14:00:55 <markwash> #startmeeting glance
14:00:57 <openstack> Meeting started Thu Apr 10 14:00:55 2014 UTC and is due to finish in 60 minutes.  The chair is markwash. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:58 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:01:00 <openstack> The meeting name has been set to 'glance'
14:01:37 <arnaud> o/
14:01:39 <markwash> roll calll!
14:01:40 <markwash> o/
14:01:44 <nikhil___> o/
14:01:47 <ameade> o/
14:01:48 <zhiyan> hi
14:01:58 <zhiyan> o/
14:02:06 <markwash> hi zhiyan
14:02:20 <zhiyan> hi markwash
14:02:39 <markwash> okay lets talk about icehouse rc2
14:02:43 <markwash> #topic icehouse rc2
14:03:26 <markwash> #link https://launchpad.net/glance/+milestone/icehouse-rc2
14:03:53 <esheffie1d> o/
14:04:08 <markwash> also
14:04:59 <markwash> #link https://bugs.launchpad.net/glance/?field.tag=icehouse-rc-potential&search=Search
14:05:05 <zhiyan> esheffie1d: hi
14:05:37 <markwash> icehouse-rc2 is supposed to go out today
14:06:25 <markwash> if you wanna see any of these or other things land, we'll need to jump on it crazy fast in terms of landing it in master and creating another branch against milestone proposed
14:06:47 <jokke_> markwash: I don't think I will have test out for that oneliner today ... I'm really not familiar enough with the testing system for that and haven't got too much suggestions either how to do that
14:07:34 <markwash> hi jokke_!
14:08:00 <markwash> let me find a link to the test stuart was talking about
14:09:41 <jokke_> markwash: the test Stuart sugggested to remove the config line would not fly ... as per my comment underneath, that test seems to verify that the function works as intended
14:10:11 <markwash> yeah I see that
14:10:21 <jokke_> so my problem is how to test that the config has actually been done at least some point
14:10:24 <markwash> I kinda thought he was meaning to say we could remove calls to that function from test setup
14:10:44 <jokke_> yeah that was his proposal, but looking the nature of that test, it would not make sense
14:11:28 <jokke_> and that seems to be the only place in v2 tests the function is called
14:11:39 <jokke_> I was hoping as well that it would have been that easy
14:12:22 <markwash> jokke_: maybe removing it from https://github.com/openstack/glance/blob/master/glance/tests/functional/db/test_registry.py#L25 would work?
14:12:52 <markwash> I think if we don't explicitly call that function from our functional testing of the registry db driver would be sufficient to prevent regression
14:14:34 <jokke_> markwash: that's true ... if you are happy about where the configure is called now, it would be called on the db.registry.api import
14:15:15 <jokke_> which would make that api.configure_registry_client() obsolete
14:15:38 <markwash> oh sorry, was there an issue with where the configure is called in your patch?
14:15:41 <jokke_> and ensure that the function call stays in the import level
14:15:55 <markwash> I know its a little tricky to do that stuff at import time
14:16:12 <jokke_> markwash: I did it first in the images ImageController so it would have been consistent with v1
14:16:37 <markwash> jokke_: consistency with v1 doesn't make a ton of sense here, honestly
14:16:41 <markwash> v2 doesn't always use the registry
14:16:47 <markwash> so it shouldn't hardwire registry setup
14:16:58 <markwash> v1 *does* always use the registry
14:17:04 <jokke_> fair enough and that was what we discussed with Stuart as well
14:17:09 <markwash> right
14:17:18 <markwash> so doing the configure at import time is a bit weird, however
14:17:28 <markwash> I think we don't normally import glance.db.registry.api
14:17:48 <markwash> instead we do a normal import for glance.db and then later call get_api()
14:17:50 <jokke_> it is, _but_ it seems to work and does not look horrible ;)
14:18:10 <markwash> which will do importutils.import_module('glance.db.registry.api')
14:18:21 <markwash> so the import is actually like a class constructor in the final analysis
14:18:24 <jokke_> ok
14:18:29 <markwash> still gross, I'll grant
14:18:30 <jokke_> yeah
14:18:33 <markwash> but best not to rock the boat :-)
14:18:39 <markwash> for rc2 at least! :-)
14:19:16 <jokke_> I agree ... so if we want to throw that in as it is, fine or I have couple of hours left to try to do some test hacking hoping nothing will break badly :P
14:19:42 <markwash> okay cool, thanks for that discussion jokke_  I'm glad we can probably get that item unblocked
14:19:56 <jokke_> thanks
14:20:16 <markwash> this bug is sort of on the bubble https://bugs.launchpad.net/glance/+bug/1285617
14:21:06 <markwash> it has a few options for fixes
14:21:08 <markwash> https://review.openstack.org/#/c/77379/
14:21:17 <markwash> and https://review.openstack.org/#/c/83327/
14:21:31 <markwash> zhiyan: that latter one is yours I believe
14:21:41 <zhiyan> markwash: yes
14:22:10 <ameade> I vote for the quick fix atm
14:22:14 <zhiyan> actually i think maybe we can land https://review.openstack.org/#/c/83327/ first, then ask #77379 use "tox -egenconf" to update template conf file
14:23:09 <zhiyan> ok, if so, as ameade said, i prefer owner of #77379 just cherry-pick 383327 down to local dev env, and do  "tox -egenconf" to update template conf file
14:23:13 <markwash> zhiyan: are you thinking of this for milestone-proposed icehouse-rc2 ? or for juno in general?
14:23:15 <zhiyan> IMHO
14:23:50 <zhiyan> i think rc2 needs 77379 probably
14:24:14 <markwash> okay I think I see
14:25:01 <markwash> are there other big differences that would surface if we used config sample generation, other than this db section?
14:25:29 <zhiyan> i believe there has some
14:25:46 <markwash> zhiyan: also, is there a particular reason you want to make sure the config is the same before and after switching to config sample generation? or just general consistency?
14:26:29 <zhiyan> the complete config items https://review.openstack.org/#/c/83327/1/etc/glance/glance.conf.sample
14:27:50 <ameade> i like 77379 as-is for now i think
14:27:51 <zhiyan> just for general consistency
14:27:58 <markwash> one thing I'm a tiny bit worried about is if there are options where our sample description is a lot better than our description in the option definition
14:28:18 <zhiyan> ameade: me too. but i think it miss some db related options
14:28:19 <markwash> in juno we should have some time and exposure to have someone audit the sample config for sanity and follow up with description changes in the code
14:28:48 <arnaud> +1
14:29:07 <zhiyan> markwash: ok, i'm fine to make 77379 go for rc.
14:30:20 <markwash> okay let's do another quick pass through that patch and see if there are changes that need to be made before it can land
14:30:20 <zhiyan> and i think probably it need to add missing options
14:30:51 <markwash> zhiyan: +1
14:31:15 <zhiyan> so i still think my latest comments in 77379 is useful
14:31:25 <markwash> looking
14:32:02 <markwash> I see, similar to what you said here
14:32:07 <zhiyan> markwash: and btw, in J, I'd like to figure out a way to allow Glance to support automatic generation of separated sample configuration files.
14:32:25 <markwash> except just paste the relevant sections from tox -e genconfig, rather than the whole config
14:32:33 <arnaud> zhiyan: that would be nice yes
14:32:44 <zhiyan> maybe need to leverage https://bugs.launchpad.net/oslo/+bug/1300546
14:33:04 <markwash> okay I'll reinforce your comment in that patch
14:33:30 <markwash> but I"m worried that shi han zhang won't have time to get a fix uploaded today
14:33:55 <zhiyan> arnaud: yes, manual sync-up is boing and error-prone
14:34:03 <arnaud> yep...
14:34:22 <zhiyan> markwash: ok, i do it.
14:35:17 <markwash> zhiyan fixes all the bugs
14:35:19 <zhiyan> markwash: (after the handling for bug 1298698)
14:35:35 <markwash> if you want to share the burden I'm sure someone whose day is closer to beginning could give it a look ;-)
14:36:33 <markwash> zhiyan: but okay. . I'll be back online in about 2 hours and I'll look for and blindly approve changes
14:36:34 <zhiyan> :)
14:37:00 <markwash> anything else anyone is worried about for icehouse rc2?
14:38:01 <markwash> yay crickets!
14:38:09 <markwash> #topic open discussion
14:38:21 <markwash> so today is my "oh crap gotta catch the bus" day
14:38:25 <markwash> so I need to head out in about 10 minutes
14:38:58 <jokke_> looking good ... I think for Juno we need to get the testing bit more clear for v2 sqlalchemy/registry
14:39:36 <markwash> ameade and I had a good chat yesterday about figuring out some other solutions to the use case of rapidly populating a volume with glance image data
14:40:06 <markwash> ameade: does it still seem like image locations / fs store / image location metadata may do the trick?
14:40:11 <jokke_> I'm also bit concerned about the performance of the v2 api
14:40:33 <markwash> jokke_ hmm me too, do you have any measurements of it we could look at?
14:41:05 <ameade> markwash: i think so, one limitation though is you can only have one backend with glance, cinder however has a scheduler and things that allow spreading the load
14:41:11 <ameade> not sure how much that matters though
14:41:37 <jokke_> markwash: just quick test on our RnD cluster ... Grizzly v1 image-list on about 53k records took less than two minutes ... havana v2 + couple of backports to get reg working same data the image list took 8+ minutes
14:41:58 <zhiyan> ameade: btw, does the wiki page ready?
14:42:00 <ameade> for example, glance could only use one nfs share, but cinder can have as many as it wants and decides which to use
14:42:27 <markwash> jokke_: hmm that's no good, but i wonder if is any better with glance api talking directly to the db rather than to the v2 registry
14:42:31 <ameade> zhiyan: no
14:43:00 <jokke_> markwash: with sqlalchemy in the data_path it was still around 6min+
14:43:08 <markwash> ameade: there is some logic in the fs driver about sharing the load among different directories on the fs, which I think may be similar
14:43:19 <markwash> jokke_: holy smokes! that's no good
14:43:41 <ameade> markwash: interesting
14:43:59 <markwash> jokke_: any theories / patches yet? :-)
14:44:01 <jokke_> markwash: I know ... just haven't had opportunity to dig into what is causing that yet
14:44:01 <zhiyan> markwash: do you like reopen https://blueprints.launchpad.net/glance/+spec/s3-multi-part-upload bp? i believe https://review.openstack.org/#/c/49316/ change is a worth stuff, and Masashi Ozawa really like to do it. (actually he do it in whole I cycle)
14:45:23 <markwash> zhiyan: sure, I can't find why I marked it obsolete
14:45:45 <markwash> jokke_: well I'll be very eager to hear about anything you find out!
14:45:46 <zhiyan> markwash: thanks
14:45:56 <markwash> I would like to put fixes for that into icehouse as a backport if posible
14:46:07 <jokke_> markwash: will let you know as soon as I do
14:46:11 <markwash> okay, gotta run folks, thanks so much for this very tactical meeting!
14:46:14 <markwash> #endmeeting