20:02:03 <markwash> #startmeeting glance
20:02:04 <openstack> Meeting started Thu Jul 25 20:02:03 2013 UTC.  The chair is markwash. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:02:05 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:02:06 <markwash> hi everybody
20:02:07 <openstack> The meeting name has been set to 'glance'
20:02:15 <rosmaita> hi mark
20:02:16 * jbresnah waves
20:02:16 <ameade_> hey
20:02:29 <zhiyan> \o/
20:02:35 <markwash> my apologies if I have seemed or do seem distracted this week
20:02:52 <markwash> I caught a bug earlier on, and treatment has kept me a bit spaced-out
20:03:29 <jbresnah> Did you file the bug on launchpad?
20:03:43 <jbresnah> I would be happy to triage it
20:03:44 <markwash> the broad topic i want us to address during the meeting today is what we expect/plan to accomplish in H-3
20:03:47 <markwash> lol
20:03:58 <markwash> but due to being out sick, I haven't developed a specific agenda
20:04:32 <markwash> so lets do a quick once-around-the-room to get agenda items from everyone, to make sure we stick to a sensible structure
20:04:36 <jbresnah> For h3 i would like to get a patch for storage quota
20:04:44 <zhiyan> ify, we discussed in last meeting: https://etherpad.openstack.org/glance-havana3-blueprints
20:04:44 <jbresnah> but that is all i have in mind right now
20:06:14 <rosmaita> new upload-download workflow recast as "tasks" api
20:06:40 <rosmaita> property protections
20:06:44 <markwash> as an additional item, i think it makes sense to take another look at https://launchpad.net/glance/+milestone/havana-3
20:07:25 <markwash> I wanna talk a little bit about focusing on bug triage and fixes during H-3, so I'll add that to my list as well
20:08:11 <markwash> okay, I've got 4 items, storage quotas, property protections, task api, and bug triage
20:08:13 <markwash> anything else?
20:08:51 <jbresnah> not from me
20:09:06 <markwash> well, and "other h-3 blueprints" if we have time
20:09:15 <markwash> okay, rosmaita wanna kick things off with tasks api?
20:09:17 <zhiyan> i have 3, we discussed yesterday
20:09:23 <rosmaita> sure
20:09:25 <markwash> #topic task api
20:09:36 <rosmaita> https://wiki.openstack.org/wiki/Glance-tasks-api
20:09:39 * markwash hands over the mic
20:10:05 <rosmaita> the idea is that we can unify import/export/clone in a new uri path
20:10:19 <rosmaita> for these operations, we will return a "task" resource
20:10:29 <rosmaita> that you can poll to see what's up
20:10:41 <rosmaita> eventually  you either get success or failure status
20:10:52 <rosmaita> and the think has an expiration date, so it can be deleted
20:11:07 <rosmaita> that doc above gives the basic outline
20:11:21 <rosmaita> it's got links to proposed use for import, export, and clone
20:11:32 <rosmaita> there was some discussion on the mailing list
20:11:53 <rosmaita> some people would prefer using the image response for this
20:12:13 <rosmaita> but i think that's a bad idea, these operations may result in a non-image
20:12:16 <jbresnah> i missed that discussion on the ML
20:12:19 <rosmaita> so why clog up the db
20:12:31 <rosmaita> there's a ref on the above doc if you want to review
20:12:31 <jbresnah> what is an example non-image?
20:12:34 <rosmaita> was a few months back
20:12:42 <markwash> jbresnah: an image import that fails
20:12:51 <jbresnah> fair
20:12:59 <markwash> rosmaita: I agree 100%
20:13:00 <jbresnah> yeah, i like it to not be an image
20:13:14 <markwash> I don't want to always use my PTL hat, but I'm putting it on to say no to more complex internal image state
20:13:16 <rosmaita> i think the current glance people are ok with this
20:13:21 <jbresnah> why is it v2?
20:13:26 <jbresnah> as in /v2/tasks
20:13:51 <nikhil> caz v1 prolly should not support that
20:14:02 <rosmaita> well, we (as in rackspace) are thinking of only going public with v2
20:14:03 <jbresnah> the task portion seems like a new API all together ot me
20:14:26 <jbresnah> currently v1 v2 ar efor image handing, and now there is new API for tasks
20:14:31 <markwash> rosmaita: thanks for the overview, this looks good to me, is there a best place for folks to provide more feedback?
20:14:35 <jbresnah> but very minor point
20:14:37 <jbresnah> just caught my eye
20:14:47 <jbresnah> API-wise i think that is great
20:14:59 <jbresnah> i would like to hear about some impl details and architecture
20:15:05 <rosmaita> don't have a feedback point, what would be best to use?
20:15:26 <rosmaita> don't think there's much impl details yet, been working on the api mostly
20:15:37 <rosmaita> there are implied details in the individual docs, though
20:15:39 <markwash> I don't really want to trigger another internal vs external state discussion honestly, but I guess the ML is still best
20:15:51 <rosmaita> ok, i will post after the mtg
20:15:58 <rosmaita> and hope for the best
20:16:18 <markwash> rosmaita: one thing I want to make sure you know, is that there are some other folks outside of your org interested in working on this
20:16:31 <rosmaita> that is good news
20:16:35 <markwash> in particular I know flwang is interested in some of the implementation details, and he and I have had some discussions
20:16:56 <markwash> rosmaita: how much of this are you imagining will be done for H-3 at this point?
20:17:09 <nikhil> markwash: rosmaita yeah flwang and I had a chat
20:17:17 <markwash> nikhil: oh good :-)
20:17:23 <nikhil> markwash: he seemes to be of the opinion that everone is on design phase
20:17:32 <jbresnah> i would like to work on it as well (i have some reasons to want this as well)
20:17:39 <nikhil> not sure how far along are you both on that one
20:17:40 <rosmaita> i think a preliminary implementation, e.g., import with just 1 file format conversion
20:17:41 <jbresnah> tho i am not sure i will be able to in the h3 time
20:17:48 <rosmaita> same for export
20:17:48 <nikhil> jbresnah: i mentioned it to him that you might be interested too
20:17:53 <rosmaita> not sure about cloning, TBH
20:18:00 <markwash> with rosmaita's recent work, i think we're at a great state in terms of the API design
20:18:16 * rosmaita blushes
20:18:18 <markwash> there are some details to be worked out internally, but I think we can iterate on that pretty safely
20:18:27 <nikhil> jbresnah: just based on our old discussion
20:18:39 <jbresnah> nod
20:18:47 <markwash> rosmaita: okay, so up to "import" with a simple conversion step for H-3?
20:19:00 <jbresnah> i am very interested in what it would take to add a new backend task
20:19:02 <nikhil> markwash: would like to sync up with you on cloning bp after the meeting
20:19:09 <nikhil> if you'r free for a few mins?
20:19:15 <rosmaita> markwash: i think so
20:19:33 <markwash> nikhil: not sure yet, but sometime soon for sure
20:19:35 <rosmaita> definitely import, possibly export too
20:19:35 <nikhil> jbresnah: +1
20:19:45 <markwash> okay, jbresnah wanna move on to talking about storage quotas?
20:19:46 <nikhil> markwash: will send an email
20:19:51 <jbresnah> sure
20:19:56 <markwash> #topic storage quotas
20:20:11 <jbresnah> I am working through a patch to do something very simple
20:20:33 <jbresnah> just enforce a maximum allowed storage usage across all storage systems
20:20:51 <jbresnah> for now the quota would be set as a single value in conf and apply to all users
20:21:23 <jbresnah> the idea being that when keystone has more sophisticated ways of setting user based quota information taht the value will come from a keystone query
20:21:36 <jbresnah> but the enforcement code will all be the same
20:21:44 <markwash> sounds pretty straightforward
20:21:58 <markwash> I was a bit meh on quotas earlier in the cycle, but this sounds reasonable to me
20:21:59 <jbresnah> yeah not much to it
20:22:09 <jbresnah> oh, and i was thinking of only putting it into v2 as well
20:22:16 <jbresnah> at leastfor the first patch
20:22:24 <jbresnah> to keep it small and easier to review
20:22:28 <jbresnah> how does that sound?
20:22:36 <markwash> jbresnah: sounds like something that could be easily done in H-3, do you just wanna take feedback on the patch? or do folks need a way to talk this out more before you finish the impl?
20:22:55 <jbresnah> yeah h3
20:23:00 <jbresnah> i am good with just getting feedback on the patch
20:23:07 <jbresnah> unless people are fully against the idea
20:23:16 <jbresnah> in which case i would like to know sooner
20:23:23 <markwash> anybody worried about just giving feedback on the patch? I assume the default setting is "no limit" so it shouldn't be too binding
20:23:32 <jbresnah> right
20:24:04 <zhiyan> i just think how it work in multiple-location context
20:24:24 <jbresnah> each location counts against the quota
20:24:37 <zhiyan> you mean 'each store'?
20:24:50 <markwash> that is an interesting question
20:25:01 <jbresnah> right
20:25:09 <jbresnah> it is about total usage
20:25:18 <markwash> if I use my own swift account, and add a location from it, does that count against my shared storage quota?
20:25:31 <jbresnah> i could imagine a later patch putting quota on a per-storage system basis
20:25:44 <jbresnah> markwash: for this yes
20:26:00 <markwash> okiedokie, let's keep thinking about it but I think we can still just look at the patch
20:26:21 <markwash> can I move on to bug triage?
20:26:24 <zhiyan> i'd like check details from patch for that.
20:26:26 <jbresnah> nod
20:26:29 <zhiyan> pls
20:26:33 <markwash> #topic bug triage in H-3
20:26:59 <markwash> since we pretty much can't merge anymore after H-3, I think it makes sense to make a few passes through the bug list to try to fix anything critical
20:27:17 <markwash> so all I'm really asking is for core folks to devote a little extra time to bug management during the next months
20:27:41 <markwash> after we do some triage, it might make sense to have an actual "bug squash" day, but let's hold off
20:27:50 <jbresnah> yeah that sounds like a reasonable request
20:28:18 <ameade_> markwash: oh hey just thought about the mox thread on the ML, should we talk about that?
20:28:47 <markwash> ameade, added to list
20:28:51 <markwash> okay, I think that's all from me
20:29:01 <markwash> rosmaita, wanna move on to prop protections?
20:29:30 <rosmaita> ok, so here's something to look at: https://wiki.openstack.org/wiki/Glance-property-protections-product
20:29:35 <markwash> #topic property protections
20:30:00 <rosmaita> so the above is a "product" approach, it's kind of self-explanatory, would appreciate feedback
20:30:22 <rosmaita> anyway, i think we're ready here to start work on this tomorrow
20:30:38 <markwash> okay, cool, so H-3 feels good?
20:30:45 <rosmaita> definitely
20:31:01 <markwash> one note on the FAQ, you suggest that communication about the protections would be out-of-band
20:31:09 <markwash> but could we also use the schema to communicate those restrictions?
20:31:25 <markwash> maybe some restrictions are supposed to be 'secret' though. . .
20:31:34 <rosmaita> that's what i was thinking
20:31:41 <rosmaita> the nonreadable properties
20:31:46 <jbresnah> rosmaita: recently we added a notion of "meta data" on storage system locations
20:32:07 <jbresnah> i would be great if these protection mechanisms could apply there as well
20:32:37 <rosmaita> ok, will have to look
20:32:47 <rosmaita> not sure about h-3 for that though
20:32:50 <markwash> hmm, jbresnah would you buy that as a second pass? it sounds like it might be hard to do at first
20:33:07 <jbresnah> yeah, also i can make that my problem
20:33:17 <rosmaita> once we get it right for image properties, might not be hard to port to location
20:33:26 <markwash> jbp - jbresnah's problems
20:33:34 <jbresnah> just keep me abreast of whats up
20:33:38 <jbresnah> heh
20:33:39 <rosmaita> ok
20:33:51 <jbresnah> 99 jbps but protection aint one?
20:34:09 <jbresnah> ...i probably should have gone with properties there
20:34:11 <markwash> rosmaita: have you heard any more from smclaren about property protections, do you think most folks are on board?
20:34:16 <markwash> haha
20:34:31 <markwash> jbresnah: careful, we'll quote that in a court in sweden and then you'll have to learn a lot about extradition law
20:34:45 <rosmaita> markwash: have not heard more, will contact him to ask
20:35:01 <rosmaita> not sure if he saw your comments on the etherpad
20:35:15 <jbresnah> heh
20:35:26 <rosmaita> should i send out a mailing list item on property protections?
20:35:36 <markwash> sounds good, some sort of stub for more feedback
20:35:42 <rosmaita> ok, will do
20:35:56 <markwash> all right, one more from me
20:36:02 <markwash> #topic other havana-3 blueprints
20:36:10 <markwash> #link https://launchpad.net/glance/+milestone/havana-3
20:36:24 <markwash> we need to take a very realistic view of these to try to keep our promises this time around
20:36:57 <markwash> if you're assigned there, and think there's any chance your bp won't make it for H-3, talk to me and see what we can do to either pare down scope, or push off to I
20:37:07 <jbresnah> 2 - 3 seem to be all about the task work, is that right?
20:37:24 <markwash> some of those we already covered a bit in this meeting, so I"ll be updating based on my notes later today
20:38:30 <markwash> I should also add I'm still very happy about what we delivered in H-2
20:38:32 <zhiyan> markwash: do you think we can talk about 'global state machine to maintain image status'? or discussing that later when i start do that?
20:38:48 <markwash> zhiyan: looks like we should have time
20:38:53 <markwash> #topic mox thread
20:38:56 <markwash> ameade, take the lead
20:39:05 <ameade_> #link http://lists.openstack.org/pipermail/openstack-dev/2013-July/012474.html
20:39:06 <markwash> also can I get a link on that? I'm not up to speed
20:39:08 <markwash> yes!
20:39:10 <markwash> read my mind!
20:39:21 <ameade_> basically mox isn't python3 compatible
20:39:39 <ameade_> I think we shoudl do both things chuck suggests there
20:39:56 <ameade_> #2 is the quick solution and we can also slowly phase out mox
20:40:01 <markwash> I would like to see if we can slowly move towards mock
20:40:13 <markwash> mostly because I tend to like the tests it results in a bit better
20:40:14 <ameade_> +1
20:40:26 <ameade_> russell sent an email about this
20:40:29 <ameade_> #link http://lists.openstack.org/pipermail/openstack-dev/2013-July/012484.html
20:40:29 <jbresnah> ameade: just reading that email i would say pymox sounds perfect
20:40:36 <jbresnah> but i wonder what the rest of the debate is
20:40:41 <markwash> so maybe we port to pymox, and set up a policy that says "no more mox" ? and try to enforce that in review?
20:41:06 <ameade_> +1, the first part of that is the important part, just want to make sure everyone is aware
20:41:07 <markwash> apparently I'm walking the same path russellb has paved
20:41:34 <esheffield> markwash: are you saying existing mox -> pymox but new tests should use mock?
20:41:38 <jbresnah> in glance there are only a handful of mox tests
20:41:41 <jbresnah> so this should not be so hard
20:41:53 <markwash> esheffield: something like that, if folks don't hate the idea too much :-)
20:42:12 <jbresnah> markwash: i would go further and say pick the right one and port the tests
20:42:19 <jbresnah> just because glance does not have that many mox tests
20:42:31 <markwash> jbresnah: +1 if that's easy enough
20:42:44 <markwash> I can't say I can do all the porting required, but if someone can I'd be eager to review and approve
20:42:47 <jbresnah> markwash: did i just dig myself a hole?
20:42:50 <markwash> :-)
20:43:07 <jbresnah> i can be involved
20:43:16 <jbresnah> do some porting
20:43:18 <jbresnah> and reviewing
20:43:20 <markwash> jbresnah: let's just say someone can opt-in to porting all the tests :-)
20:43:52 <markwash> #topic global state management
20:44:04 <markwash> zhiyan can you help me understand what kind of changes you mean here? I'm still a little lost
20:44:44 <ameade_> dibs! i'll port glanceclient to mock
20:45:00 <zhiyan> we just have concern about consistent image status management in multiple-locations patchs.
20:45:48 <jbresnah> zhiyan: that seems like a big change to me
20:46:00 <zhiyan> so for that maybe we need extract those code from controller layer to a dedicated/common place
20:46:01 <jbresnah> zhiyan: i honestly dont see it happening to glance at this stage in its life
20:46:07 <zhiyan> jbresnah: yes, probable
20:46:28 <markwash> zhiyan: I think I can see what you're saying then, I'd actually like to see the domain model restructured a bit to move all status updates into glance/domain/__init__.py
20:46:32 <jbresnah> maybe, but it would be neccessarily disruptive.
20:46:33 <markwash> there are some challenges presently to doing that
20:46:53 <markwash> this is honestly the kind of refactoring that I'd really like to see happen in between H-3 and the I summit
20:47:02 <jbresnah> dont take my negativitly wrong tho, i would love to see it happen
20:47:35 <jbresnah> markwash: yeah probbly best to have ti at the beginning of a cycle then the end
20:48:00 <jbresnah> but at the same time, i am reminded of netscape re-writing its code base
20:48:03 <zhiyan> markwash: ok, i think so. i can think about it to get a draft idea later, and discuss with team, then to start it.
20:48:11 <jbresnah> and the advice i once got that 'all software sucks'
20:48:11 <markwash> zhiyan: that might also free up more time for you to focus on some of your other bps, and in other projects where there might be some headwinds
20:48:30 <jbresnah> i fear such a change could be more disruptive than helpful
20:48:38 <markwash> jbresnah: I think its less code than you're thinking
20:48:49 <jbresnah> markwash: ok cool
20:48:51 <markwash> but not sure I have the same picture in my mind as others
20:48:55 <jbresnah> markwash: if so then i am all about it
20:49:07 <markwash> okay, I think we've got a good bit for open discussion
20:49:11 <markwash> #topic open discussion
20:49:16 * markwash hopes he didn't miss anything!
20:49:26 <zhiyan> markwash: do we have plan in h3 to update glanceclient to get fully v2 api support?
20:49:36 <markwash> oh man that is a good topic
20:49:39 <jbresnah> if people have a chance please checkout this patch: https://review.openstack.org/#/c/38606/
20:49:51 <jbresnah> flavio fixed a bug that was hurting me
20:49:59 <jbresnah> it is pretty straight forward
20:50:01 <markwash> zhiyan: we really need that, and I think esheffield has some existing work on it
20:50:09 <markwash> esheffield: does that sound correct?
20:50:15 <jbresnah> zhiyan: +1
20:50:18 <jbresnah> i would like to have that too
20:50:47 <esheffield> yes, I have a couple of patches out, and the Thoughtworks team we're working with have done some work as well
20:51:07 <markwash> great, esheffield please feel free to keep contacting me directly to keep me on task for the reviews
20:51:15 <zhiyan> esheffield: markwash: thanks for the update. yesterday cinder folks ask me again.
20:51:25 <esheffield> Sure thing
20:51:50 <esheffield> The changes I have out are for create https://review.openstack.org/#/c/35504/
20:51:54 <esheffield> and upload https://review.openstack.org/#/c/38359/
20:52:49 <markwash> fantastic
20:53:18 <markwash> well, sounds like we might have dried up on open discussion topics
20:53:22 <zhiyan> an other one from me is that: do you think it will be better if we move some properties from image level to image location's level? such as 'is_public' is good for a particular location since currently we have multiple locations for an image?
20:53:22 <esheffield> the cli side will take a bit more work, but I think the library side of this is pretty close
20:53:29 <markwash> nm
20:53:49 <jbresnah> zhiyan: on the surface that makes sense
20:53:53 <markwash> zhiyan: in general, yes we need to move some things down to the location
20:54:01 <markwash> I wonder if status could actually move there too
20:54:06 <markwash> or some portion of it
20:54:14 <zhiyan> nod
20:54:21 <markwash> like pending delete, or saving
20:54:39 <zhiyan> we be we can think about it also, and discussing on this summit too
20:54:50 <jbresnah> gotta be careful with backward compat in wire protocol
20:55:10 <markwash> is it time for us to rewrite glance as a protobuf rpc service in go?
20:55:19 <jbresnah> heh
20:55:38 <ameade_> yes long overdue
20:55:40 <zhiyan> markwash: do you think so?
20:55:59 <markwash> zhiyan: :-) only a little
20:56:55 <zhiyan> sorry, i think it's worth thing, or say under multiple-location situation, some properties should belong with location but image...
20:57:31 <markwash> zhiyan: oh sorry thought you were asking about my protocol buffers and go comment
20:57:32 <markwash> :-)
20:57:41 <jbresnah> zhiyan: i agree with you
20:57:43 <markwash> zhiyan: definitey +1 for moving the appropriate properties down to locations
20:57:47 <jbresnah> that should probably be a new bp
20:57:49 <markwash> I think that's going to be a big improvement
20:57:54 <jbresnah> and put through those paces
20:58:03 <markwash> it might be hard to do with the proper backwards compat, but I think it can be done to some degree
20:58:05 <zhiyan> markwash: do you think it should happen in h3?
20:58:21 <markwash> zhiyan: I'm not worried about that right away
20:58:34 <markwash> I'd love to see a strong definition of what's needed, so we can hammer away at it during the summit
20:58:44 <jbresnah> +1
20:58:47 <zhiyan> so, i think we can discussing it in summit...and make it in I?
20:58:56 <markwash> sounds good to me
20:58:58 <jbresnah> . o O ( Tho I wont be at the summit <tears> )
20:58:59 <zhiyan> great!
20:59:39 <zhiyan> jbresnah: don't worry, i will think about how to involve you in
20:59:54 <iccha> if the internet works :)
20:59:56 <zhiyan> and hope you can get there. :)
20:59:56 <markwash> jbresnah: can't you just take a boat?
21:00:17 <jbresnah> markwash: you would think i could, i am not all that far
21:00:50 <markwash> okay, i got another meeting, thanks everybody!
21:00:54 <markwash> #endmeeting