:44 pdardeau: that's one of my things :) 21:34:59 acoles: yes, please :-) 21:35:48 the first few patches make supporting changes to core swift code without introducing any encryption specific concepts - and there are some pretty significant changes there 21:36:12 (eg adding the multipart mime/footers stuff to replicated policies instead of just EC policies) 21:36:19 adding the encryption middleware is pretty well isolated to the final couple of patches 21:36:43 there will be about 7 or 8 total patches, right? 21:36:43 notmyname: yes, that's one, and adding metadata support, and changing how we override container update headers 21:36:54 notmyname: yes. 7 I think. 21:37:27 there's one that just adds generic tests and shuffles test helper code around 21:37:43 so plenty of scope to get involved even if you have no care for the encryption pieces 21:38:15 so no excuse to not review the patch chain :-) 21:38:30 My second point was please don't push changes over the gerrit review - usually that is fine for a small fix - but here it will make managing the chain of dependencies harder 21:39:01 but please do link to diffs that i can apply if you have suggestions for better code 21:39:11 +1 21:39:59 pdardeau: sound good? 21:40:20 finally, and touching on pdardeau question again, there is an overview doc patch that could be a good place to start understanding the encryption side. 21:40:23 yep! 21:40:51 any other questions about the crypto review process? 21:41:23 sounds good to me as well. thanks acoles :) 21:42:31 yep. huge thanks to acoles, jrichli__, mahatic, and others for getting the crypto branch ready. and acoles will have a lot of work in front of him managing the patch chain 21:42:56 it's a great feature 21:43:21 let's move on to the last topic 21:43:26 #topic golang status update 21:43:45 the TC meeting yesterday brought the golang decision to a vote 21:44:02 the proposal was "golang is allowed as a supported language across openstack projects" 21:44:46 although the TC agreed with our technical reasoning, they rejected the proposal because of the feared impact to the community as a whole and the ability for people to do cross project work 21:44:58 :( 21:45:30 the next step is for me to work with dims on crafting a proposal for the TC to allow swift to use golang as a special exception 21:45:34 so here's what didn't happen 21:45:43 the TC did not decide to kick swift out of openstack 21:45:59 the TC did not mandate that swift needed to be broken into different repositories 21:46:08 those are good 21:46:35 I'm disappointed in the result, but some of the proposed outcomes would have been much worse 21:46:40 Maybe we should flirt with some other foundations and make them jealous. 21:47:35 that being said, while I keep working on the politics-of-openstack-governance side of things, as a project, we're still going forward with what we talked about in austin 21:47:44 we still have the feature/hummingbird branch 21:47:55 the technical reasoning for using golang still stands 21:48:35 the agreement amongst ourselves to go forward with using golang for the object server still stands 21:49:03 the next steps are to further define what the specific limits of the golang code we want to bring into master are 21:49:29 (which I had hoped to make more progress on by this point, but someone I got distracted by some rather epic mailing list threads) 21:49:34 *somehow 21:50:26 so the conclusion from yesterday's TC meeting is that I need to do some different yet-to-be-defined things with the TC, but nothing else has changed 21:50:51 if you're frustrated by the whole situation, I completely understand. I'm right there with you 21:51:06 please feel free to pm me, email me, call me, whatever and vent :-) 21:51:22 please don't ragequit :-) 21:51:31 * notmyname looks at dfg_ 21:51:44 huh? 21:51:52 dfg_: don't ragequit ;-) 21:51:53 he enjoys rage too much to quit 21:52:20 ya- 4:30 ragequite. 4:45 start drinking , other stuff 21:52:24 so...what questions do you have? 21:53:31 what are the chances that the new proposal will not receive the same treatment? 21:54:14 nadeem: well, I have to believe that it won't, otherwise there's just lots of despair and unhealthy habits ;-) 21:55:03 but seriously, based on what the TC said in yesterday's meeting and on the ML, having a per-project special case seems to be the direction that most TC memebers seemed to dislike the least 21:55:28 yeah, the TC seemed to fundamentally agree that golang is the right move for swift. just not the rest of openstack 21:55:37 is being 'the special project that can use golang' just a precursor to being forced out though? 21:55:41 okay 21:56:07 or to graduate :D 21:56:14 :) 21:56:14 the 'you guys don't play nice' feeling is already pretty strong 21:56:25 as long as I'm ptl, I'm going to do my best to make sure that swift is focused on solving end-user/deployer problems, and as long as we're doing that, I can deal with the politics. and golang is the right technical choice that benefits users 21:56:47 so that's what I'll keep pushing for 21:56:54 notmyname: I'm glad you continue to be willing to take all this crap on your shoulders 21:57:03 notmyname: IMO you are handling a difficult situation with great patience and wisdom. thanks for your efforts. 21:57:20 ^^^ what acoles said :) 21:57:22 Is it true that TC didn't know about the golang experiment? 21:57:30 earlier I mean 21:57:46 nadeem: I find that hard to believe 21:58:11 they were all at my "OMG objects" talk i think :p 21:58:26 because it seems like they were surprised by the proposal 21:58:39 but swift is an openstack project, and it's good for us to be part of openstack. we have a larger community of contributors, more exposure to users, and frankly a great marketing platform by being an openstack project. although there are frustrations that come with it, I don't want to seriously entertain the idea of not being part of openstack 21:59:04 nadeem: i had the same impression, and that's why i asked nearly the exact same question here a little earlier 21:59:10 nadeem: dfg_: in addition to a ML post I send out over a year ago, you've got talks at 3 summits 21:59:25 true 21:59:57 please feel free to reach out to me if you have questions or concerns about anything going on 21:59:58 pdardeau: nadeem It's hard to be sure from irc minutes but I only really detected that sense of surprise on the part of one TC member 22:00:15 looks like were at our end meeting time 22:00:27 feel free to continue discussing in -swift, of course 22:00:39 not sure if i should ask :), but patch 238799? 22:00:39 onovy: https://review.openstack.org/#/c/238799/ - swift - Change schedule priority of daemon/server in config 22:00:46 thanks for coming, everyone. it's a privilege to work with you on swift 22:01:01 #endmeeting