14:00:13 <markwash> #startmeeting glance 14:00:14 <openstack> Meeting started Thu Aug 1 14:00:13 2013 UTC and is due to finish in 60 minutes. The chair is markwash. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:15 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:17 <openstack> The meeting name has been set to 'glance' 14:00:39 <brianr> hi mark 14:00:52 <markwash> brianr: o/ 14:01:02 <flwang> o/ 14:01:16 <iccha> \o 14:01:20 <zhiyan> ops, time changed? 14:01:29 <jbresnah> o/ 14:01:40 <nikhil> o/ 14:02:42 <flwang> markwash: what's the agenda today? 14:02:52 <markwash> good question! 14:02:54 <markwash> #topic agenda 14:03:06 <markwash> I meant to talk about documentation stuff some last week but forgot 14:03:21 <flwang> yep, good topic 14:03:32 <flwang> the homework assigned to you :) 14:03:49 <markwash> and I'd like to review any blueprint progress we've made, give people a chance to highlight reviews 14:04:01 <zhiyan> good point 14:04:07 <markwash> so, from me *blueprint progress and *docs 14:04:22 <markwash> any other items folks want to talk about that aren't covered under bp progress? 14:05:56 <markwash> okay, maybe we'll think of something along the way 14:06:01 <markwash> #topic docs 14:06:10 <markwash> so I did finally contact annegentle 14:06:13 <jbresnah> perhaps a review day/bug squash day 14:06:41 <iccha> markwash: and? 14:07:01 <markwash> and I got a great response, main items were 14:07:30 <markwash> glance/doc/source should contain stuff for contributors, which is probably a good catchall for design, relevant history, and api spec plans 14:07:47 <markwash> and openstack/image-api should contain our spec documentation 14:08:31 <markwash> also openstack/api-site/ is for api reference, but I'm not really sure what that means apart from the spec. . . 14:08:51 <markwash> I haven't taken the chance to determine what kind of state those various places are in 14:08:58 <flwang> markwash: i think the api-site is update to date, just need recheck 14:08:59 <iccha> yeah the different between image-api and api-site? 14:09:27 <flwang> iccha: +1 what's the image-api mean here? 14:10:15 <markwash> one final thing, before I forget, is openstack/openstack-manuals for installation / administration info, there might be some elements of that we want to contribute to as developers 14:10:28 <markwash> iccha: I'm as confused as anyone about the distinction 14:11:39 <iccha> i feel like we need a doc day or something, or atleast split the effort. the current document is in different states from some info being incorrect, inconsistent to some latest details being there 14:11:48 <iccha> dont know if thats stretching it 14:12:41 <markwash> something to that tune would make a lot of sense 14:13:30 <markwash> it also looks like there's a divide in the image-api repo 14:13:31 <flwang> so until now, we mentioned: 1. openstack/doc/source 2. api-site 3. installation/admin doc 14:13:43 <markwash> with v1 in docbooks, and v2 in markdown 14:14:24 <brianr> api-site looks like a split from openstack-manuals, it's got the quickstart guides in it 14:14:52 <brianr> i wish they would stop moving stuff, it is very confusing 14:15:09 <zhiyan> brianr: +1 14:15:24 <markwash> I'm stuggling for a strategy, how we can have the most impact before H is released 14:15:58 <iccha> markwash: maybe we should start with the doc/source becuase thats something as devs we can start working on right away 14:16:10 <flwang> iccha: +1 14:16:13 <markwash> I buy that 14:16:17 <iccha> and then if we have the meat information, it can be ported to other places 14:16:34 <iccha> the installation guide may require some additional work tho 14:16:52 <brianr> i will get info into the "operator docs" (which you haven't mentioned!) 14:16:54 <flwang> I can recheck the api-site 14:17:17 <markwash> sorry, slightly distracted, cats attacking each other 14:17:39 <brianr> flwang: +1 , that needs to be updated for the current glanceclient 14:17:50 <markwash> brianr: which operator docs? 14:18:00 <flwang> operator guide? 14:18:05 <brianr> https://github.com/openstack/openstack-manuals 14:18:17 <brianr> it's where the common images properties went for grizzly 14:18:24 <markwash> okay gotcha 14:18:37 <brianr> it's where some config info for protected props should go 14:18:51 <markwash> it might also make sense as a first effort for people to go through and submit patches to remove the out-of-date or incorrect portions of whatever is out there 14:18:56 <iccha> so apart from the 3 points flwang listed there is a 4. operator docs? 14:19:01 <markwash> I'm not sure there is a lot of that, but there might be some 14:19:23 <markwash> oh and, FWIW we're not necessarily on the hook for all of these 14:19:34 <markwash> hence, starting with the one in our source tree does make a lot of sense 14:19:54 <flwang> yep 14:20:01 <markwash> so it sounds like we have some volunteers to look into and assess the ones on our list? 14:20:29 <brianr> yep 14:20:47 <flwang> flwang: api-site, brianr: operators doc 14:20:53 <iccha> markwash: +1 there must be some sort of split up so we dont step on notes. an etherpad or soem sort of communication so we all know whose looking into what 14:21:15 <iccha> even if its just docs in glance repo 14:21:16 <boris-42> Hi all 14:21:18 <boris-42> =) 14:21:23 <iccha> hey boris-42 14:21:27 <flwang> iccha: can you help create a etherpad link to track this? 14:21:54 <iccha> https://etherpad.openstack.org/glance_documentation_efforts 14:21:58 <flwang> then we can update our status on that and everyone know each other's focus to avoid overlap 14:22:00 <flwang> cool 14:22:02 <flwang> thanks 14:23:55 <markwash> so did we agree the first step is essentially review? 14:24:04 <markwash> or is that already done to some extent elsewhere? 14:24:59 <flwang> markwash: i agree the first item we need to do is figuring out the out-of-date stuff 14:25:06 <brianr> i think iccha and eddie already did a review 14:25:14 <brianr> it's on an etherpad somewhere 14:25:19 <markwash> ah, lord, apparently I'm just completely behind on this issue :-/ 14:25:22 <iccha> markwash: i think it could be done simelatanesouly. while some ppl identity the purpose/state of soem resources. we need to identify a central source of truth where we start contributing 14:25:53 <brianr> we should probably file documentation bugs 14:26:03 <iccha> https://etherpad.openstack.org/glance_etherpad_list - etherpad of all glance etherpads . the stuff esheffield and me worked on is like api related 14:26:04 <brianr> that way, we won't step on each otehr 14:26:30 <markwash> brianr, iccha: +1 14:27:33 <markwash> brianr: are those like bugs to a specific project? or just glance bugs with some docs tag? 14:28:08 <brianr> markwash: it's kind of complicated, like take a look at this one:https://bugs.launchpad.net/openstack-manuals/+bug/1195473 14:28:33 <brianr> so it's a bug for openstack-manuals, and it's linked back to glance 14:29:15 <markwash> gotcha, that makes sense 14:29:16 <brianr> i think if we are going to write doc, we do something like that 14:29:34 <brianr> if we just want to alert the doc people, we put the doc impact tag in the regular bug 14:29:46 <iccha> hmm ok so here is the thing. if we make it too complicated it might deter ppl from going into quick update glance docs mode. so either we need one person owning maintaining this docs bugs and make sure they get tagged and etc or ppl being able to just enter what they re doing and jump onto it. 14:30:08 <brianr> iccha: you are right, it is a PITA 14:30:31 <brianr> how about i follow up with Anne and get better instructions about how/where we should file stuff 14:30:55 <iccha> brianr: sounds good and if meanwhile ppl feel like updating glance/doc/source they are welcome to file a bug and do it? 14:31:04 <iccha> thanks for taking the onerous task brianr :) 14:32:15 <markwash> okay, we have *some* action items here 14:32:28 <brianr> yes, i think we own glance/doc/source, so we do whatever we want there 14:32:36 <iccha> so maybe action item is brianr will give us a step by step how to file glance doc bugs and which projects to associate it with? 14:33:05 <iccha> brianr: +1 on us gettign started with this if we want and keep contributing to glance/doc/source but remembering to file a bug for it before 14:33:26 <markwash> and flwang, you had said you were going to look into api-site somewhat? 14:33:52 <flwang> markwash: yep, will check if it's update to date and update it 14:34:08 <flwang> make sense? 14:34:30 <markwash> good 14:34:40 <markwash> let's move on to blueprints, any objections? 14:34:46 <brianr> no objections! 14:34:47 <iccha> go for it! 14:34:54 <markwash> #topic current blueprint progress 14:35:15 <markwash> iccha: brianr indicated you would most likely begin work on api protected properties this week 14:35:22 <markwash> did that come to pass? 14:35:43 <iccha> markwash: some intial work has started https://github.com/isethi/glance/tree/protected_p 14:36:29 <markwash> timeline for H-3 still feels reasonable? 14:36:55 <markwash> iccha: ^^ 14:36:58 <iccha> markwash: yes def doable 14:37:00 <markwash> cool 14:37:09 <brianr> anyone interested in protected properties, please take a look at https://wiki.openstack.org/wiki/Glance-property-protections-product and put questions on the FAQ if you have 'em 14:37:24 <markwash> how about async work, there are a number of people interested in that. . any progress? 14:37:34 <markwash> flwang: brianr ^^ ? 14:37:53 <brianr> flwang and nikhil have been working on an etherpad 14:38:10 <iccha> whats the link ? :p need to add it to https://etherpad.openstack.org/glance_etherpad_list 14:38:15 <brianr> i think anyone interested should take a look, and let's meet today or tomorrow? 14:38:16 <flwang> Ihttps://etherpad.openstack.org/havana-glance-requirements 14:38:24 <markwash> sketching out code design, basically? 14:38:31 <brianr> https://etherpad.openstack.org/havana-glance-requirements 14:38:45 <brianr> and some use cases to provoke design 14:38:49 <nikhil> markwash: code design, rearch, more use cases(corner cases) etc 14:39:01 <nikhil> basically, a careful one to implement so it seems 14:39:16 <nikhil> added some comments based on convo last night 14:39:23 <flwang> nikhil: did you touch the task api part? 14:39:36 <iccha> i remember markwash had some poc work on async workers 14:39:37 <markwash> brianr: meeting tomorrow could be great 14:39:44 <iccha> +1 14:40:04 <nikhil> people state adding in your names there 14:40:11 <flwang> iccha: yep, I know that. I'm going to implement it based on markwash's work 14:40:20 <brianr> flwang: what time zone are you in? 14:40:34 <nikhil> flwang: brianr markwash iccha we need to divide tasks 14:40:36 <flwang> Beijing, China 14:40:49 <nikhil> flwang: he means wrt UTC 14:40:52 <flwang> so I think UTC+8:00 14:41:45 <markwash> I'd like to attend as well 14:41:57 <brianr> would this time work? 14:42:14 <markwash> probably 14:42:15 <nikhil> brianr: may be a doodle? 14:42:35 <nikhil> in UTC :) 14:42:52 <flwang> brianr: you can plan it on my night, it's ok 14:42:57 <brianr> so 14:00 UTC, i think? 14:43:06 <markwash> Friday night glance party! 14:43:10 <flwang> the tomorrow, the same time? 14:43:12 <markwash> byob 14:43:23 <jbresnah> heh 14:43:41 <brianr> etherpad party! 14:43:47 <flwang> brianr: it works for me 14:43:50 <markwash> let's sort this out after the meeting if we aren't already settled 14:44:02 <markwash> zhiyan: got any reviews to highlight for us for your various bps? 14:44:17 <zhiyan> https://review.openstack.org/#/c/37222/3 14:44:26 <zhiyan> probable need flaper87 input: https://review.openstack.org/#/c/37421/1 14:44:57 <markwash> ah yes, that second one looks a little stuck 14:45:12 <markwash> I tried to wrap my head around the issues, and couldn't quite figure it out 14:45:26 <markwash> but I know jerdfelt is a smart guy, so I'm a little worried 14:46:12 <markwash> zhiyan: maybe sometime we can track that down on irc with flaper87 and jerdfelt 14:46:22 <markwash> today or tomorrow. . I think he's in my timezone 14:46:30 <markwash> zhiyan: does that cover it for now? 14:46:35 <zhiyan> and for Scrubber-refactoring, i'm working on it, as our discussed, but from tomorrow to next Tuesday i need attend a meetings in office, will spend some time on that.. 14:46:42 <markwash> gotcha 14:46:56 <markwash> jbresnah: any work on super simple quotas? 14:47:07 <zhiyan> markwash: so, sorry, seems i'm not available tomorrow... 14:47:08 <markwash> (which I forgot to add to the H-3 list, but will fix immediately) 14:47:23 <markwash> zhiyan: I'll ask flaper about it next I see him then 14:47:25 <jbresnah> markwash: yeah i have a patch out there 14:47:32 <jbresnah> https://review.openstack.org/#/c/37993 14:47:49 <jbresnah> zhiyan: left some comments that i am addressing, but the general idea is there 14:48:19 <jbresnah> i also have a nova patch, but that is OT here 14:48:35 <jbresnah> but it relates to direct-url work 14:48:48 <markwash> cool 14:49:11 <markwash> boris-42: your group has a number of db-related patches out for the oslo db blueprints 14:49:24 <markwash> care to mention them or other concerns here? 14:49:29 <boris-42> markwash not so much=) 14:49:54 <boris-42> markwash so we are working in whole openstack, we would like to use common DB code in all projects 14:49:58 <zhiyan> jbresnah: btw, as i mentioned, seems nfs-driver patch maybe drop some challenge in for your 'metadata' idea (which based on file-store) ... 14:50:18 <boris-42> nova, neutron, cinder, ironic already use this code 14:50:27 <boris-42> and there is no problems with it 14:50:49 <markwash> yeah, it looks like there are a number of good fixes in the patches 14:50:59 <iccha> i do know that this will offer some benefits like tpooling and softdelete code. but i am not sure if it has anything missing which glance needs 14:51:30 <markwash> I'm a little meh on the timestamp mixins. . not sure that DRY really goes that far I guess 14:51:40 <boris-42> markwash It will 14:51:46 <boris-42> markwash but it requires migrations 14:51:53 <jbresnah> zhiyan: perhaps, but there are other filesystem stores besides nfs 14:51:53 <zhiyan> iccha: +1 ,also interested in that part. 14:52:03 <jbresnah> and other storage systems besides file, etc 14:52:15 <boris-42> markwash using oslo db code is just start point 14:52:24 <boris-42> markwash of global DB cleanup 14:52:51 <boris-42> markwash such as soft_delete + unique constraints, alembic instead of sqlaclhemy-migrate 14:53:10 <boris-42> markwash and so on 14:53:36 <zhiyan> seems cool 14:53:47 <markwash> boris-42: out of curiousity, have you or your team thought much about non-sql based backends? 14:54:14 <boris-42> we would like to use non-sql, to allow easier way to make now downtime updates 14:54:27 <boris-42> but it will be soooo hard to put it in OpenStack=) 14:54:34 <iccha> markwash: was it you or someone else who was mentioning some sort of db abtraction so its database indepedent 14:54:50 <boris-42> in nova, cinder we have db.api 14:54:54 <boris-42> iccha ^ 14:54:56 <markwash> iccha: I have some work in that direction, but not a lot of push from anyone internally to push it further 14:55:08 <boris-42> it allows us to isolate and implement different backends 14:55:18 <boris-42> But this thing is connected with unit tests 14:55:30 <boris-42> and unit tests should be run also with different backends 14:55:42 <flwang> markwash: did you mean we are interested in using no-sql as the db backend? such as mongo? Hbase? 14:55:51 <markwash> flwang: yes 14:56:03 <flwang> great, I love it 14:56:14 <markwash> mostly I'm interested in redefining the db abstraction layer so that it doesn't implicitly put business logic in the db drivers 14:56:15 <boris-42> markwash flwang first of all we should make our tests independent from sqlalchemy 14:56:32 <markwash> boris-42: I think we have that to a fair degree in glance 14:56:46 <markwash> anyway, I'd be interested in your perspective on what we do have there 14:57:04 <markwash> boris-42: any reviews you want to highlight here? I know we can find them as well, so maybe that is enough 14:57:08 <nikhil> markwash: +1 interested in db abstraction (a lot) 14:57:29 <boris-42> markwash about db abstraction 14:57:42 <flwang> we can take some reference from ceilometer 14:57:50 <esheffield> +1 likewise - I was actually giving some thought to non-sql approaches this morning 14:57:51 <boris-42> markwash I am not sure that full isolation of logic and DB is good poing 14:57:57 <flwang> I will do some investigation 14:57:58 <markwash> #topic open discussion 14:58:02 <boris-42> markwash it is clear 14:58:05 <markwash> (seems like that's what we're doing) 14:58:17 <markwash> boris-42: there are some definite risks 14:58:18 <brianr> quick tasks question: so we have asynchronous image tasks, s'pose someone starts an export and then deletes the image, do we: 14:58:27 <brianr> 1 - end task with error, they were stupid 14:58:28 <brianr> 2 - "lock" the image somehow until the task is completed 14:58:38 <brianr> this will be a problem for cloning since the glance-to-glance coordination may take some time, you originate task request in region T but the image is in region S, someone else working in your account in image S doesn't know about the clone request, and could delete image before it gets cloned 14:58:52 <brianr> i'm wondering whether something like the task_state that nova has on instances might be needed here? 14:58:55 <jbresnah> brianr: there are many issues regarding the asycn task work like that which concern me actually 14:59:14 <boris-42> markwash hmm I have some thoughts about DB code and tests in glance 14:59:27 <iccha> jbresnah: do u see any overlap with staccato? 14:59:29 <boris-42> markwash as we should make baby steps 14:59:30 <bourke> will throw this out, have added a blueprint for porting Swift's ratelimiting to Glance. Is this something people would be interested in? 14:59:34 <markwash> brianr: also, 3 - use reference counts for deletes. . . task would hold a temporary reference. . that might be very strained 14:59:37 <flwang> brianr: we need some "lock" just like Nova does 14:59:44 <brianr> guess we r out of time, should we discuss on mailing list? 15:00:00 <flwang> or the glance irc channel? 15:00:00 <jbresnah> iccha: good questions, in some areas yes 15:00:00 <boris-42> markwash probably it will be best solution to implement db.api as in Nova (it is pretty easy), and remove dependency from sqla in tests?) 15:00:23 <jbresnah> iccha: but i worry about going to far down the road into becoming a generic job scheduler 15:00:25 <brianr> flwang: maybe after async worker meeting tomorrow? 15:00:31 <nikhil> flwang: not on irc 15:00:45 <nikhil> seems like many poepl are interested 15:00:45 <ameade_> time 15:00:46 <ameade_> lol 15:00:51 <nikhil> and don't want to miss on input 15:01:07 <brianr> ok, i will send out my question to mailing list 15:01:07 <flwang> nikhil: ok, got 15:01:11 <jbresnah> brianr: i think you would need tpo reference count the image 15:01:11 <nikhil> thanks 15:01:26 <nikhil> we need atomicity 15:01:26 <flwang> it's time to run 15:01:33 <markwash> #endmeeting