14:03:38 <markwash> #startmeeting glance
14:03:39 <openstack> Meeting started Thu Jul 18 14:03:38 2013 UTC.  The chair is markwash. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:03:40 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:03:42 <openstack> The meeting name has been set to 'glance'
14:03:43 <zhiyan> \o/
14:03:45 <flaper87> \o/
14:03:48 <zhiyan> :)
14:03:52 <markwash> #topic agenda
14:04:03 <markwash> hi folks
14:04:09 <zhiyan> hi markwash
14:04:12 <flaper87> Hi markwash
14:04:14 <flaper87> lol
14:04:44 <markwash> at the end of next meeting, I wanna have a pretty good idea of what we're trying to deliver over havana-3
14:04:45 <zhiyan> btw, thank you all, for review and landing my patch to H2
14:04:57 <markwash> but is there other stuff folks want to talk about today?
14:05:04 <flaper87> zhiyan: thank you for working on that
14:05:05 <zhiyan> https://etherpad.openstack.org/glance-havana3-blueprints
14:05:25 <zhiyan> flaper87: :-D, thanks you!
14:05:38 <markwash> yes, thanks to zhiyan and all the reviewers. . I think ttx was impressed that we didn't have to drop anything in the last two weeks of h2
14:05:54 <zhiyan> coool!
14:06:26 <zhiyan> markwash: could you pls take a look above link? do you think they are make your sense?
14:07:02 <markwash> looking
14:07:23 <markwash> I think those are all good, but I'm still a little unclear about "global state management"
14:07:33 <markwash> that sounds kind of like something nova does that is problematic
14:07:41 <zhiyan> maybe i can create BPs for that (except Enable task, others are all I added)
14:08:14 <flaper87> markwash: following our former plan to get Glance "Out There" - reading, making it a public service - I'd love to see async workers move forward
14:08:14 <markwash> #topic havana 3 blueprints
14:08:25 <markwash> flaper87: +1
14:08:29 <flaper87> This is the patch for that
14:08:31 <flaper87> #link https://blueprints.launchpad.net/glance/+spec/async-glance-workers
14:08:35 <flaper87> blueprint*
14:08:38 <markwash> I think we might have a critical mass at this point
14:08:45 <zhiyan> markwash's draft : https://review.openstack.org/#/c/31874/2
14:08:53 <zhiyan> flaper87: +1
14:08:57 <markwash> I have a better idea how I think it should work (my draft is not up to date)
14:09:16 <flaper87> since the registry stuff is ready, I can focus on that and help you with that blueprint
14:09:49 <zhiyan> markwash: yes, i'd like post our discussing summary (IRC) to a etherpad page, if you like
14:10:28 <flaper87> also, I think this is important - still sticking to our former plan - to have some quotas in glance. IMHO, we should track this down as well
14:10:31 <flaper87> #link https://blueprints.launchpad.net/glance/+spec/glance-basic-quotas
14:11:35 <zhiyan> flaper87: could you pls add them to that etherpad page also?
14:11:43 <markwash> yeah, I think that is just looking for workers at this point
14:12:06 <markwash> s/workers/developers
14:12:11 <zhiyan> :)
14:12:15 <markwash> not to be confused with image workers / async processing
14:12:42 <flaper87> markwash: I know John was looking at it (re quotas)
14:12:44 <zhiyan> maybe that's truth, 'developer' = 'worker' :)
14:13:19 <markwash> we could implement async just as a mechanical turk with glance developers manually doing image conversions on their laptops
14:13:29 <flaper87> markwash: LOOOOL
14:13:52 <markwash> that might cause a security problem I guess :-)
14:13:55 <zhiyan> https://etherpad.openstack.org/glance-async-task-discussing
14:14:24 <markwash> zhiyan thanks
14:15:04 <markwash> flwang: o/
14:15:20 <flwang> markwash:o/
14:16:15 <markwash> zhiyan: can you describe the scrubber changes in some more detail?
14:16:31 <markwash> I'm interested in refactoring it as well, principally to make testing it less brittle and slow
14:17:11 <zhiyan> markwash: ok, for now, the biggest issue for it is about 'multiple-locations' supporting IMO
14:18:38 <markwash> zhiyan: okay cool, makes sense
14:18:44 <zhiyan> currently scrubber not support that at all, actually if image has more then one location there, and glance configured using scrubber to remove image, the glance-api will raise exception when use delete image.
14:19:01 <zhiyan> s/use/user
14:19:15 <flaper87> zhiyan: I think I lost the first message, re scrubber. Sorry. Can you re-send it ?
14:19:28 <zhiyan> and other code also need be refactoring
14:19:47 <zhiyan> ok, for now, the biggest issue for it is about 'multiple-locations' supporting IMO
14:19:57 <flaper87> zhiyan: +1
14:20:28 <markwash> I'm also a bit confused about its architecture, not to complicate the discussion, but. . .
14:20:45 <markwash> it tends to use a local process consuming a file-backed queue of locations to delete
14:20:46 <zhiyan> flaper87: actually, i checked scrubber code, IMO it's not so clear...
14:21:09 <markwash> though it can also periodically query the database and shove images into the local queue for processing
14:21:14 <markwash> its very scatter brained
14:21:37 <markwash> and yet, its only a small LOC
14:21:58 <markwash> meaning to say, its not that hard to replace
14:22:23 <zhiyan> markwash: yes, but i'm not followed your last point ..
14:22:46 <markwash> when I was looking at it for testing purposes, I thought the best thing would be a rewrite
14:22:47 <zhiyan> markwash: are you meaning scrubber can be removing? or some change
14:23:07 <markwash> I don't think we can remove it, because we need something to handle pending deletes
14:23:08 <zhiyan> markwash: ok, for change what?
14:23:45 <zhiyan> markwash: yes, i think so. but not clear what you want to rewrite..sorry
14:24:06 <markwash> zhiyan: I was just thinking of completely rewriting the component that handles pending_deletes
14:24:13 <markwash> so probably completely replace glance-scrubber
14:24:26 <zhiyan> markwash: using async-task?
14:24:37 <zhiyan> or just internal refactoring?
14:24:43 <markwash> just internal refactoring
14:24:47 <zhiyan> for glance-scrubber .. ok
14:25:15 <zhiyan> i will check it when i do that, and I'd like / we can discussing details.
14:25:29 <flaper87> I haven't looked at scrubber that much
14:25:48 <zhiyan> flaper87: it's easy, small
14:26:34 <zhiyan> but as markwash said, currently it read delete 'task/request' from file
14:26:36 <markwash> so, I mention that only as a green light to folks if they want to be a bit drastic with it
14:26:46 <flaper87> zhiyan: yeah, I wonder if we'll be ever be able to get rid of it, perhaps using async-workers
14:27:13 <zhiyan> markwash: aha, thanks.
14:27:44 <markwash> zhiyan: are those blueprints in the etherpad that you added all things you plan to work on?
14:27:54 <markwash> or are some of them just general suggestions of things you think need to be done
14:28:27 <zhiyan> flaper87: yes, async-task will be good. asynchronous delete image. but there is a issue for async-task, it need support reliable persistent task queue supporing.
14:29:10 <zhiyan> markwash: i'd like create them, but i'm not sure i have enough bandwidth do all of them, you know.
14:29:16 <flaper87> zhiyan: agreed
14:29:30 <flaper87> markwash: btw, I remembered another blueprint that I think is important for H-3
14:29:41 <markwash> zhiyan: makes sense, can you put your name by the ones you *definitely* want to work on, including teh ones already in progress?
14:29:41 <flaper87> markwash: store credentials being saved in the DB
14:29:55 <markwash> flaper87: ah good point, I think iccha has some current work on that
14:29:58 * flaper87 is trying to find the blueprint
14:30:19 <markwash> unfortunately the rackspace folks cannot be here today, they have a corporate thing, not just protesting us :-)
14:30:28 <zhiyan> markwash: sure, will do.
14:30:30 <flaper87> markwash: yeah, she mentioned it, we should sync with her about that
14:30:32 <flaper87> markwash: LOL
14:30:51 <flwang> markwash: do we want plan the task bp into h-3?
14:31:06 <markwash> flwang: I think we do
14:31:12 <zhiyan> https://review.openstack.org/#/c/34801/ ?
14:31:16 <markwash> flwang: it is very important and there are several interested folks
14:31:17 <flaper87> markwash: and this one: https://blueprints.launchpad.net/glance/+spec/glance-cache-path
14:31:51 <flwang> markwash: yep, but obviously, we need more discussion
14:31:55 <flaper87> I think the cache one is pretyt important as well.
14:32:10 <flwang> flaper87: agree +1
14:32:26 <flwang> I'd like to see the cache management is improved
14:32:37 <markwash> yeah, there have been some definite requests around that
14:33:21 <zhiyan> flaper87: markwash: i really think we should write them down to etherpad page
14:33:32 <markwash> +1
14:33:35 <markwash> I just added the cache path one
14:33:55 * flaper87 gets the other ones
14:33:56 <zhiyan> thanks
14:34:48 <zhiyan> btw, i just know iccha is a lady...since flaper87 use 'she'..
14:34:55 <markwash> flwang: you're planning on working on the task / async stuff during h3, correct?
14:35:05 <flwang> yep
14:35:10 <flaper87> zhiyan: she is and she rocks! :D
14:35:14 <flwang> but I think I need you guys support
14:35:22 <flwang> to make sure the design is correct :)
14:35:23 <markwash> flwang: sure thing
14:35:28 <flaper87> flwang: +1
14:35:42 <flaper87> flwang: where are you based?
14:35:43 <markwash> also there are some other folks for whom async processing is very important, and we need to make sure they are in the discussion
14:35:57 <flwang> flaper87, Beijing, China
14:36:10 <flaper87> flwang: cool!
14:36:11 <markwash> currently nikhil is assigned on the blueprint, but I know he's been very busy with some other work as well and would welcome your work
14:36:41 <flwang> markwash: yep, we have started some discussion since yesterday
14:37:11 <flwang> flaper87: where are you based?
14:37:14 <zhiyan> flwang: paste again for you, a summary we discussed yesterday. https://etherpad.openstack.org/glance-async-task-discussing
14:37:34 <flwang> zhiyan: great, thanks
14:37:36 <flaper87> flwang: I'll take the cache management one, hopefully, the common oslo cache will land soon as well and we can pull that in
14:37:40 <flaper87> flwang: Como, Italy
14:38:03 <flaper87> flwang: pull me in into the async workers discussions, pls :)
14:38:27 <markwash> flaper87: have you seen https://blueprints.launchpad.net/glance/+spec/refactoring-move-caching-out-of-middleware ?
14:38:33 <flwang> flaper87: nice, I love pizza
14:38:43 <zhiyan> flwang: markwash: and me pls
14:38:47 <flwang> flaper87: sure thing
14:39:21 <flwang> zhiyan: sure
14:40:19 <markwash> flaper87: I'd be interested in seeing if we can address that refactoring during the process, or if its too much to take on
14:40:30 <flaper87> markwash: mmh, I think I did and forgot. That makes more sense than having a cache service as we discussed once
14:40:41 <markwash> I'm pretty sad about the way the cache lives at the wsgi middleware layer and thus has to duplicate all the formatting logic
14:40:54 <flaper87> markwash: I think we can do it and it makes sense
14:41:06 <flwang> markwash: +1
14:41:15 <flaper87> also, multiple-location landed already, that will help us on addressing john's comments as well
14:42:05 <markwash> I think we've got good initial coverage of our h3 blueprints, I'll try to update the pages this week and follow up with questions for folks
14:42:10 <zhiyan> markwash: flaper87: the BP make sense to me, but i don't think jbresnah's thoughts is right (in whiteboard) maybe i not followed his comments well
14:42:59 <flaper87> zhiyan: that came from another discussion about having some kind of service for it, IIRC. I'm pretty sure we have to discuss that a bit further
14:43:05 <flaper87> perhaps in our next meeting
14:43:05 <markwash> zhiyan: I think the idea is that for admins, cache could be managed as another location on the image
14:43:25 <markwash> I'd like to open up the dicussion in case folks have general issues they want to bring up
14:43:34 <flaper87> markwash: o/
14:43:43 <markwash> #topic open discussion
14:43:44 <zhiyan> markwash: i'd like to know them all
14:43:55 <flaper87> o/
14:44:14 <markwash> flaper87: go for it
14:44:28 <zhiyan> markwash: do you think it's time to take a look again on https://blueprints.launchpad.net/nova/+spec/image-multiple-location
14:44:38 <zhiyan> it's for nova side, but related with glance
14:44:44 <flaper87> I'm pretty concerned about the client and it's support for glance's v2, I know iccha and esheffield are working on it
14:44:58 <zhiyan> thanks for you earlier input for that btw.
14:45:20 <flaper87> so, I was planning to take less bps for H-3 and dedicate some more time to the client
14:45:31 <zhiyan> flaper87: oh, yes, cinder guy ping me again for it, in yesterday cinder weekly meeting
14:45:38 <flaper87> but, the client needs some review-help
14:45:39 <markwash> flaper87: that would make me crazy happy
14:45:56 <zhiyan> flaper87: just because glanceclient has gay on v2 client, causes cinder has some limitation
14:46:00 <zhiyan> markwash know that
14:46:04 <zhiyan> a defect
14:46:48 <flaper87> markwash: cool, so, I'll sync with iccha and esheffield and dedicate more time to it
14:46:55 <markwash> flaper87: noted, I'll double down my efforts to review client patches
14:47:12 <markwash> should be easy to do, after all 2 * 0 = 0 :-)
14:47:14 <flaper87> markwash: question: How are client releases planned? I know those are just tags but when do they happen?
14:47:30 <markwash> it essentially has its own roadmap
14:47:49 <markwash> for now, the roadmap issues are v2 support and deprecating legacy I believe
14:47:59 <flaper87> markwash: ok, so it's feature based
14:48:09 <flaper87> once those are covered, we can release a new version
14:48:11 <markwash> yeah, it is not tied into the general openstack release
14:48:28 <flaper87> oke-doke!
14:48:39 <markwash> I'm still really interested in someone basically taking on the client as pseudo-ptl, honestly
14:48:52 <markwash> because I keep accidentally ignoring it to focus on the server side
14:48:58 <flaper87> markwash: I raised my hand the last time :D
14:49:27 <markwash> I think you'd be a good fit, but we probably have to sync up more to make the transition
14:49:33 <zhiyan> markwash: flaper87: can you take a look on https://review.openstack.org/#/c/37616/ ?
14:49:51 <zhiyan> for resolve this: https://bugs.launchpad.net/glance/+bug/1202391
14:49:56 <markwash> zhiyan: yes, thanks for that
14:50:24 <flaper87> markwash: yup, we should also ask iccha!
14:50:33 <zhiyan> +2 pls, if you ok, just don't like other guys say my patch block something. :)
14:50:35 <flwang> is it time to call for reviewer ? :)
14:50:39 <flaper87> oke, that was my point for the open discussion
14:50:41 <flaper87> :D
14:51:16 <bourke> was wondering did anyone have any thoughts on the topic of unicode in exceptions...
14:51:29 <flaper87> bourke: ?
14:51:37 <zhiyan> bourke: https://review.openstack.org/#/c/37421/ ?
14:51:47 <bourke> flaper87: both yourself and zhiyan have a nice fairly similar solution there
14:52:03 <flaper87> zhiyan: https://review.openstack.org/#/c/36658/
14:52:05 <flaper87> :D
14:52:07 <zhiyan> yes, i lean from flaper87 :)
14:52:30 <bourke> flaper87: but was wondering what about other exceptions. is it something that could be wrapped in a function and used in multiple places?
14:52:32 <zhiyan> aha, since you give me a comment, so i checked your patch, and got it
14:52:50 <flwang> markwash: flaper87: may I get your comments about this https://review.openstack.org/#/c/37511/   ?
14:53:06 <zhiyan> bourke: can we take care that be case by case? defect driven maybe more easy
14:53:15 <flaper87> zhiyan: +!
14:53:17 <flaper87> zhiyan: +1
14:53:36 <bourke> sure thing
14:54:03 <zhiyan> bourke: thanks you for you raise this up, i just forget it...
14:54:25 <markwash> flwang: I like that you are getting the latest oslo.rpc code, but I'd still slightly prefer to see it in another patch
14:54:42 <markwash> flwang: maybe you can ask the other submitter to update their patch with the latest code?
14:54:47 <zhiyan> markwash: yes https://review.openstack.org/#/c/37184/
14:55:05 <flaper87> flwang: have you looked at the new oslo.messaging?
14:55:47 <flaper87> I wonder whether it is worth to pull oslo's rpc code into glance instead of waiting 'til new oslo.messaging is ready
14:55:57 <zhiyan> 5 mins left
14:56:24 <flwang> flaper87: nope, what's up?
14:56:26 <markwash> flaper87: I think the goal here is to use notifier, which depends on rpc
14:56:47 <flwang> markwash: yes
14:56:49 <flaper87> markwash: indeed, that notifier will depend on oslo.messaging by the end of H-3
14:57:05 <markwash> flaper87: well, that's pretty nice
14:57:05 <bourke> oh, one other thing, if someone could take a look at https://review.openstack.org/#/c/35286/ that would be much appreciated
14:57:18 <flaper87> is it worth to pull all that in instead of waiting for that migration to happen ?
14:57:20 <flwang> flaper87: any blueprint or bug to track that?
14:57:31 <flaper87> flwang: erm, let me check
14:57:37 <markwash> bourke: looks like an easy +2 approve, will check immediately after mtg
14:57:52 <bourke> markwash: thanks!
14:58:07 <markwash> flaper87: do you know if its like "early h3" for being ready? or last minute?
14:58:13 <zhiyan> and https://review.openstack.org/#/c/36884/  and https://review.openstack.org/#/c/37222/
14:58:30 <flwang> flaper87: it oslo.messaging is the direction, it would be nice to do that
14:58:32 <flaper87> markwash: I'd say last minute
14:59:14 <markwash> any other folks have last minute notices or announcements?
14:59:19 <flwang> markwash: so I still can't get approve for my fix ? O:-)
14:59:34 <flwang> flaper87: any comments?
14:59:50 <flaper87> flwang: TBH, I'd wait for that before pulling oslo's notification in
15:00:11 <flaper87> because oslo.rpc has some config parameters that might change when oslo.messaging is complete
15:00:30 <zhiyan> markwash: just notice, pls review https://review.openstack.org/#/c/37616/ for fix https://bugs.launchpad.net/glance/+bug/1202391
15:00:35 <flaper87> among other things
15:00:58 <markwash> flaper87: that makes sense
15:01:30 <flwang> markwash: how do you think? IMO, I think we can do the refactoring for notifier and follow up the change of oslo.messaging
15:01:50 <markwash> we can, but it increases the operational burden
15:01:56 <markwash> I think we have to clear out now folks
15:02:06 <markwash> thanks errbody
15:02:10 <flaper87> kk, good meeting! Thanks to all!!
15:02:11 <markwash> #endmeeting