20:00:50 <robcresswell> #startmeeting horizon
20:00:50 <openstack> Meeting started Wed Sep 21 20:00:50 2016 UTC and is due to finish in 60 minutes.  The chair is robcresswell. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:00:52 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:00:54 <openstack> The meeting name has been set to 'horizon'
20:01:05 <david-lyle> o/
20:01:12 <r1chardj0n3s> o/
20:01:17 <robcresswell> o/
20:01:26 <robcresswell> hey 3 ptls in a row
20:01:37 <r1chardj0n3s> well, that's the usual suspects ;-)
20:01:38 <tsufiev> o/ ;)
20:02:05 <robcresswell> Well thats everyone then :p
20:02:17 <robcresswell> #topic Notices
20:02:29 <robcresswell> Few things to talk through
20:02:37 <robcresswell> Agenda is here:
20:02:40 <robcresswell> #link https://wiki.openstack.org/wiki/Meetings/Horizon
20:03:00 <robcresswell> First off, we've tagged RC1 and master is now open for Ocata development
20:03:05 <robcresswell> \o/
20:03:11 <r1chardj0n3s> \o/
20:03:32 <robcresswell> Ive added newton-rc-potential and newton-backport-potential tags to Launchpad
20:03:36 <tsufiev> Some people are still grumpy at us for glanceV2 timing
20:03:39 <robcresswell> So please tag bugs appropriately
20:03:49 <robcresswell> tsufiev: Are "some people" sahara?
20:03:58 <robcresswell> Or, are there others too?
20:04:00 <tsufiev> Heh, and murano
20:04:32 <r1chardj0n3s> interesting. do you have email threads to point to, 'cos I think  Imissed them
20:04:36 <tsufiev> They implemented a horrible hack setting default glance version to 1 to make their CI pass
20:04:45 <robcresswell> Unfortunately, the choice was made to to go for V2 so that glance can deprecate properly
20:04:59 <robcresswell> Yeah, there are services still relying on v1
20:05:05 * tsufiev looking for a horrible hack link
20:05:30 <robcresswell> It is unfortunate that we caused problems late in the day, but thats just how things are. I think everything has been fixed now in time for release.
20:05:58 <tsufiev> That's it https://bugs.launchpad.net/murano/+bug/1624423
20:06:00 <openstack> Launchpad bug 1624423 in Murano ocata "[dashboard] Murano should show error if glance v2 is used" [High,In progress] - Assigned to zhurong (zhu-rong)
20:06:21 <tsufiev> I guess they were just too busy/desperate to complain in ML
20:07:01 <tsufiev> but yes, eventually they went through it
20:07:04 <r1chardj0n3s> (that images upload change in glance v2 was something I always meant to look into to fathom why they made the change)
20:07:05 <kzaitsev_mb> about the bug — we in murano relied on glance v1 to download images that are required for specific murano apps (when importing murano apps)
20:07:25 <r1chardj0n3s> kzaitsev_mb: s/download/upload yes?
20:07:26 <robcresswell> kzaitsev_mb: Sahara had a similar issue. They relied on copy-from iirc.
20:07:42 <robcresswell> r1chardj0n3s: No, they mean download from a remote URL, right?
20:07:49 <kzaitsev_mb> so if glance v1 is unavilable — we would just put up an error and say where to download the image by hand
20:07:59 <r1chardj0n3s> robcresswell: they can download from a remote URL now, they can't upload a file blob
20:08:11 <kzaitsev_mb> yep, well we downloaded/uploaded it to glance from a different source
20:08:16 <r1chardj0n3s> oh, maybe I'm confused. let's not get stuck on this
20:08:27 <robcresswell> r1chardj0n3s: I thought it was the other way round. You cant copy from a remote URL any more, but you can still upload from local.
20:08:40 <r1chardj0n3s> robcresswell: other way around, I thought
20:08:52 <kzaitsev_mb> well downloading a 500+mb image in dashboard and uploading it to glance all in a http server's process somewhere seems like a bad idea to me
20:09:06 <kzaitsev_mb> so we would just say that we can't and direct users to the url
20:09:14 <kzaitsev_mb> so they can upload it to glance by hand
20:09:33 <tsufiev> kzaitsev_mb: as i said in private, setting glance version to 1 was really an awful idea, since it disabled metadata features
20:09:34 <kzaitsev_mb> at least that's what I plan to have in newton release
20:09:45 <robcresswell> kzaitsev_mb: Thats how upload works, straight to Glance
20:09:54 <robcresswell> in the CORS implementation, iirc.
20:10:00 <kzaitsev_mb> tsufiev: as I pointed to you — there is a commit that's going to revert that =)
20:10:08 <tsufiev> So maybe it's partially our fault - i mean the lack of docs on fallback scenarios for that case
20:10:15 <tsufiev> Okay
20:10:23 <tsufiev> Just sharing my thoughts
20:10:50 <robcresswell> tsufiev: Can you clear up mine and r1chardj0n3s debate? What is it that V2 doesnt support? local upload? or download from url?
20:11:16 <robcresswell> bpokorny_: o/ if you're around, chime in :p
20:11:21 <tsufiev> robcresswell: upload by url i think
20:11:22 <r1chardj0n3s> robcresswell: I'm just firing up the interface now to clear it up ;-)
20:11:49 <robcresswell> upload by url
20:11:51 <robcresswell> lol
20:11:55 <r1chardj0n3s> lol I broke my config "The current Horizon settings indicate no valid image creation methods are available. Providing an image location and/or uploading from the local file system must be allowed to support image creation."
20:12:02 <robcresswell> \o/
20:12:22 <tsufiev> robcresswell: not my invention ;)
20:12:25 <robcresswell> My understanding was that in V2 you can no longer pass a URL for Glance to copy the image from.
20:12:50 <tsufiev> Yep, that's what i meant by 'upload via url'
20:12:54 <bpokorny_> Hi, sorry, I got pulled away. Reading..
20:13:02 <robcresswell> yes. right. thanks tsufiev
20:13:10 <kzaitsev_mb> robcresswell: but if the image is on remote http — that's still a bad idea to download it from remote to some local storage and then to glance, isn't it?
20:13:15 <robcresswell> But you can still upload from your local system fine
20:13:17 <kzaitsev_mb> or even directly to glance
20:13:28 <kzaitsev_mb> it's still a long-running process that can easily fail
20:13:43 <robcresswell> kzaitsev_mb: Its a security concern, not the time length
20:13:43 <kzaitsev_mb> sorry if I missed part of the conversation
20:13:56 <robcresswell> kzaitsev_mb: Image upload/download will always be long-running, that is unavoidable.
20:14:15 <tsufiev> kzaitsev_mb: as we told, there is nothing that horizon can do about that
20:14:18 <kzaitsev_mb> robcresswell: yep, the question is who should do the heavy-lifting
20:14:22 <robcresswell> kzaitsev_mb: The disabling of copy-from I believe is a security concern,  I believe
20:14:24 <tsufiev> We cannot reinvent glance
20:14:44 <kzaitsev_mb> tsufiev: not blaming horizon on glance v2 design decisions =)
20:14:54 <kzaitsev_mb> sorry if I sounded like that =)
20:14:55 <robcresswell> kzaitsev_mb: Well, the only difference really is that you must have the image locally first via download, then upload to glance. Its time consuming.
20:15:06 <robcresswell> No, I'm just being direct for clarity :)
20:15:20 <robcresswell> Otherwise we will keep going in circles.
20:15:34 <kzaitsev_mb> true )
20:15:50 <tsufiev> kzaitsev_mb: does glare support this feature?
20:16:03 <robcresswell> SO, yes, in summary, we broke things, it would have been nice not to, but I still consider V2 compatibility more important than the risk of breakages.
20:16:49 <robcresswell> Next time we will try to get api features before FF, but if things slide, then there is little to be changed for critical work.
20:16:50 <kzaitsev_mb> tsufiev: idk =)
20:17:23 <robcresswell> We have a few open RC2 bugs:
20:17:24 <r1chardj0n3s> yeah, as you say robcresswell, it's a timing thing. should have been in before FF, but unfortunately wasn't
20:17:26 <robcresswell> #link https://launchpad.net/horizon/+milestone/newton-rc2
20:17:58 <robcresswell> r1chardj0n3s: Yah :( Not ideal, but I don't regret allowing it in, either.
20:18:03 <r1chardj0n3s> yup
20:18:25 <robcresswell> So stable cores, please keep your eyes on stable/newton, and backport theese fixes if appropriate
20:18:41 <tsufiev> There was an increased rate of weirdness in integration tests recently (again!)
20:18:44 <robcresswell> I would like to tag RC2 in the next 48 hours tbh.
20:18:47 <bpokorny_> Chiming in late: There's a way to switch to use v1 if people still need the glance v1 features.
20:19:13 <tsufiev> One related to associate ip test, another to firefox rejecting selenium
20:19:23 <robcresswell> hang on please
20:19:27 <robcresswell> tsufiev ^^
20:19:33 <tsufiev> Ack
20:19:42 <bpokorny_> If people need the Glance v1 behavior still, they can configure Horizon to use v1.
20:20:16 <robcresswell> bpokorny_: Yeah, I think its that the o_d/api/glance.py changed some handling related to metadata
20:20:29 <robcresswell> which broke services using it.
20:20:38 <robcresswell> amongst other small changes.
20:21:22 <bpokorny_> I see. Yeah, handling of metadata had some changes at the API layer.
20:21:54 <robcresswell> Reminder about the summit planning:
20:21:57 <robcresswell> #link https://etherpad.openstack.org/p/horizon-ocata-summit
20:22:09 <robcresswell> Please drop any thoughts or +1s and your names inline
20:22:17 <robcresswell> Its a little rob-centric right now :p
20:22:36 <robcresswell> Not that I don't love the sound of my own voice, clearly.
20:23:01 <r1chardj0n3s> I like that rob will continue running things
20:23:07 <robcresswell> #topic Integration tests
20:23:36 <robcresswell> r1chardj0n3s: I plan on being tipsy the entire summit :p
20:23:47 <r1chardj0n3s> +1
20:23:52 <robcresswell> tsufiev: You wanted to talk about integration test issues? :)
20:23:59 <tsufiev> Well, integration tests could use more than 1 troubleshooter
20:24:05 <robcresswell> Sorry for interrupting earlier, I just wanted to get through notices.
20:24:09 <tsufiev> Sure
20:25:05 <tsufiev> I cannot dedicate all my time to them unfortunately, for example their refactoring to do less redundant things is completely stalled
20:25:21 <robcresswell> Yes :(
20:25:42 <robcresswell> I'd like to discuss testing at length in the summit sessions... its becoming problematic
20:26:00 <tsufiev> Our other guy is fascinated by idea of writing his own testing framework, which i'm a bit skeptical about
20:26:23 <robcresswell> We need to consider how move forward practically. Integration tests are currently relying on 1 or 2 people for all their fixing, which is not good.
20:26:42 <tsufiev> Mostly because tests are not a feature, it's a process of maintenance
20:27:21 <robcresswell> Perhaps we should be much stricter on features, and simply block others so more bugs can be solved. But thats r1chardj0n3s decision :)
20:27:23 <r1chardj0n3s> indeed, though david-lyle and I do have warm bodies we should be able to throw at the problem
20:27:39 <robcresswell> \o/
20:28:00 <tsufiev> r1chardj0n3s: meaning tests improving?
20:28:01 <r1chardj0n3s> robcresswell: well, ultimately, we're at the mercy of corporate goals
20:28:17 <r1chardj0n3s> tsufiev: meaning whatever related to integration tests
20:28:30 <r1chardj0n3s> (well, tests in general)
20:28:32 <tsufiev> \o/
20:28:42 <robcresswell> r1chardj0n3s: yeah :/
20:29:14 <tsufiev> Also at least for 2 cycles we're going to do something about lightweight integration tests
20:29:34 <tsufiev> Aka functional ones, but through browser
20:29:49 <robcresswell> selenium?
20:29:51 <tsufiev> The last sentence is my vision for them
20:29:54 <r1chardj0n3s> yep, that never turned into a plan after the midcycle
20:29:56 <tsufiev> Yes
20:30:07 <r1chardj0n3s> I think a BP is needed, to solidify the vision
20:30:21 <r1chardj0n3s> so we can all agree on that "lightweight" and "functional" mean :-)
20:30:38 * tsufiev makes a mental note
20:30:44 <r1chardj0n3s> and then we can point contributors at that BP
20:31:47 <robcresswell> Got a bug I'd like to discuss, if we're done on testing
20:31:53 <tsufiev> Planning is the hard task. Once you have the plan, it's very hard not to implement it yourself ;)
20:32:09 <tsufiev> I mean, to resist the temptation
20:32:42 * tsufiev finished
20:32:56 <robcresswell> :)
20:33:03 <r1chardj0n3s> robcresswell: go for it, I also have a bug to discuss
20:33:03 <robcresswell> #topic RC2 bugs
20:33:16 <robcresswell> #link https://bugs.launchpad.net/horizon/+bug/1624743
20:33:17 <openstack> Launchpad bug 1624743 in OpenStack Dashboard (Horizon) "Project image table: admin user sees images which are not shared with me" [High,New]
20:33:35 <tsufiev> Who is me ;)?
20:33:51 <robcresswell> This is a sort of odd one. Its a bug because its a change in behaviour
20:34:26 <tsufiev> bpokorny: ^^
20:34:33 <robcresswell> I think its because of how the ng implementation is detecting the current context.
20:34:46 <robcresswell> So it simply checks if they are an admin, but not if they are in the admin panel.
20:35:03 <robcresswell> But thats just a theory
20:35:13 <robcresswell> That sounds to me like the project and admin panel are identical.
20:35:22 <r1chardj0n3s> if you're logged in as admin
20:35:24 <tsufiev> Not the glanceV2 aftermath?
20:35:26 <bpokorny_> I think that bug is only for the django images panel.
20:35:43 <lcastell_> Is it?
20:35:44 <robcresswell> this isnt v2, I think its the new ng panel
20:35:59 <robcresswell> As it says "incompatible with the existing behavior"
20:36:32 <robcresswell> Effectively, if you are logged in as admin, the project panel is treated as the admin panel.
20:36:51 <bpokorny_> Ah, yep, you're right robcresswell.
20:37:02 <bpokorny_> That's for the ng images panel.
20:37:38 <bpokorny_> It's due to how the glance v2 API returns images for admins.
20:38:20 <bpokorny_> With glance v2, there's no way to specify as an admin "I want the list of images for a specific tenant, or that are shared with that tenant"
20:38:24 <bpokorny_> In 1 api call.
20:38:35 <bpokorny_> Glance v1 had a way to do that.
20:38:54 <robcresswell> Oh interesting
20:39:03 <robcresswell> So its more a v2 issue
20:39:05 <bpokorny_> When you get an image list for Glance v2 as an admin, you get all the images in the cloud.
20:39:19 <robcresswell> I wonder how it behaves on v1 in ng images
20:40:07 <bpokorny_> There are filters that can be used in the API call that might be able to replicate what glance v1 used to return.
20:40:41 <bpokorny_> I'll test it out with v1 on ng images to see if it still works the same as before.
20:41:07 <robcresswell> The question I had really, was whether we actually want to, given that you can use the faceted search with it
20:41:50 <robcresswell> if its consistent with glance's behaviour, then I'd expect to stick with that, rather than filtering on the API call to change the data
20:41:51 <r1chardj0n3s> but you can't filter by project in the Project Images list, can you?
20:42:13 <robcresswell> I think you can if you are an admin
20:42:26 <r1chardj0n3s> I'm looking at it right now, and you can't ;-)
20:42:37 <robcresswell> Ah well, that changes things
20:42:52 <r1chardj0n3s> huh, you can't filter by project (or owner) in the admin list either
20:42:57 <r1chardj0n3s> that's suboptimal
20:43:14 <robcresswell> bpokorny_: If you get some time to dig into that, that would be very helpful. Could you leave comments on the bug report?
20:43:31 <tsufiev> But you can by Owned facet, nope?
20:43:33 <bpokorny_> robcresswell: Yep, will do.
20:44:06 <tsufiev> r1chardj0n3s: ^^
20:44:35 <robcresswell> thanks bpokorny_
20:44:38 <r1chardj0n3s> tsufiev: nope http://i.imgur.com/N6e9Ggi.png
20:44:50 <r1chardj0n3s> that's the admin view
20:44:57 <bpokorny_> np
20:45:09 <r1chardj0n3s> projects view is the same
20:45:27 <tsufiev> Maybe we should add the missing facet then...
20:45:30 <robcresswell> I like that security testing image there too
20:45:52 <r1chardj0n3s> robcresswell :-)
20:45:54 <r1chardj0n3s> tsufiev: yeah
20:46:19 <robcresswell> agreed, tsufiev. Lets investigate a bit more and discuss tomorrow, I think. We need more info.
20:46:31 <robcresswell> r1chardj0n3s: Did you want to use the last ten mins or so for the other bug?
20:47:39 <r1chardj0n3s> robcresswell: so I couldn't help noticing that https://bugs.launchpad.net/horizon/+bug/1462922 is still marked Critical, a year and a half on, and we've been migrating away from run_tests anyway...
20:47:40 <openstack> Launchpad bug 1462922 in OpenStack Dashboard (Horizon) "run_tests without venv broken" [Critical,Confirmed]
20:48:12 <tsufiev> Haha, horizon is very robust piece of software
20:48:27 <r1chardj0n3s> totes is
20:48:32 <david-lyle> this has been around for a release or two
20:48:32 <tsufiev> It can live with a critical bug for more than a year
20:48:49 <david-lyle> ah, as you said
20:48:51 <r1chardj0n3s> I think it speaks to how "critical" the bug truly is, and I think it should be downgraded.
20:48:58 <david-lyle> ++
20:49:33 <r1chardj0n3s> (I pinged mrunge/zigo about it yesteday, but got no response)
20:49:42 <robcresswell> Ah yes, the packagers were not fond of these changes
20:49:46 <zigo> o/
20:49:59 <robcresswell> r1chardj0n3s: They are both UTC +2 I *think*
20:50:15 * zigo reads backlog
20:50:53 <r1chardj0n3s> zigo: this is regarding whether https://bugs.launchpad.net/horizon/+bug/1462922 is still Critical, or even a valid bug any longer in light of the move to tox
20:50:54 <openstack> Launchpad bug 1462922 in OpenStack Dashboard (Horizon) "run_tests without venv broken" [Critical,Confirmed]
20:51:15 <zigo> It's valid, but we can always have distro specific bugs.
20:51:18 <zigo> Sorry, patch.
20:51:22 <zigo> Which we do in such a case.
20:51:59 <zigo> It'd be nice to have it fixed for the final release, I'd say.
20:52:15 <zigo> Hum...
20:52:19 <zigo> It was open a long time ago.
20:52:25 <r1chardj0n3s> yes :-)
20:54:07 <zigo> There's still "site-packages" which doesn't even exist in Debian (it's "dist-packages").
20:54:16 <zigo> What is this .js for?
20:55:13 <r1chardj0n3s> right, so on looking into it further, I'm not sure the tox switch alters the basic problem that mrunge has, which is installation over the Internet during test setup, for the js tests
20:55:30 <robcresswell> Yup :/
20:55:45 <robcresswell> Vendor everything? :p
20:55:58 <r1chardj0n3s> when I'm in charge, we're gonna vendor EVERYTHING
20:56:16 <david-lyle> you're fired
20:56:23 <r1chardj0n3s> uhoh
20:56:24 <robcresswell> ...I've made a terrible mistake
20:56:28 <robcresswell> :p
20:56:32 <david-lyle> :D
20:56:45 <jlopezgu> lol
20:56:49 * zigo opens a bug against ftp.debian.org to remove horizon
20:56:56 <zigo> :)
20:56:57 <r1chardj0n3s> ouch
20:57:00 <david-lyle> haha
20:57:18 <david-lyle> seems to be going really well for you r1chardj0n3s :/
20:57:20 <robcresswell> Almost at time
20:57:30 <r1chardj0n3s> \o/
20:57:32 <robcresswell> The undermining has already begun
20:57:41 <robcresswell> Hasn't even started yet
20:57:58 <r1chardj0n3s> so we have a "critical" bug that downstream work around but we have no real solution for
20:58:00 <david-lyle> I think he's tryiing to back out
20:58:00 <r1chardj0n3s> hurm
20:58:17 <david-lyle> there's no escape
20:58:19 <robcresswell> david-lyle: Too late, traps been sprung
20:58:20 <r1chardj0n3s> can't back away any further... unless I move to Tasmania
20:58:24 <r1chardj0n3s> it's nice down there
20:59:29 <robcresswell> Lets defer this. Might be worth talking about at summit as a general to do, regarding dropping run_tests.
20:59:30 <david-lyle> resolved: it's nice in Tasmania
20:59:36 <david-lyle> good meeting
20:59:38 <tsufiev> r1chardj0n3s: what's about syberia ;)?
20:59:44 <robcresswell> david-lyle: That sounds like a #success to me...
21:00:04 <robcresswell> Thanks everyone!
21:00:07 <robcresswell> #endmeeting