19:02:49 <dtroyer> #startmeeting OpenStackClient
19:02:50 <openstack> Meeting started Thu Sep 24 19:02:49 2015 UTC and is due to finish in 60 minutes.  The chair is dtroyer. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:02:52 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
19:02:55 <openstack> The meeting name has been set to 'openstackclient'
19:03:01 <stevemar> i/
19:03:03 <stevemar> o/
19:03:05 <dtroyer> ping: dhellmann, stevemar, briancurtin, terrylhowe, lhcheng, sigmavirus24
19:03:13 <dhellmann> o/
19:03:21 <lhcheng> o/
19:03:24 <dtroyer> I did put together an agenda this morning...
19:03:30 <stevemar> i'm on a call, so i apologize in advance if i don't pay 100% attention
19:03:38 <dtroyer> #link https://wiki.openstack.org/wiki/Meetings/OpenStackClient#24_Sep_2015
19:04:00 <stevemar> dstanek: ^ maybe you wanted to join this time?
19:04:27 <dtroyer> so lets get started
19:04:30 <terrylhowe> o/
19:04:35 <dstanek> howdy
19:04:36 <dtroyer> #topic Image v2
19:05:03 <dtroyer> We got a chunk of Image v2 work in 1.7.0, there's more in the pipeline now,
19:05:08 <stevemar> popular topic these days
19:05:13 <dtroyer> do we have an idea of what is still missing?
19:05:13 <stevemar> mordred: ^
19:05:31 <dhellmann> yeah, I recommend not doing much with image create until after the summit when we can see what the glance team is going to do there
19:06:23 <dtroyer> ok, thats actually mostly done for direct upload, the alternate methods are not there yet.
19:06:33 <mordred> yah. it has PUT
19:06:33 <dtroyer> so that makes sense to wait
19:06:36 <dhellmann> ++
19:06:38 <mordred> does not have tasks
19:06:42 <mordred> but I agree on waiting on that
19:06:58 <dtroyer> mordred: do you know if that list of v2 deployments also supports v1 at all?
19:07:29 <mordred> dtroyer: I don't have a full catalog of v1 support - but I believe many of them did
19:07:36 <dtroyer> I'm thinking we might want to make falling back simpler for that specific case
19:07:38 <mordred> dtroyer: would it be helpful to know that?
19:07:48 <mordred> nod
19:08:04 <mordred> that's what glanceclient is doing in general these days
19:08:06 <mordred> so I agree
19:08:15 <dtroyer> this is one reason I never finished v2 create in the first place… we might just use v1 create for now
19:08:18 <dtroyer> ok, yes
19:08:58 <dtroyer> otherwise, we have a bunch of changes to set in review
19:09:08 <dtroyer> #link https://review.openstack.org/226843
19:09:19 <dtroyer> #link https://review.openstack.org/226749
19:09:30 <dtroyer> #link https://review.openstack.org/227397
19:09:57 <dtroyer> I think bunting is going to add unset to https://review.openstack.org/226843
19:10:20 <dtroyer> I know find_resource() is broken for v2 also, what else?
19:11:55 <dtroyer> nothing?
19:12:38 <dtroyer> ok, so that's a shorter list than I thought.
19:13:12 <dtroyer> I had a question about the 'extra' options in v2 set (—architecture, —os-distro, etc) and think I may have been the one to put those there
19:13:24 <dtroyer> but disagree with my former self if so and want to remove them.  objections?
19:14:57 <dhellmann> I'm not sure what that means?
19:15:57 <dtroyer> there are options new in v2 set (and my version of create) that weren't in v1, I'd like to remove them as they are really just wrappers around --property
19:16:48 <dtroyer> actually, there were in v1, and were added for compat
19:16:57 <dtroyer> duh, I need to read more before I speak
19:17:46 <dhellmann> so we would just want folks to use --property instead of the more specific options?
19:18:05 <dtroyer> long-term, yes.  These were for CLI compatibility
19:18:23 <dhellmann> makes sense
19:18:27 <dtroyer> and I think I re-convinced myself to keep them
19:18:39 <dtroyer> I wish I could blame pain killers for that...
19:18:54 <dtroyer> anything else about Image to talk about?
19:20:10 <dtroyer> #topic SDK integration
19:20:52 <dtroyer> terrylhowe has been working with this on and off for a while now, and has a chain of reviews starting at https://review.openstack.org/#/c/138745/ to add Netowrk support using the SDK
19:21:40 <dtroyer> The recent image fun has re-motivated me to look at getting off glanceclient so I've been looking at that too
19:21:42 <terrylhowe> it is kind of big.  I’d like to get the keystoneauth stuff merged to the sdk before doing a real sdk integration.  I think that work is very close in the sdk
19:22:18 <dtroyer> agreed, but we can start laying the foundation for it and exploring the command-level changes required
19:22:31 <terrylhowe> I’ve just been stuck on the special project team and had no extra time in the last 2 weeks
19:23:02 <dtroyer> I generalize the basic include-sdk-as-dependency bits from terrylhowe's work in https://review.openstack.org/#/c/227037/
19:23:23 <terrylhowe> I think if the sdk supported ksa, we could treat the ksc session just as if itwas a ksa one
19:23:50 <dtroyer> good to hear
19:23:57 <dtroyer> but, here's what I really want to talk about…
19:24:15 <dtroyer> we've said for a while that adding SDK would be 2.0
19:24:29 <dtroyer> and that we don't want to do a 'big switch' to do that
19:25:11 <dtroyer> but I'm a little concerned about needing to make changes that would also require a major version bump and that we'd land at 5.0 sooner than later
19:25:19 <dtroyer> (maybe this is misplaced concern?)
19:25:44 <dtroyer> and the recent ML thread about semver and dependencies made me think about it more.
19:25:47 <dtroyer> thanks dhellmann ;)
19:25:49 <dhellmann> I wouldn't worry too much about that
19:26:50 <dhellmann> I'm not actually certain that adding the sdk or replacing the client libs with the sdk represents an API change in OSC that requires a major bump
19:26:58 <dtroyer> ok.  user expectations for a 2.0 release that has no user-visible changes (other than installing openstacksdk) seems odd
19:27:14 <dtroyer> that's one of what I was thinking about since the thread
19:27:28 <dhellmann> right, as long as the supported API of OSC is not changing it should be fine
19:27:39 <dhellmann> the only time I expect to be able to install something without updating dependencies is for a patch release
19:28:23 <dtroyer> and I think from an OSC user perspective that is also the case
19:28:55 <dtroyer> I think earlier we were concerned about possible behavioural changes with the changing of the API client implementation
19:29:05 <dtroyer> good testing should handle that
19:29:13 * stevemar lurks back in and reads scroll up
19:29:15 <dtroyer> but our testing is spotty
19:31:55 <dtroyer> stevemar: I know you have opinions here, speak up when you've caught up
19:33:05 <stevemar> dtroyer: i know i said recently that including the sdk should be a bump to 2.0, but i guess i don't see an issue with releasing 1.8.0 with sdk included
19:33:41 <dtroyer> ok, I think we're on the same page now
19:34:09 <dtroyer> the next question though, is the SDK team ready to have us include it as a production dependency?
19:34:22 <dtroyer> I've told them before that we wanted to wait until they release 1.0 to do that
19:34:49 <stevemar> they 1.0.0'ed right?
19:34:51 <dtroyer> but I think we can do this incrementally and work in the ksa bits as they release them
19:35:00 <dtroyer> I'm getting 0.6.2 from pypi right now
19:35:23 <stevemar> 0.5.2 ?
19:35:38 <dtroyer> but the stuff terrylhowe did works fine with where it is now
19:35:40 <stevemar> not released since july 30th
19:35:46 <dtroyer> look for openstacksdk
19:35:47 <stevemar> thats a lot of missing
19:35:57 <dtroyer> they appear to have dropped python-
19:36:01 <stevemar> ah okay
19:36:06 <stevemar> smart move ;)
19:36:19 <stevemar> unlike these other chumps we know
19:36:20 <dtroyer> which we're still going to do at 2.0, whenever that happens
19:36:34 <terrylhowe> there won’t be a 1.0 until ksa is done
19:37:00 <terrylhowe> it changes the way the user uses the sdk and makes it a lot easier to integrate
19:37:05 <dtroyer> are you guys OK with OSC starting to require sdk before then?
19:37:20 <dtroyer> I'm not worried about our internal change required there
19:37:38 <terrylhowe> the python-openstacksdk project depends on openstacksdk, but it shouldn’t be used
19:37:43 <dtroyer> and maybe that's our trigger to go to 2.0
19:38:02 <stevemar> dtroyer: i'm okay with that
19:38:20 <stevemar> dtroyer: how many commands do you expect will leverage the sdk?
19:38:32 <terrylhowe> yeh, I think that is fine dtroyer as long as it isn’t used in something key
19:38:35 <dtroyer> eventually all of them?
19:38:49 <stevemar> dtroyer: obvssss
19:38:54 <dtroyer> ok, thanks terrylhowe
19:39:14 <dtroyer> I think though our start will be with Network, Image and Object-Store goven their general state overall now
19:39:27 <stevemar> dtroyer: so i;m looking at this gap, how much exposure will we have between now and when sdk is 1.0?
19:39:46 <dtroyer> exposure to what?
19:40:16 <stevemar> from a requirements POV it's easy, sdk only requires stuff we already use
19:40:54 <stevemar> dtroyer: if we release 1.8.0 with sdk, what commands will use it, only object store? only image? both?
19:41:05 <dtroyer> right.  and I don't expect for the functionality to change with 1.0, just the SDK API
19:41:14 <stevemar> right
19:41:19 <stevemar> which could be churn for us
19:41:25 <dtroyer> whatever is ready when we get there?  terrylhowe has started Network, I've started Image
19:41:25 <stevemar> but won't break anyone
19:41:49 <stevemar> i say lets do it
19:41:51 <terrylhowe> I don’t think anyone uses network much, so that is pretty safe
19:41:53 <dtroyer> my hidden agenda is to remove the glanceclient and neutronclient dependencies
19:41:55 <stevemar> FILGTM!
19:41:58 <dtroyer> and their deps
19:42:00 <stevemar> oh my yes
19:42:27 <stevemar> dtroyer: does the sdk have support for all the image commands we do now?
19:42:35 <stevemar> we dont do any network commands :(
19:42:44 <dtroyer> I don't know, was learning as I went.
19:43:04 <dtroyer> we do network * commands now, terrylhowe already has re-implemented them in SDK
19:43:17 <dtroyer> wokring forward from there gets easier
19:44:23 <terrylhowe> I’m not sure on coverage in the sdk image world stevemar Evertt Toews did most of the recent work there
19:44:48 <dtroyer> does it do both v1 and v2?
19:45:02 <dtroyer> that's what I need to figure out next, how to switch it...
19:45:15 <stevemar> looks like its missing stuff https://github.com/stackforge/python-openstacksdk/tree/master/openstack/image
19:45:36 <terrylhowe> yes dtroyer  and image create and basic stuff I know work, just don’t know all the details on properties etc.
19:45:45 <terrylhowe> but it has v1 and v2
19:45:48 <dtroyer> cool
19:46:00 <stevemar> fairly limited functionality there
19:46:05 <dtroyer> so we should be able to do a v2 image show with a name ;)
19:46:08 <stevemar> but we can do object and network first
19:46:42 <dtroyer> and fwiw, I'm planning to jump in and fill in holes as needed
19:47:05 <dtroyer> ok, 15 min left…movie on?
19:47:08 <terrylhowe> sdk network and object store will do everything the osc needs
19:47:13 <dtroyer> cool
19:47:34 <dtroyer> #topic reviews and bugs
19:47:51 <dtroyer> I've mentioned a couple of reviews already, are there any others that need attention?
19:48:29 <dtroyer> https://review.openstack.org/214264 looks simple and ready
19:48:32 <stevemar> dtroyer: oh coolness here: saharaclient is creating an osc plugin
19:48:59 <dtroyer> ah, that reminds me of another topic for open discussion
19:49:39 <stevemar> https://github.com/openstack/python-saharaclient/blob/master/setup.cfg#L35-L46
19:49:50 <stevemar> that was unexpected and nice to see
19:50:33 <stevemar> looks like designate and ironic are lined up too: https://review.openstack.org/#/c/133676/
19:50:39 <stevemar> https://review.openstack.org/#/c/171672/
19:51:14 <dtroyer> zaqar is working on one, was talking about command formats with them
19:51:19 <stevemar> barbican was +2'ed but it was bit-rotted: https://review.openstack.org/#/c/198732/
19:51:39 <stevemar> oh, i spoke with thingee and hemnalunch today over at -cinder
19:51:43 <dtroyer> and so since stevemar took us down this road, do we list the known plugins anywhere in the docs?
19:52:01 <stevemar> dtroyer: duh http://docs.openstack.org/developer/python-openstackclient/plugins.html
19:52:10 <stevemar> though it needs an update
19:52:14 <terrylhowe> nice stevemar on sahara
19:52:14 <dtroyer> i a word there
19:52:21 <dtroyer> known plugins objects
19:52:33 <terrylhowe> designate merged for v2
19:52:39 <dtroyer> for collision avoidance purposes…
19:52:41 <stevemar> ahhh, the command objects?
19:52:44 <stevemar> no we do not
19:52:58 <stevemar> that could be done...
19:53:31 <dtroyer> I'm thinking we might want to do that, or link to the objects in the plugin docs to make it easier to see what is in general use
19:53:51 <stevemar> i'll think of a way to try and represent that
19:54:29 <stevemar> instead of Command List we can have Plugin Command List ?
19:54:49 <stevemar> and list the known ones there... hopefully autogenerate them
19:54:56 <dtroyer> that may be a lot?  if generated, sure
19:55:10 <dtroyer> the issue was in knowing what object names are already in use in other plugins
19:55:15 <stevemar> right
19:55:20 <stevemar> let me think about it
19:55:27 <dtroyer> ok, thanks
19:55:37 <dtroyer> any other reviews to mention?
19:55:44 <stevemar> dtroyer: back to cinder talk - spoke with thingee and hemnalunch - you can read the -cinder scroll back
19:55:58 <stevemar> they are entertaining the idea of deprecating the CLI for cinder
19:56:22 <dtroyer> cool…  I don't have -cinder in znc, was that today?
19:57:38 <stevemar> dtroyer: yes, today
19:57:41 <stevemar> a few hours ago
19:58:05 <dtroyer> ok, I'll find it
19:58:16 <stevemar> we can chat with them at the summit, but i think they are ready to take the plunge :)
19:58:29 <terrylhowe> awesome
19:58:36 <stevemar> their only concern is microversions
19:58:55 <dtroyer> if they did it similar to nove (i think they did) that's about 2/3 worked out
19:59:09 <stevemar> dtroyer: they haven't even done it yet, i think
19:59:20 <stevemar> just something on the horizon
19:59:45 <stevemar> we are at le end
20:00:05 <dtroyer> ok, even better then, we might be able to streamline them a bit.  once we implement a geneeralized discovery/negotiation we'll know what nova and ironic did wrong ;)
20:00:20 <dtroyer> yup.  thanks everyone!
20:00:32 <dtroyer> #endmeeting