21:00:48 #startmeeting swift 21:00:49 Meeting started Wed May 18 21:00:48 2016 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:50 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:53 The meeting name has been set to 'swift' 21:00:55 who's here for the swift meeting? 21:00:57 o/ 21:00:59 * onovy 21:01:00 o/ 21:01:03 o/ 21:01:04 o/ 21:01:06 \o/\ 21:01:07 o/ 21:01:15 o/ 21:02:18 agenda this week is 21:02:19 #link https://wiki.openstack.org/wiki/Meetings/Swift 21:02:26 hi 21:02:37 evening 21:03:00 o/ 21:03:04 o/ 21:03:52 . 21:04:01 ok, we can get started I think 21:04:16 not a ton on the agenda this week 21:04:23 #topic crypto update 21:04:30 #link https://trello.com/b/63l5zQhq/swift-encryption 21:04:41 acoles: jrichli: what's going on this week with crypto? 21:05:17 well, the todo list has been going down and there's a bunch of patches waiting review 21:05:25 the top 3 are on priotity reviews page 21:05:46 including one that needs no crypto knowledge to review, patch 316924 21:05:46 acoles: https://review.openstack.org/#/c/316924/ - swift (feature/crypto) - crypto - make some probe tests compatible with crypto 21:06:00 \o/ 21:06:14 great 21:06:42 we now have copy middleware and therefore func tests pass and we have crypto in pipeline for jenkins functional test jobs on feature/crypto :) 21:06:56 woohoo! 21:07:27 cool 21:07:27 and if that patch 316924 lands then we have a good story on probe tests 21:07:28 acoles: https://review.openstack.org/#/c/316924/ - swift (feature/crypto) - crypto - make some probe tests compatible with crypto 21:08:32 jrichli has been collatingsome questions for "crypto experts" to validate our final choices over how we handle key and iv, ready for final review 21:08:57 good. jrichli when/how is that happening? 21:09:20 the strategy involves both etherpad and the doc in review 21:09:32 and some email and pings on IRC 21:10:03 i wasn't planning on a meeting yet - got to get everyone on same page on what we are doing now first, i guess 21:10:05 we 21:10:18 we'll decide from then if a meeting needs to happen. prob will 21:10:31 ok 21:11:04 notmyname: so although we are closer to being ready for final review, in that tasks are being addressed at a good rate, we have a review backlog to get through on feature/crypto before I can start to prepare for a feature/crypto-review branch 21:11:11 looking at the TODO list on the trello page, the only one that looks tricky is the PUT is really post-as-copy 21:11:34 acoles: ok. you're reading my mind. I was going to ask that :-) 21:11:35 notmyname: ah, i wanted to raise that for opinions here 21:12:22 currently a post-as-copy will cause a previously unencrypted object body to get encrypted (as it is copied), which differs from a fast-post 21:12:56 right. makes sense 21:13:29 we think we know how to prevent that (more code!) but question is do we care? on one hand I feel it is inconsistent wrt fast post, on other hand clients cannot tell, its an internal anomaly 21:14:19 and I could even buy an argument that if something is unencrypted and we get chance to encrypt it then we should! 21:14:20 what's the end result? a POST may or may not be (re-)encrypted? 21:14:34 but otherwise everything works? 21:15:16 an already encrypted object will always remain encrypted after a POST, an unencrypted object will become encrypted by post-as-copy 21:15:32 but not fast-post 21:15:32 doesn't matter to me... you copy A to B, GET B, and A's contents come back 21:15:42 what's not to like? 21:15:48 and unencrypted object with fast-post will have encrypted metadata and listing info? 21:15:50 otherwise it all works, and client cannot tell 21:16:10 notmyname: yes 21:16:13 notmyname: just encrypted metadata, thelisting etag does not change on a fast-post 21:16:19 ok 21:16:22 my gut reaction is that I'm completely fine with crypto only ever working with fast-post. so yeah, sounds like a non-issue to me. it Just Works 21:16:24 ok, partly right ;-) 21:17:15 I am obviously biased towards not writing more code. 21:17:27 lol 21:17:28 I mean that I would be fine if you told me that crypto is broken with post-as-copy. so I'm definitely ok with it working with both fast-post and post-as-copy, evne if they have slightly different internal effects 21:17:49 yes! definitely write less code :-) 21:18:02 ok. any contrary opinions? 21:18:04 since we are going in that direction anyway, i will also bring up the fact that it has been expressed that those that really want to ensure all data is encrypted will no be starting with pre-existing unencrypted objs 21:18:53 so, just more reason not to write more code! 21:19:11 acoles: I don't hear any 21:19:24 If you've turned on encrypted you want things encrytpted.. if you have objects taht aren't if they start being encrypted even partially then that's what you want.. so I think its ok 21:19:25 notmyname: that's helpful, that was probably one of the bigger/trickier remaining todo's. 21:19:47 #agreed crypto doesn't need to detect if a PUT is from a POST-as-COPY 21:19:48 and now its the easiest! 21:20:01 notmyname: coding todo's that is, we have lots of reviewing todo's ;) 21:20:10 :-) 21:20:32 ok, so the top 3 reviewing todos are the ones on the priority reviews page? 21:20:35 i know! i keep adding to my list to review throughout the day - the stack is not shrinking 21:20:38 #link https://wiki.openstack.org/wiki/Swift/PriorityReviews 21:21:46 notmyname: yes, and i know I have mentioned it before but I'd really appreciate feedback on patch 316924 (probe test changes) because if the approach I have there is not acceptable then I need to think really hard 21:21:46 acoles: https://review.openstack.org/#/c/316924/ - swift (feature/crypto) - crypto - make some probe tests compatible with crypto 21:22:20 acoles: added to the top of my stack 21:22:28 and thinking is even harder than coding! 21:22:36 :) 21:22:55 acoles: the approach seems sane to me. i should actually run the probe tests for that one... 21:23:02 acoles: jrichli: thanks for the update 21:23:08 now speaking of priority reviews... 21:23:16 #topic review the reviews 21:23:26 looking at https://wiki.openstack.org/wiki/Swift/PriorityReviews 21:23:37 last week we had some busted probetests 21:23:45 I think those are all ok now, right? 21:24:19 those problems were fixed, but i have had more lately - but it might just be me 21:24:20 I reviewed something to fix a probe test today, testing expirer 21:24:36 ah, yeah, that's my fault. 21:24:52 patch 318167 21:24:52 acoles: https://review.openstack.org/#/c/318167/ - swift (feature/crypto) - crypto - test varying proxy pipeline configs 21:25:03 no not that! sorry 21:25:32 patch 315918 21:25:32 kota_: https://review.openstack.org/#/c/315918/ - swift - Fix probe failure and small things 21:25:40 yes, that ^^ 21:25:47 got it 21:25:59 I'll make sure that's on the priority reviews page, too 21:26:13 i saw that one, and i ran with those changes to see if my issues were fixed. seems like i still had issues, but i will test that again and leave comments. 21:26:21 makes me think we should reexamine getting the probetests running in the gate 21:26:25 jrichli: thanks 21:26:44 jrichli: i think it could be racey as it is now 21:26:58 patch 238799 has some comments and has a merge conflict 21:26:58 notmyname: https://review.openstack.org/#/c/238799/ - swift - Change schedule priority of daemon/server in config 21:27:07 onovy: is it something peter will be able to look at again? 21:27:12 peterlisak: ^ 21:28:34 ok, I'm going to assume "no" then 21:29:17 we might need to bug someone to take that over. mattoliverau were your review comments something that would be reasonable to write and push over the current version? 21:30:06 yeah 21:30:24 mattoliverau: ok. might be good to do that in order to move it forward 21:30:58 I'll also add patch 317475 to the priority page. looks like a rather ugly bug 21:30:58 notmyname: https://review.openstack.org/#/c/317475/ - swift - Send correct size in POST async update for EC object 21:31:03 acoles: thanks for picking that one up 21:31:18 notmyname: let me take a look at it again. If no one else will I'll rebase and push up a new patchset and we can take it from there. 21:31:19 notmyname: i caused it :/ 21:31:42 acoles: but importantly you're solving it :-) 21:32:14 container sync and crypto stuff on the priority reviews page i'll leave as-is 21:33:06 somewhat related to the priority reviews page (and as a small status update from me)... 21:33:21 if you haven't seen it already, I added a new section to http://not.mn/swift/swift_community_dashboard.html 21:34:00 the "needs follow-up" is a list of patches that has had reviews but doesn't have any reviews on the current patch set 21:34:42 ie stuff where the patch owner has been doing stuff but is likely waiting on reviews 21:35:12 and although not published in a web page or irc bot yet, I've also built on that to build some personalized lists 21:35:25 (I talked about this some in bristol) 21:35:59 nice 21:36:14 so eg stuff that you own that has a negative review or stuff that youv'e starred but haven't reviewed 21:36:41 and I'll be working on some more complicated things like stuff you've previously negatively reviewed but has a new patch set 21:36:54 all sorted by the community stars. etc etc 21:36:58 still WIP :-) 21:37:24 cool 21:37:31 #topic next release 21:37:42 I'd like to have a release *before* we land crypto 21:38:00 or rather, not wait until crypto lands to do a release 21:38:24 AFAICT patch 317475 is the biggest thing not landed that should 21:38:25 notmyname: https://review.openstack.org/#/c/317475/ - swift - Send correct size in POST async update for EC object 21:38:37 but if there's other stuff, please let me know 21:39:31 does that sound reasonable? any reason to rush one out now or wait until crypto lands? 21:40:38 nobody is very chatty today ;-) 21:41:03 sounds reasonable. Sorry -ETOOEARLY :P 21:41:17 was just looking at open patches 21:41:19 +1 21:41:19 hehe, nothing special to push for it :) 21:41:29 :-) 21:41:32 ok, a couple of miscellaneous things before talking about the state of golang 21:41:38 #topic misc stuff 21:41:58 yesterday was swift's "birthday". six years in production 21:42:08 congrats! 21:42:21 \o/ 21:43:06 congrats!! 21:43:10 it's pretty cool, I think. and the fact it's still a thing that people are using and taking into account is a huge testament to the people who are writing and reviewing the code. thanks for your work 21:43:53 has anyone tried putting swift on 16.04 yet? I've been putting together a new SAIO on 16.04 21:44:19 for me, it's working pretty well so far. just a couple of weird new things that I need to transcribe into the SAIO docs, but nothing major 21:44:42 I haven't yet. 21:45:10 well I was still on precise so I figured after 4 years it was a good time to upgrade :-) 21:45:32 quick note about the hackathon 21:45:41 I tried an openstack-ansible AIO on 16:04 and had some issues, but they I think is me and nothing to do with Swift. :) 21:46:12 basic info is july 11-14 in san antonio. rackspace is sponsoring and hurricanerix is doing all the logistics coordination 21:46:12 s/they/that/ 21:46:35 thanks hurricanerix! 21:46:43 like normal, we'll have an eventbrite page set up for registrations and likely some sort of cap on how many 21:46:43 mattoliverau np 21:46:48 \o/ 21:46:50 I'm working with hurricanerix to get all those details 21:47:05 hmm, that sounds like I'll be flying home on my birthday.. that's exciting :P 21:47:07 thanks hurricanerix :-) 21:47:18 mattoliverau: maybe you'll have 2 if you time it right? 21:47:19 good luck hurricanerix! 21:47:33 notmyname: wrong way. there is a chance I'd miss it :P 21:47:38 lol 21:47:43 thanks, jrichli gave me her checklist, which is making this much easier. 21:47:55 mattoliverau: do that every year and stay forever young 21:47:58 yw 21:48:09 I'll pass out more info on the hackathon as we have it 21:48:20 lol, yeah if your birthday day doesn't actually happen, do I get to stay younger? 21:48:27 of course! 21:48:29 #topic what's up with the whole golang conversation? 21:48:51 and if anybody has any questions or requests, feel free to pm/email me 21:48:52 so there's been a *huge* mailing list thread and several rather intense IRC meeting discussions 21:49:36 nothing is "decided" per-se, but many of the cross-project teams are working on figuring the details of what supporting golang means 21:50:15 eg we've got to figure out how we identify dependencies and have some sort of reproducable builds. and how to run golang tests in the -infra environment. and how to find and build docstring-style docs 21:50:19 stuff like that 21:50:23 so those are going well, I think 21:50:40 the TC meeting this week, however, was...interesting 21:51:28 basically it very quickly got steered from "do we support golang" to "what *is* openstack and is it ok to actually have a dataplane service project in openstack or not" 21:52:01 wow 21:52:02 there will likely be another (very long) email thread about this question in the near future 21:52:08 mattoliverau: yeah. wow 21:52:37 I still say we were here first, and they should be the ones to leave. 21:52:41 lol 21:52:44 ha ha 21:52:44 hahaaha 21:52:48 redbo: rofl 21:52:52 I actually don't htink anyone is going to leave :-) 21:53:04 so, because there's a whole lot of different things being said, I wanted to be available to answer any questions you have about it, if you've been paying attention 21:53:30 but reading through tc meeting logs...it seems like some guys want Swift to leave openstack 21:54:20 i think it's more of they want the front-end of swift to stay, but the back-end (which we'd be porting to go) should be its own separate service 21:54:21 nadeem: I'm not sure that's the case. I do think there's some misconceptions and a perception that "swift is different" 21:54:49 bkeller`: and that's the misconception, I think. where the (rather arbitrary) line between front-end and back-end is 21:54:52 yep true 21:55:16 so that's part of what I'm trying to work with a few people on the TC and in defcore about 21:56:08 as of now, with the current defcore standard, you can run swift on top of some other system and still pass defcore. however, that doesn't mean swift is an object storage provisioning service 21:56:11 notmyname: I would say you did a good job on all of this. 21:56:21 nadeem: thanks. still trying to figure out most of it as I go :-) 21:57:04 if anyone does have concerns, please feel free to ask me publicly or privately. concerns with either what other people are saying or with what i'm saying 21:57:10 Does that mean all the other projects would have to spin off their default backends? Like keystone's mysql backend and cinder's lvm backend? 21:57:50 redbo: actually i think the extreme conclusion is that swift would become like mysql or lvm (a service that is used but not part of openstakc) 21:58:09 now, I don't htink that will actually happen, but it is the logical conclusion of one answer 21:58:32 redbo: at least, that's how ttx presented it in the TC meeting 21:59:19 personally, I'm not sure why this conversation needs to be had or needs to be had at this particular time. I'm not sure of the actual problem that's being solved or the issue that's actually broken 21:59:52 there are about zero people who have actually approached me to talk about these things, although i have had several people relay these things to me anonymously 22:00:23 so in general, stuff is happening. we;ll see where it goes. I'm not particularly worried about swift's place in openstack in the future 22:00:34 and I'd be happy to talk with you if you want to go over it more 22:00:58 sorry I'm dumbfounded with the craziness. 22:01:10 heh. me too, sometimes :-) 22:01:17 looks like we're over time 22:01:25 notmyname: thanks as always for being the voice and taking the swift fight 22:01:38 +2 22:01:40 notmyname: maybe we can have group therapy at hackathon ;) 22:01:41 notmyname: I'll second that - thanks 22:01:43 thank you for your participation and work on swift. 6 years now! 22:01:48 pdardeau: lol 22:01:48 Yeah 22:01:54 I've been working on it for 7 years :( 22:02:00 I appreciate your support :-) 22:02:05 redbo: 6.5 for me :-) 22:02:13 #endmeeting