21:00:05 #startmeeting swift 21:00:06 Meeting started Wed Aug 17 21:00:05 2016 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:08 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:11 The meeting name has been set to 'swift' 21:00:15 who's here for the swift team meeting? 21:00:20 o/ 21:00:22 sup 21:00:23 hello 21:00:23 o/ 21:00:26 o/ 21:00:33 here 21:00:49 o/ 21:00:52 welcome 21:01:11 agenda this week is https://wiki.openstack.org/wiki/Meetings/Swift 21:01:17 #link https://wiki.openstack.org/wiki/Meetings/Swift 21:01:45 \o 21:01:59 I want to go over "this week in the tc..." and then hear from people as to what you're workign on and how's it going 21:02:24 o/ 21:02:25 timburke: torgomatic: courtesy ping 21:02:31 ohai 21:02:36 hi 21:02:40 hi 21:02:59 #topic tc goals 21:03:17 you may have seen this email thread on the mailing list 21:03:30 here's the summary of the tc decision this week: 21:03:52 the TC will set openstack-wide goals for every project to implement in a given development cycle 21:04:34 the 2 proposed goals for the octa release are (1) stop using any oslo incubated code and (3) full py35 compatibility, including all tests and all services running under py35 21:05:01 if the second one sounds ... ambitious, you aren't alone :-) 21:05:12 :) 21:05:35 and there are likely other projects (including nova) that will likely not be able to implement that during the octa cycle 21:05:39 so the debate continues 21:05:53 but that being said, there's the current status and something we'll all be affected by 21:06:12 if you want the details, I'll point you to the ML thread and/or the gerrit review 21:06:30 notmyname: can you link them here? 21:06:40 ok. give me a moment 21:06:55 notmyname: you mean you weren't prepared :P 21:07:03 :-o 21:07:25 https://review.openstack.org/#/c/349068/ 21:07:26 notmyname: patch 349068 - governance - describe a process for managing community-wide goals (MERGED) 21:07:57 #link https://review.openstack.org/#/c/349068/ 21:07:57 mattoliverau: patch 349068 - governance - describe a process for managing community-wide goals (MERGED) 21:08:03 the TC meeting log is here http://eavesdrop.openstack.org/meetings/tc/2016/tc.2016-08-16-20.02.log.html 21:08:21 notmyname: mattoliverau: acoles: thx a lot! 21:08:52 and http://lists.openstack.org/pipermail/openstack-dev/2016-August/100537.html seems to be close enough to the top of the thread to pick it up 21:09:22 the recent branch of it is at http://lists.openstack.org/pipermail/openstack-dev/2016-August/101348.html 21:09:48 so peruse those and feel free to ask questions 21:10:31 any questions or comments on that topic before we move on? 21:11:15 ok, let's move on then 21:11:22 notmyname: what useful input can be made to the debate? 21:11:29 from swift I mean 21:12:19 when it comes to the goals themselves, we'll need to give realistic feedback about what it will take for us to implement them 21:12:45 as to the process itself, it seems that the debate is over 21:13:25 understood re process, just wondering how they will decide on the py35 for ocata goal 21:14:55 my understanding is that the goal is proposed and approved by the tc and then the feedback is given by the projects by a relatively early date in the release cycle 21:15:18 although there seemed to be some debate in the meeting on if that was actually a good process or not. so it might be tweaked 21:15:30 but that's what's on record now 21:16:01 I guess what I'm saying is that "py35 in octa" is decided. we can give feedback later on what we're doing to make that happen 21:16:11 at least, that's my understanding of the current situation 21:17:32 acoles: is that how you saw it? 21:18:06 notmyname: thanks for clarifying. I was not sure if the py35 goal had been adopted or just proposed 21:18:36 actually, it hasn't landed yet, so it's not formally adopted. so there might be some small changes, I suppose 21:18:47 https://review.openstack.org/#/c/349069/ 21:18:47 notmyname: patch 349069 - governance - add ocata goal "support python 3.5" 21:19:19 looks like a new patch set was uploaded in the past hour, so i'll need to look at that 21:19:25 anythign else on this topic? 21:20:45 ok, moving on 21:20:50 #topic status updates 21:21:13 it's been a crazy few weeks, with the golang stuff and the other tc debates 21:21:25 or at least, it's been crazy for me. I hope you've been more productive :-) 21:21:51 so that being said, I want to hear what you're working on, how's it going, and where we can all pitch in 21:22:06 Looking for final +A for ionice / nice patch 21:22:23 onovy: yeah, I think between mattoliverau and clayg, we should have that 21:22:30 still working on repconn patch..hoping to finish it this week. 21:22:40 nadeem: great!! 21:22:48 nadeem: I'm looking forward to seeing that 21:23:10 timburke: you've been working on the history mode versioning, right? what's going on there? 21:23:15 onovy: yeah, if clayg is only co-authored by doeing some nits, then I can change mine to a +A 21:23:40 mattoliverau: coauthored was added by Peter 21:23:41 notmyname: responding about as quickly as i can now that it's finally got some attention 21:23:54 Just because he used few ideas from clay 21:23:56 hasn't changed materially in months 21:24:01 I wrote multi-range GETs for SLOs; it's in gerrit. I mostly did that since I'm reusing SegmentedIterable in some other code for which I need multi-range support, not because there's some burning desire for this functionality in SLO. 21:24:09 timburke: so are we going to land it soon? 21:24:11 just uploaded a new patchset for the partpower increase - will continue with the tempurl stuff 21:24:18 cschwede: nice! 21:24:19 timburke: I owe you a reply on history mode. It's on my list. I need to talk with somebody 21:24:20 that's the hope :-) 21:24:38 oshritf and I have been working on process concurrency for container sync, patch is ready for review https://review.openstack.org/210099 21:24:45 jrichli: timburke: me to, i will do another review tomorrow morning 21:25:12 there's also the slo-etag-in-sysmeta change -- patch 347538 21:25:12 timburke: https://review.openstack.org/#/c/347538/ - swift - Store SLO Etag and swift_bytes in sysmeta 21:26:24 which fell out of some investigation for swift3 about conditional requests and SLOs. the long and short of it was, don't mess with SLO etags :( 21:26:50 (at least on swift pre-that-patch) 21:26:52 lol. large object are such a problem child! 21:27:22 joeljwright: I've still got "release swift client" on my todo list... anything to update on work needed there? 21:27:29 mattoliverau: so I think you can +A if nobody complains :-) 21:27:41 there's at least one new bug (which i'm still writing up) that would block it 21:27:42 notmyname: it would be great to see the copy method patch added 21:27:44 https://review.openstack.org/#/c/280200/ 21:27:44 joeljwright: patch 280200 - python-swiftclient - Add copy object method 21:27:55 Aha copy from Marek 21:28:09 joeljwright: welll *somebody* put a -1 one on it! ;-) 21:28:11 onovy: I'll take a look again post meeting and breakfast 21:28:15 Marek is not working for Seznam anymore. So he will not update it 21:28:18 (requests started getting picky about header types/format) 21:28:24 mattoliverau: cool thanks 21:28:25 timburke: what's the bug? 21:28:32 onovy: ah, good info. thanks 21:28:38 and there it is 21:28:54 joeljwright: can you push your typos fixes over? 21:28:55 notmyname: but Peter can work on it. But not now, too busy 21:29:05 -H 'X-Storage-Policy: EC' doesn't work. needs to be -H 'X-Storage-Policy:EC' now (no space) 21:29:16 So I am fine if someone can take it and finish it 21:29:18 notmyname: yeah, the -1 I put on could be easily fixed with a small patch 21:29:26 I can do that tomorrow 21:29:31 joeljwright: ok. thanks. 21:29:33 Cool, thanks 21:29:56 timburke: ummm...that's...weird? 21:30:24 I'm benchmarking a golang container server. And an alternative backend that uses boltdb. 21:30:25 timburke: whitespace is from the devel after all :P 21:30:50 notmyname: related (at least vaguely) to my reply on the ML: http://lists.openstack.org/pipermail/openstack-dev/2016-August/101765.html 21:31:04 I may try leveldb/rocksdb, but it's a little more complicated because there aren't transactions. 21:31:06 redbo: interesting. 21:31:49 And do we need transactions? 21:32:07 mattoliverau: speaking of containers, any progress on sharding? 21:32:21 things like merging objects kind of need to be able to read/modify/write transactionally. 21:32:30 Ah right 21:32:49 timburke: let me know if there's something I can look at when you've got a patch/bug report 21:33:01 tdasilva has been working on symlinks ( I guess he is away now ). I still need to play around with the latest patchset! 21:33:45 yeah, i've been distracted with the Flu and the wifes surgery the last week or so.. but making progress, container backend has support for the 2 databases.. still more to do there.. and about 1/2 thorugh the container consistency engine.. I have a trello board I try and keep up to date (which I shouhd update).. its linked on the ideas page 21:33:59 mattoliverau: ack 21:34:25 next steps will me sharder modification. 21:35:31 I put up a patch to get rid of .durable files so we use less inodes https://review.openstack.org/355958 21:35:34 s/me/be/ 21:35:37 acoles: yay 21:36:11 acoles: nice! adding this to my to-review-list 21:36:23 lot's going on. instead of feeling overwhelmed, it makes me happy. :-) 21:36:59 I need to update stars/priority reviews/etc. I'll do that asap 21:37:07 cschwede: notmyname 90% of the diff lines are in tests :/ but lots of those are duplicating for old and new durable scheme 21:37:30 I'll be traveling to NYC next week for OpenStack East and the ops meetup. 21:37:52 I think I'll see tdasilva there 21:38:02 anyone else going? 21:38:19 acoles: you'll be out all next week? and the following? 21:38:23 bit far for me :P 21:38:31 mattoliverau: swim ;-) 21:38:43 notmyname: lol, I see what you did there! 21:39:08 notmyname: yes I am away for next two weeks 21:39:31 kota is out this week and IIRC will be back next week 21:39:49 and joeljwright, you'll be on vacation soon right? end of this month? 21:40:06 acoles: vacation? is it suppose to be summer on the other side of the world or something. 21:40:07 no, end of sept 21:40:13 joeljwright: ah, ok 21:40:28 acoles is going to the US! 21:40:36 Btw: what is release plan? Any deadlines? :-) 21:40:52 onovy: great question 21:41:21 * onovy need to prepare Debian package :-) 21:41:30 :) 21:41:32 we need to have a release for newton by the end of september 21:41:35 for swift 21:41:49 mattoliverau: ironically it is quite summer-like in UK but we're heading elsewhere 21:42:06 I suspect that will be the next release we have (we *might* could have one before then, but I doubt it) 21:42:19 Ok 21:42:24 And client? 21:42:34 acoles: lol, you should save up your leave and come to Oz in your winter.. how theres a holiday ;) 21:42:35 and I'd like to do a client release asap. call it the next couple of weeks? 21:42:47 Nice thanks 21:42:52 http://releases.openstack.org/newton/schedule.html 21:42:59 ok, I'll get some time put aside for reviews and updates on the client 21:43:16 I propose we do not meet next week. unless someone would like to volunteer to chair 21:44:22 anything else from anyone to bring up in this week's meeting? 21:44:29 yeah, actually I have 21:44:37 while I've been hiding 21:44:40 I'm happy no meeting.. if we decide that needs to change, then we can figure something out.. I'm excited about possibly sleeping in a little :) 21:44:49 I've been working on a middleware I'd like to open source 21:44:58 joeljwright: cool 21:45:07 (based on SLO to make TAR objjects) 21:45:18 joeljwright: oh? like a reverse bulk upload? 21:45:22 but what's the best process to make this thing available? 21:45:52 I still have work to do, but it's at a point where feedback would be helpful :) 21:45:56 joeljwright: do you want it to be upstream in swift? or do you just want it to be open? 21:46:04 Create new os project with infra? 21:46:12 Or merge it to swift 21:46:21 notmyname: I suppose that depends on interest 21:46:34 joeljwright: if it's the reverse of the bulk upload, I'm interested :-) 21:46:44 * onovy too 21:46:44 joeljwright: i’m interested! had something similar in mind for some time... 21:46:46 joeljwright: start as a patch to swift, we can take it from there ;) 21:46:55 so there's 2 options 21:47:05 either just straight up propose it to swift as a patch in gerrit 21:47:26 or write up something and put it in your own code hosting thing (github or otherwise) and link it on the ideas wiki 21:47:40 +1 for first option 21:47:42 but either way, then you'll need to bug people to look at it. and we can take it from there 21:47:59 nice one joeljwright 21:48:06 the hardest part for me with option 1 is that it's targetted at 2.2.x 21:48:57 oh yeah, there's another thing i've been working on! https://review.openstack.org/#/c/342857 21:48:58 timburke: patch 342857 - swift - Add defaulter middleware 21:49:04 but making a patch might be the easiest way to get it under noses 21:49:06 thanks everyone 21:49:35 timburke: I've been meaning to look at that.. I'll try and get to it this week 21:49:48 Btw I added all manpages 21:49:59 onovy: thanks 21:50:13 Can't found patch now, I am on cellphone 21:50:36 another thing (still thinking of that requests upgrade): what do we think about developers passing headers like {'X-Object-Meta-Foo': 1000}? should we change swiftclient to str-ify all header args, or should we go fix https://github.com/openstack/swift/blob/master/test/probe/test_object_expirer.py#L72 ? probably worth considering the discussion at https://github.com/kennethreitz/requests/pull/3366 21:50:43 onovy: https://review.openstack.org/#/c/355187/ 21:50:43 notmyname: patch 355187 - swift - Added missing manpages for all cli tools 21:50:50 Yep that's it 21:51:01 Help2man is your friend 21:51:49 timburke: I'm not up to date on teh requests upgrade topic, but my gut reaction is that we should assume headers are strings and not any other particular type 21:52:44 timburke: where's the requests upgrade discussion/but/whatever happening? 21:52:49 how do I get caught up there? 21:53:36 but when they aren't, should we (a) yell loudly, (b) do nothing and let requests (maybe) yell loudly, or (c) coerce it to a string 21:53:44 And another topic we was working hls live streaming directly from swift. And guess what, it's working and we will use it :-) 21:53:49 notmyname: go see the requests pull 21:54:02 timburke: ok. has this trickled into openstack yet? 21:54:13 onovy: awesome 21:54:28 I will show demo when it will be ready 21:54:32 notmyname: i've seen a message on the ML related to it 21:54:42 timburke: I'll have a read of that requests backgroud and try to form an opinion 21:54:44 ok 21:54:50 joeljwright: thanks 21:54:59 anything else from anyone this week? 21:55:20 -> bed, thanks, gn 21:55:25 hello 21:55:32 onovy: o/ 21:55:52 kota_: little late, but you made it! 21:56:09 sorry, my alarm has been stopped. 21:56:16 kota_: no worries :-) 21:56:32 everyone, thank you for coming to the meeting this week 21:57:13 and thank you for working on swift. it's very nice, for me, (especially after all the recent discussions) to hear about the feature being written, the bugs being fixed, and the prod use cases being addressed :-) 21:57:28 no meeting next week, and I'll see you in #openstack-swift 21:57:31 #endmeeting