19:01:25 #startmeeting swift 19:01:26 Meeting started Wed Aug 7 19:01:25 2013 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:01:27 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:01:29 The meeting name has been set to 'swift' 19:01:32 who's here for the swift team meeting? 19:01:38 o/ 19:01:39 o7 19:01:45 o/ 19:01:57 o/ 19:02:02 o/ 19:02:09 * torgomatic is just here for the snacks 19:02:27 thanks everyone for joining. let's get started 19:02:28 * clayg was lead to believe there would be punch and pie? 19:02:36 topics: https://wiki.openstack.org/wiki/Meetings/Swift 19:02:44 #topic 1.9.1 release 19:02:56 so first up, notice that 1.9.1 RC was cut today 19:03:07 * portante yum, never thought irc meetings could be so tempting ... 19:03:08 this is primarily in response to the bug reported and patched by portante 19:03:27 thanks to zaitcev for asking the original question in a review 19:03:38 patches have been backported to folsom and grizzly 19:03:50 and merged into master 19:04:14 there go my plans to take down the new Intel Swift cluster with delete commands 19:04:21 redbo also has a patch out based on this to remove unlinkold() calls 19:04:23 :-) 19:04:45 assuming nothing major is reported by people this week, the current RC will be the final 1.9.1 release 19:04:59 ttx will help formalize that on monday or tuesday, depending on your timezone 19:05:20 any questions about 1.9.1 or today's release? 19:05:54 ok, moving on to the schedule for the next one 19:06:00 #topic havana release schedule 19:06:15 the openstack-wide schedule for havana is at https://wiki.openstack.org/wiki/Havana_Release_Schedule 19:06:46 from that, you can see that oct 17 is the date for havana, and the RC for havana is sept 26 (ish) 19:07:18 which means that it's likely for us (ie swift) to have just one more release before then, and that release will be included in the openstack havana release 19:08:06 Is this a good time to ask if LFS is makingit 19:08:11 * portante hopes to get diskfile refactoring and database db work from zaitcev in for that release 19:08:29 clayg has reviewed the latest diskfile proposal 19:08:29 so therefore the 1.9.2 milestone has been created and we'll shoot for the RC at sept 26 ish 19:08:42 great 19:08:43 portante: what's next?! 19:08:43 zaitcev: portante: ya, I'd love to see it included 19:08:54 clayg just asked the right questions 19:08:57 would that be 2.0 if we get lfs? 19:08:58 zaitcev: what do you need next 19:09:22 * clayg guesses he needs to review some of zaitcev's patches too... 19:09:26 chmouel: probably not (my default answer is no) 19:09:33 :) 19:09:53 * chmouel guesses that question come up every release 19:09:58 it's likely that the next release will be 1.10.0 rather than 1.9.2 (either way, just a number) 19:10:15 current_version = uuid.uuid4().hex 19:10:18 1.A.0 19:10:26 redbo: :-) 19:10:33 yes! 19:10:41 Peter asked me to take https://raw.github.com/zaitcev/swift-lfs/master/doc/source/lfs_plugin.rst and stuff it into... docstrings. Mostly. Some of it does not match docstrings entirely, so I'm going to keep that file, just make it very small. Include examples of configuration that GlusterFS/Ceph/HPSS can copy-paste. 19:11:23 The "LFS planned changes" section is supposed to list all work items and reviews, which it mostly does. 19:12:11 David was not happy about my renaming of broker.db_file, then I thought I'd help his ACL thing in exchange, then he suddenly disappeared. 19:12:12 @zaitcev: HPSS? Is someone working on this, are there any docs I didn't read yet? 19:12:52 cschwede: A guy from IBM Federal poked me about doing it. Apparently its xattr is limited and Swift does not work right on top of it. 19:13:11 zaitcev, will there be an example config for zfs? 19:13:13 cschwede: So it would be Swift's essentially unchanged back-end but with some workarounds 19:13:20 clayg: I believe that for DiskFile, we need to grind out the changes to use the new open method and then from that work do the official API definition 19:13:44 with that, we'll apply a similar discipline to the database broker based on pete's work already 19:14:31 portante: ok, feel free to ping me anytime you have a review up that I haven't looked at yet, or if you get a couple of changes going and you could use some help with one 19:14:31 zaitcev: portante: are you concerned with the ability to get changes proposed in time for inclusion in havana? 19:14:31 @zaitcev: thanks for the info. Nice to know as I know some other people interested in this 19:15:15 I think we can get DiskFile in on time, but we'll have work hard on the database db side 19:15:34 ok 19:15:42 let me know how I can help 19:16:05 anything else to discuss for the havana release? 19:16:15 and I think we need to realize that this won't be a completely polished interface, as we make changes, the ripple effects take time as folks start to see how these changes can be leveraged for other work 19:16:50 in general, the review queue has been getting really long. we all need to get better at staying on top of reviews 19:17:08 #topic swift hackathon 19:17:12 * portante has been off on perf testing for the last hour days, sorry 19:17:25 so we're going to do this :-) 19:17:32 yes 19:17:38 nice :) 19:17:39 hackathon is a good idea 19:17:43 the summit is in hong kong (and has a tendency to have lots of powerpoint 19:18:06 so for those who want to do code stuff with no powerpoint allowed, let's do a hackathon! :-) 19:18:17 sounds good! 19:18:43 I'll make the invite public shortly (ie this afternoon or tomorrow), but here's the summary 19:18:50 Tuesday, October 15, 2013 at 9:00 AM - Thursday, October 17, 2013 at 7:00 PM in Austin Texas 19:18:57 venue? 19:19:07 we'll reserve a block of hotel rooms 19:19:14 the venue will be on the UT campus 19:19:21 I replied, but feel free to bump me if you need "more core core" peope accomodated. 19:19:23 jsut north of downtown austin 19:19:23 ahh nice 19:19:28 notmyname: where at on campus? 19:20:27 creiht: in the UT commons center 19:20:39 sorry, had to find it in my email 19:20:44 hehe cool 19:21:02 http://www.utexas.edu/commons/ 19:21:05 that commons? 19:21:10 ya 19:21:23 then that is at the JJ. Pickle Research center 19:21:25 http://www.utexas.edu/commons/rooms/commons/bevoroom.php 19:21:26 not the UT campus 19:21:46 but still a cool place 19:21:59 ah ok :-) 19:22:24 isn't that place in way north austin? 19:22:33 dfg: http://www.utexas.edu/commons/maps/ 19:22:54 Nah, just North Austing 19:23:04 I used to work across the street when IBM was there 19:23:14 i still think of austin in terms of having to take the bus there... 19:23:19 haha 19:23:33 parking there will also be better than the normal campus :) 19:23:41 "Drive via US-84 E ยท 698 mi 11 h 12 min" .. hmm. I think I'll have to fly in. 19:23:46 heh 19:23:57 I'll get the rest of the logistics info out ASAP 19:24:04 lol 19:24:06 I'm looking forward to it 19:24:17 any other questions on that topic? 19:24:31 it should be no problem to join remote? 19:24:43 cschwede: not sure. we havent' discussed that yet 19:24:49 ok 19:24:49 do you know what hotel you are going to get a block of rooms at yet? 19:25:08 portante: not yet. that's next on the list fo things to do for it 19:25:11 any preferences? 19:25:36 I was going to stay in the cheapest motel. At Havana summit FBI broke into the next room and arrested a dude at 4 a.m. 19:25:45 marriot 19:25:45 lol 19:25:50 marriott 19:25:52 red hat gets a discount there 19:26:03 I'll probably just crash with family 19:26:05 I store my points there :) 19:26:13 zaitcev: whoa 19:26:39 I'll look into the marriott 19:26:56 #topic erasure codes progress 19:27:04 torgomatic: peluse: can you give an update here? 19:27:07 so I've got general multi-ring support up for review on the EC branch. Covers proxy server only and test coded updated to support the plumbing but havne't added specitic unit tests yet. Looking for feedback then will post similar changes to the replciator. 19:27:19 #link https://trello.com/b/LlvIFIQs/swift-erasure-codes 19:27:20 lots more info on teh review page 19:27:29 I provided a little feedback; I'll look more after this meeting 19:27:53 cool, thanks 19:28:13 I know Yuan has done a lot too, Yuan are you on? Its 3am there so hard to say 19:28:27 notmyname: is this a community trello board, or only for swiftstack folks? 19:28:33 portante: public 19:28:38 torgomatic, I made some early code here, c/39865/ 19:28:54 yuan: ok, I'll look at that too 19:28:58 I would not mind using this for tracking the work for DiskFile and DatabaseBroker and the rest of LFS 19:29:00 Wrt Yuan's work - he posted a draft of the get/put path changes and the EC library framework. When he cleans it up more he'll push it up for review on the EC branch 19:29:04 * torgomatic has been busy with internal SwiftStack work lately 19:29:13 portante: ok, let's talk after to get one set up 19:29:19 k thx 19:29:39 Does Kevin know about the Swift meetings? 19:29:55 peluse: I'll make sure he does 19:30:36 cool, I know he's made progress on his EC algorithms and on the algebraic signature stuff. We at Intel are also looking to open source some low level code and will contriube that as as optional library as well. Next month or so I think 19:30:43 cool 19:31:01 anything else need to be discussed on EC? any questions? 19:31:03 clayg: did you get a chance to look at the latest multi ring changes? 19:31:27 question, we want a new eccontroller or some helper func 19:32:09 whatever makes the code cleaner 19:32:12 If you mena ECObjectController I think we settled on no new controller, just helper functions. Or I read things that way anyway... torgomatic? 19:32:25 the LFS patch work at the controller level would help make this easier to do, would it not? 19:32:31 peluse: not since sam's comments 19:32:48 zaitcev already has a tree with that work that folks might want to consider 19:33:00 portante, yeah it does 19:33:01 His comments were just today I think, patch 10 is the last update I made. Lots of fun w/the test code :) 19:33:04 zaitcev: got a link for clicky goodness? 19:33:18 peluse: I'll take another look then - thanks 19:33:40 prtante: can you send the link? 19:33:44 The "big" LFS, which plugs Controller in proxy is here https://github.com/zaitcev/swift-lfs 19:33:55 thanks 19:34:01 yeah, will check it out 19:34:13 that said I'm not sure, I need to look at Paul's thing. I didn't know his "multi-ring" was baout EC 19:34:40 So to be clear the multi-ring thing and the idea of using a separate ECObjectController are two different dsicussions 19:34:52 oh 19:35:06 The separate controller would be to not require different functions in the ObjectController for get/put/etc 19:35:17 I'd kind of like to support multi-ring stuff completely before starting on EC, but that's just me 19:35:25 multi-ring I'm trying to do as very generic to apply to Clayg's storage policies soln as well 19:35:50 torgotmatic: i agree which is why I jumped on it first :) 19:36:03 i:) 19:36:12 not sure what the i is for 19:36:20 was wondering 19:36:23 * portante thought it was a solute 19:36:45 anything else on EC for now? 19:36:50 not on my end 19:36:52 before moving to open discussion? 19:36:59 nothing here 19:37:04 #topic open discussion 19:37:18 anything else to discuss in the meeting this week? 19:37:29 anyone see my general question about debug prints in the unit test code? (how to) 19:38:03 run ./.unittests --nocapture --nologcapture 19:38:08 peluse: is the problem that nose is grabbing the output streams and you aren't seeing it? 19:38:21 portante: ah, those are nice options to know 19:38:35 problem is they're going somewhere and I don't know where :) 19:38:42 I can try what portante suggests 19:39:13 i think i do -vsx 19:39:21 does not work with tox, so use that with a virtualenv with the right dependencies if not on a release that supports the latest with openstack 19:39:28 running one test at a time helps a lot too 19:39:33 yes 19:39:37 what is -vsx? 19:39:44 portante: options for nose 19:39:46 oh, sorry 19:39:53 and how do I run one test a time, that would be great 19:40:16 nosetests path/to/the/test_file.py:TestCaseName.test_name 19:40:21 -vsx 19:40:21 :P 19:40:24 swesome thanks 19:40:43 anything else? 19:41:13 thanks all for being here and for your work on swift 19:41:17 #endmeeting