14:00:01 <rosmaita> #startmeeting glance 14:00:01 <rosmaita> #topic roll call 14:00:02 <openstack> Meeting started Thu Apr 27 14:00:01 2017 UTC and is due to finish in 60 minutes. The chair is rosmaita. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:03 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:05 <openstack> The meeting name has been set to 'glance' 14:00:13 <rosmaita> o/ 14:00:31 <jokke_> o/ 14:00:58 <rosmaita> hi erno 14:01:14 <rosmaita> let's wait two more minutes 14:02:03 <jokke_> yeah 14:04:12 <rosmaita> ok, i guess we can get started 14:04:36 <rosmaita> topic #updates - agenda archive 14:04:46 <rosmaita> oops 14:04:53 <rosmaita> #topic updates - agenda archive 14:05:05 <jokke_> I guess we might be quick today 14:05:11 <rosmaita> yeah 14:05:25 <rosmaita> ok, i moved each year of agenda onto its own etherpad 14:05:44 <rosmaita> the were like 50K lines in the current etherpad, made it slow to load 14:05:44 <slaweq_> hello 14:05:52 <rosmaita> slaweq_: welcome 14:06:02 <rosmaita> #topic updates - new operator survey 14:06:13 <jokke_> aha! ... good call ... I'm not even sure how worthy it was to save ;) 14:06:21 <rosmaita> we have a new survey going to get info about glance cache usage 14:06:32 <rosmaita> #link http://lists.openstack.org/pipermail/openstack-operators/2017-April/013236.html 14:06:57 <rosmaita> that's what i sent to the ops list, if any of you also operate a cloud, feel free to take the survey 14:07:04 <rosmaita> (link is in that email) 14:07:06 <nikhil> .\ 14:07:13 <rosmaita> hi nikhil 14:07:28 <jokke_> any responses so far? 14:07:48 <rosmaita> haven't looked, been kind of distracted lately :) 14:08:32 <rosmaita> 7 responses! 14:09:14 <jokke_> \\o \o/ o// o/7 14:09:16 <rosmaita> i'll send a reminder out on monday, i think the survey closes on Thursday 14:09:22 <jokke_> sounds good 14:09:30 <rosmaita> that's a good response, i think our best so far has been 9 14:09:56 <rosmaita> #topic updates - wsgi goal 14:09:59 <nikhil> Maybe the word "all" helped 14:10:06 <rosmaita> :) 14:10:30 <rosmaita> ok, at the midcycle we said that the wsgi goal would ahve to slip 14:10:44 <rosmaita> but in the meantime, mtreinish has volunteered to take it on! 14:10:51 <rosmaita> so it may happen after all 14:11:03 <rosmaita> he's got 2 patches up 14:11:09 <jokke_> yeah, I haven't even looked why the gate fails on it 'though 14:11:22 <rosmaita> me neither 14:11:56 <rosmaita> anywya, just wanted to publicly say thanks to matt and cue us to keep an eye on his patches 14:12:12 <rosmaita> not sure if the gate prob is related to the next topic 14:12:28 <rosmaita> but before i move on, anyone else have an update to share? 14:12:37 <jokke_> quick one 14:12:53 <rosmaita> the floor is yours 14:13:06 <jokke_> the glance_store backport before EOLing does not look good 14:13:27 <jokke_> the change does not even pass Ocata gate due some cinder errors 14:13:38 <rosmaita> arrrrrgh 14:13:43 <jokke_> so I haven't been able to W+ it in to newton 14:14:08 <rosmaita> will tony kill us if we ask to wait to eol until after the summit? 14:14:41 <jokke_> I don't know and I'm not sure if it is really worth of it 14:15:21 <jokke_> as I haven't had opportunity to chase it either 14:16:15 <rosmaita> ok, i'll take a look after this meeting ends early, it seems like bad bug, but i need to look at teh bug again 14:16:34 <jokke_> yeah, lets sync up after and see what we do about it 14:16:40 <rosmaita> ok cool 14:16:44 <rosmaita> moving on 14:16:50 <rosmaita> #topic breaking change to deal with 14:17:01 <rosmaita> olso.config 4.0.0 released on Monday 14:17:16 <rosmaita> #link http://lists.openstack.org/pipermail/release-announce/2017-April/001370.html 14:17:37 <rosmaita> impact on us is that it's going to break a location_strategy test 14:17:50 <rosmaita> gcb submitted a patch 14:17:53 <rosmaita> #link https://review.openstack.org/#/c/457050/ 14:18:14 <rosmaita> but, it fixes it by removing the test 14:18:34 <rosmaita> the problem is, we're using the 'choices' feature of oslo.config to make sure that an operator doesn't misspell the location_strategy they want to configure. We also have code that checks for the presence of the configured location_strategy module. That code is tested by overriding the oslo.cofig 'choices' (using an option that's basically been removed in 4.0.0) so that the configured value is the name of a non-existent strategy. So we nee 14:18:43 <jokke_> I'd say we rather block the version 14:19:04 <jokke_> aha 14:19:15 <jokke_> so it's just breaking the test, not the actual code 14:19:22 <rosmaita> exactly 14:19:56 <rosmaita> yeah, in that series of patches there were 2 ways to fix 14:20:09 <jokke_> I still don't like remiving the test as solution. But otoh do we have anyone interested/able/willing to fix it? 14:20:14 <rosmaita> first one (i -1'd) was to remove 'choices' from the config 14:20:21 <rosmaita> current one is to remove the test 14:20:40 <jokke_> and does the test actually make sense? 14:20:52 <rosmaita> both good questions 14:21:03 <jokke_> we have quite a few tests lingering around that does not 14:21:20 <jokke_> I'll have a look if I have any input on that 14:21:21 <rosmaita> well, it's possible that someone could remove the module 14:21:27 <mtreinish> jokke_: sry to backtrack it's failing because of grenade on the sql image cache driver 14:21:27 <rosmaita> but, not likely 14:21:43 <rosmaita> mtreinish: oh, that's why you have that 2nd patch 14:21:58 <mtreinish> I used an external lock to synchronize between multiple worker processes (which uwsgi or apache) because they were racing on the db init 14:22:13 <mtreinish> but it added a mandatory config option for the lock path 14:22:14 <jokke_> mtreinish: ok, thanks! And thanks for stepping up and helping with that community goal. 14:22:18 <rosmaita> mtreinish: by the way, thanks for pitching in on this! 14:22:38 <mtreinish> sure, np 14:23:55 <mtreinish> the devstack patch: https://review.openstack.org/#/c/459451/ is more interesting 14:24:14 <mtreinish> it's stuck right now on image uploads because the wsgi protocol doesn't support chunked transfer encoding 14:24:21 <mtreinish> I need to figure out a workaround for that 14:24:54 <rosmaita> so you know, we ran up against that recently 14:25:11 <rosmaita> it was something dharinic was working on 14:25:49 <rosmaita> but i don't remember the patch, will have to look 14:26:16 <rosmaita> #action rosmaita - find wsgi chunked transfer encoding workaround 14:26:42 <rosmaita> #action rosmaita jokke_ - discuss glance_store EOL situation after meeting 14:26:45 <mtreinish> rosmaita: ok cool, thanks. 14:28:01 <rosmaita> ok, back to the location_strategy test 14:28:21 <rosmaita> i think it was really there for if/when we allowd operator configurable location_strategies 14:28:28 <rosmaita> i mean, custom strategies 14:28:36 <rosmaita> which so far we have not really supported 14:28:51 <rosmaita> so ATM, it's not really an essential test 14:29:27 <rosmaita> because if an operator removes parts of glance, they should expect stuff not to work right 14:29:46 <rosmaita> so i'm kind of inclined to accept gcb's current patch 14:29:52 <rosmaita> any thoughts? 14:32:53 <jokke_> might be 14:33:03 <jokke_> I have a look on that patch and comment on it 14:33:29 <rosmaita> ok, thanks, i will add a comment with what i said above 14:33:34 <rosmaita> and we can think about it 14:34:03 <rosmaita> mtreinish: take a look at this: https://review.openstack.org/#/c/423366/10/glance/common/wsgi.py 14:34:50 <mtreinish> rosmaita: ah ok, thanks. 14:35:03 <rosmaita> workaround for webob 1.7 not supporting chunked transfer encoding 14:36:41 <mtreinish> rosmaita: the issue here is that the request is being responded to outside of glance by uwsgi/apache and it never even makes it to glance code 14:39:04 <jokke_> I have a feeling that this was the blocker last time as well when we tried to get glance-api working as wsgi plugin 14:40:57 <rosmaita> yuck 14:41:31 <rosmaita> jokke_: i thihk you are correct 14:41:51 <rosmaita> i remember there was "one little thing" that was a blocker 14:42:02 <rosmaita> and i think this is it 14:42:21 <rosmaita> i guess no one else in openstack has this problem? 14:42:32 <rosmaita> trying to think of what other data apis there are 14:42:52 <mtreinish> I imagine swift will have similar problem 14:43:13 <mtreinish> and there is a note in the keystone apache docs about not using mod_wsgi if you use chunked requests 14:43:25 <mtreinish> but the keystone docs say mod_proxy-uwsgi works 14:43:33 <mtreinish> so I feel like I'm missing something 14:43:35 <jokke_> hmm-m 14:43:48 <mtreinish> in my patch 14:44:41 <mtreinish> uwsgi says it should work too: http://uwsgi-docs.readthedocs.io/en/latest/Chunked.html 14:44:47 <mtreinish> so I'm probably configuring something wrong 14:46:39 <jokke_> ok, sorry but did we have something else still in agenda? We're running short in time. If not I'm happy to continue this discussion or if there is lets move this to #openstack-glance after the meeting 14:47:20 <rosmaita> that's all ... we have 3 issues to discuss: 14:47:28 <rosmaita> 1 eol glance_store mitaka 14:47:40 <rosmaita> 2 gcb's patch 14:47:49 <rosmaita> 3 mtreinish 's wsgi situation 14:49:15 <jokke_> cool 14:50:51 <rosmaita> ok, in reverse order, i don't have anything useful to say about 3 ATM 14:51:48 <rosmaita> anyone else about #3? 14:52:08 <jokke_> not at the moment, I'm not that familiar with wsgi mods 14:52:22 <rosmaita> ok, sorry mtreinish 14:52:27 <jokke_> I just remember some people banging their heads against the wall with this before 14:52:50 <mtreinish> no worries, I'll figure something out eventually :) 14:53:12 <rosmaita> for #2, i am going to +2 the patch with my reasons as given above, and jokke_ and nikhil you can decide what you think 14:54:35 <rosmaita> as far as #1 goes, let's adjourn this meeting early and discuss it in #openstack-glance 14:54:51 <rosmaita> that way we won't get interrupted by having to end the meeting on time 14:55:07 <jokke_> ++ 14:55:08 <rosmaita> but first i should ask: 14:55:13 <rosmaita> #topic open discussion 14:55:25 <rosmaita> anyone got anything? 14:55:41 <jokke_> just want to flag that the IIR patches are still in review Q 14:55:52 <jokke_> And we're closing Pike-2 14:56:00 * rosmaita hangs his head in shame 14:56:34 <rosmaita> which isn't very helpful 14:57:01 <rosmaita> i'm anticipating a bit of a disruption, i have to turn in all my equipment tomorrow 14:57:08 <rosmaita> but will try to look over the weekend 14:57:18 <rosmaita> ok, let's re-convene in #openstack-glance 14:57:22 <rosmaita> thanks everyone 14:57:24 <rosmaita> #endmeeting