19:01:02 #startmeeting swift 19:01:03 Meeting started Wed Apr 2 19:01:02 2014 UTC and is due to finish in 60 minutes. The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:01:04 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:01:07 The meeting name has been set to 'swift' 19:01:09 howdy 19:01:11 o/ 19:01:14 welcome to the weekly fun times! 19:01:20 Hello :) 19:01:20 who's here? 19:01:59 #link https://wiki.openstack.org/wiki/Meetings/Swift 19:02:11 sorry, just catching up. and going slow. 19:02:20 here 19:02:24 just got to my desk about 3 minutes before the meeting :-) 19:02:57 ok, status updates on icehouse 19:03:08 #topic icehouse final things 19:03:34 by my count, there are 2 patches that should land today 19:03:50 and then we can safely do an RC tomorrow am (california time) 19:04:02 if-none-match https://review.openstack.org/#/c/81646/ 19:04:05 I am still hoping the in-process tests will land 19:04:10 maybe that is too much to expect 19:04:11 which creiht and acoles and myself are looking at 19:04:16 I am almost done 19:04:28 and https://review.openstack.org/#/c/84456/ about swift-{account,container}-info 19:05:03 the 2nd there is important because we haven't realeased with those tools yet, and they are currently out of sync on master 19:05:14 ie different outputs etc 19:05:30 and unifying it before the release means less migration pains 19:06:03 the if-none-match patch is a nice feature and close 19:06:24 portante: if the func tests can get finished and reviewed today, that's great to include. but is that likely? 19:06:53 dunno. :) 19:06:55 just hopeful 19:06:58 ok 19:06:59 perhaps unfounded 19:07:03 I'll look at if-none-match 19:07:07 zaitcev: thanks 19:07:21 notmyname: i'm also looking :) 19:07:26 right now I'm trying to figure out why the if-none-match stuff isn't getting test coverage 19:08:06 anything else on icehouse patches? 19:08:16 creiht: me too, but i start thinking that the problem is coverage report, not the coverage itself 19:08:19 fyi, here. (WFH and having some VPN issues though) 19:08:24 ohai! 19:09:12 ok, if nothing else for icehouse rc, let's move on 19:09:26 #topic tempest fyi 19:09:40 this is really just to make sure people know what's going on 19:09:47 a couple of things here 19:10:06 first, I've put some swift-related patches for tempest on https://wiki.openstack.org/wiki/Swift/PriorityReviews 19:10:12 so taking a look there would be good 19:10:33 second, and the bigger thing, is that tempest is going to get a lot more important to us in the future 19:11:00 the current discussions in the openstack board of directors is that the tempest tests will actually be what determine what can be called "openstack" 19:11:23 ie if it's tested by tempest and passes in most installs, then that can be openstack 19:11:35 if it's not tested in tempest, then it can't be called openstack 19:11:39 heh 19:11:45 lol @ creiht 19:11:47 * creiht goes to write minimal-stack 19:11:57 (there are a few other things, but the tempest part is most relevant to us now) 19:12:27 I don't really have much to add there. this is an FYI about the openstack steamroller process that is underway 19:12:42 any questions? 19:12:52 notmyname: but tempest will only work with keystone? 19:12:59 s/will/is/ 19:13:11 cschwede_: that may be on purpose 19:13:25 keystone is not that bad 19:13:25 not sure what the keystone requirements will be 19:13:26 ya. that's the way it is today 19:13:33 cschwede_: I think openstack wants to test how swift will be used in a full openstack setup 19:13:52 zaitcev: but a *lot* of swift providers do not use keystone 19:13:53 portante: "openstack" thinks there is only a "full openstack setup" 19:14:04 sure 19:14:07 notmyname: well that's the setup they're most concerened with 19:14:16 notmyname: except for those clouds that don't run swift that will still be called openstack 19:14:18 and they leave individual testing to the respective projects 19:14:23 right. I'm not (currently) complaining about that. just acknowledging it 19:14:27 hehe 19:14:34 can we propose its not openstack w/o swift? 19:14:36 yeah we don't need to beat the dead horse here 19:14:36 :) 19:14:50 that said afaik tempest tests for swift are in a (very) good shape 19:14:52 portante: actually, ya. but that's dependent on tempest testing 19:14:58 err peluse^ 19:15:05 cschwede_: that's good :-) 19:15:13 good coverage for most functionalities 19:15:20 they didnt' used to be, but I know there has been some efforr to improve things 19:15:35 although they aren't using /info, which I discovered last night 19:15:50 who maintains the tempest tests for swift? 19:15:51 notmyname: but the plan is to use /info, might even be a patch for it 19:15:59 cschwede_: great! 19:16:02 are not unit tests per-project enough? 19:16:03 ;) 19:16:25 peluse: the tempest team. there's a PTL and core and everything 19:16:53 so do they keep in pretty clsoe contact with the various projects to assure decent coverage? 19:17:23 peluse: I think everyone wants the current state of things to get better 19:17:31 like... for example... do we need to do anything there wrt policies? 19:17:32 (dodge, dodge, political answer) 19:17:50 peluse: yes, probably. I don't know what yet 19:17:58 its almost like you're running for some office or something :) 19:18:51 any other questions about tempest? I don't think there's actually much to do in here today about it, other than "don't ignore it, please" 19:19:16 I'll try. If-none-match first though 19:19:26 zaitcev: thanks :-) 19:19:41 #topic storage policy update 19:19:50 clayg: peluse: can you give us a quick status update? 19:20:15 I've got a small refactor and a bug fix ready for eyes. Have a 3rd that depends on clayg's update on reconclier. 19:20:26 aside from some polish the primary code paths are all complete 19:20:44 next steps? 19:21:31 there is work remaining for the reconciler, and particularlly with container replication - working through probetests and manual functional testing 19:21:49 we'll need docs and some rebase'ing to prepare the full set of changes for merge on master 19:22:09 OK, can we get some more folks to get the other lower hanging fruit ready-to-reivew patches pushed through so all that's remaining is clayg's reconciler? 19:22:15 ...and docs 19:22:21 clayg: you've taken over torgomatic's branch, right? 19:22:44 reminds me, notmyname, do you ahve someone good/at planninng on doing a doc section for SP or want me to take a stab? 19:22:44 ie, you'll be the one proposing those to master? 19:23:01 notmyname: yeah i'm still poking holes in the cleanup_table approach and trying to hybridize his approach with the queue based reconciler 19:23:16 peluse: we're it. I'll be happy to help write those with you 19:23:37 notmyname: cool, let me know if you want me to start it if you want want to 19:23:37 clayg: great! 19:23:43 peluse: go ahead :-) 19:24:03 peluse: I'd like to go over the unified diff and look for drive by refactorings and particularlly cleanup in tests that could be brought over to master before the storage policy patches 19:24:37 so next steps: (1) someone please rview the 2 patches I have that are small/ready (2) clayg to continue wrapping reconciler (3) clayg to udpate github patch set once those are done (4) me/notnmyane on docs (5) core folks review github patch set (6) clayg start to porse to master. Yes? 19:24:44 peluse: I think we've been doing that quite a bit as we go but there may be some stuff that would look find on master as it and simplify the final change line 19:25:11 peluse: somewhere in there is an icehouse release 19:25:29 well, yeah, I didn't mean to imply an urgnecy ... 19:25:32 but we are on the sp topic :) 19:25:35 clayg: tomorrow am the branch will be cut, then we're free to do whatever on master 19:25:42 whoo 19:26:07 clayg: so you'll start proposing sp to master at lunch tomorrow? ;-) 19:26:16 it's done, right? 19:26:36 2 patches I mentioned above: https://review.openstack.org/#/c/83252/ https://review.openstack.org/#/c/83840/ 19:26:39 heh, maybe 19:26:56 peluse: thanks 19:26:57 maybe eat lunch first :) 19:27:07 any other questions or updates on storage policies? 19:27:35 just a huge thanks to clayg for the exyta effort w/torgomatic's abse3nce 19:27:39 it's probably best to get the first changes up and get some fresh eyes on them to see where we might want to add comments/logging rephrase this way or that, why this approach or that 19:28:09 yup 19:28:23 you mean up on github or up in gerrit? 19:28:31 gerrit IMO 19:28:36 ... unittest coverage, do we recommend multiple policies on a saio, should functional tests look for multiple policies. 19:28:41 yup, agreed 19:28:53 eah i ment gerrit 19:29:18 #topic open discussion 19:29:26 anything else to bring up today? 19:29:34 creiht: everything ok in RAX-landia? 19:29:43 oh, the hackathon! 19:29:56 notmyname: as far as I know things are fine 19:30:03 cool 19:30:04 I'm against recommending multiply SP on SAIO. The install is intimidating enough as it is, and we love new adopters. Functests are different, they are free to say "install SAIO and post-configure it to add SPs like so" 19:30:13 hackathon link is public at https://www.eventbrite.com/e/openstack-swift-june-hackathon-tickets-8309569145 19:30:13 IMHO 19:30:32 zaitcev: that's reasonable, although it's really just changing the example swift.conf 19:30:51 zaitcev: agree, not sure I'd see value in a deault multiple policy install on SAIO but back to clayg's comment - I do think functinoal test coverage needs to at least run on a non-default policy 19:30:56 clayg: what does it do in absense of multiply rings, then? 19:31:26 zaitcev: heh fair point, but you can cp object.ring.gz to object-1.ring.gz 19:31:38 yup, that's mostly how I test 19:31:43 zaitcev: but I think keeping the saio simple is reasonable 19:31:51 and if you change swift.cnf and don't make another ring file you'll get puked on 19:32:01 same as if you forget to make the regualr obj ring 19:32:05 zaitcev: the only reason I find it particularlly interesting is probetests which are targeted at multiple policies 19:32:25 hmm 19:32:33 double hmm 19:32:46 yeah, we do run probetests in the end of SAIO document 19:32:58 but that question doesn't need to be solved right here right now, does it? 19:33:15 don't think so 19:34:17 anything else to discuss in the meeting today? 19:34:17 notmyname: do you know which *day* swift has at the design summit? (planning travel) 19:34:46 donagh: there is a tentative schedule that has us split across thursday and friday. it's not a final schedule, though 19:35:08 ok, that;s good enough for now 19:35:13 k 19:35:52 thanks for all your work. you're what makes swift awesome! 19:35:57 thanks for coming 19:36:00 #endmeeting