Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2017-11-29
22:05:05 mordred adriant: well, this isn't nearly as useful a link as I'd hoped, but ...
22:05:15 mordred http://git.openstack.org/cgit/openstack/python-openstackclient/tree/openstackclient/network/client.py#n39
22:05:38 adriant mordred: that still uses profile :P
22:05:41 mordred adriant: yah- the answer for you is gonna be to make a CloudConfig object basically just like you've been making a profile
22:06:02 mordred adriant: lemme make a quick patch to something else that shows the ACTUAL answer
22:06:02 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: Release note cleanup https://review.openstack.org/524004
22:06:07 mordred adriant: and also a doc patch :)
22:06:53 adriant mordred: k, I had a feeling i'd end up having to make a config object :P
22:07:03 mordred adriant: out of curiosity - do you want your use to explicitly *not* support clouds.yaml - or more you don't have any clouds.yaml and don't care
22:07:28 mordred (example patch is different depending on which of those is true - I should probably make both examples)
22:08:02 adriant mordred: the thing I'm working on in question is a django site/service that has it's own config file. So clouds.yaml doesn't make sense.
22:08:12 mordred adriant: nod. cool - example coming
22:08:15 adriant same again with a different API service I've got
22:08:33 adriant then the rest of my tools are built on top of the openstackclient, so they natively just use clouds.yaml
22:09:19 adriant I find that if I'm writing any scripts for openstack, I just make them an openstackclient plugin and forget all the tiresome stuff and focus on functionality
22:09:53 adriant mordred: and thanks :)
22:31:31 mordred adriant: https://review.openstack.org/524011 Update openstacksdk construction to be forward compatible
22:32:05 mordred adriant: I'm going to extract that into a proper doc - but that should have all the pieces you need (and more)
22:34:02 openstackgerrit Merged openstack-infra/shade master: De-client-ify baremetal node_set_provision_state https://review.openstack.org/499773
22:38:53 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: Release note cleanup https://review.openstack.org/524004
22:42:44 adriant mordred: ty, and love the comment
22:46:17 mordred adriant: :)
22:48:35 adriant mordred: I have to admin, that's much nicer than the profile stuff. The profile stuff was my main reason for not liking the SDK since it just seemed overcomplicated
22:48:40 adriant admit*
22:49:29 adriant like sure, you could have swift in one region, nova in another, and cinder in a third, all in the same connection object. the question though is why!?
23:05:30 adriant mordred: and the thing you were talking about at the summit was "get_session_client" right? So in the context of openstackSDK you can just do raw API calls if the code doesn't support the feature you.
23:06:36 adriant mordred: but that appears to be on the config class, rather than the connection/service class. Which means I need to keep the config class around if I need to call APIs directly, or is there a way I can get at that on the connection/service class?
23:07:49 adriant I was imagining it as "resp = conn.object_store._client.get("/thing")
23:09:03 adriant I'm sure the adapter is in there somewhere, and i can probably find it by trawling the code some more, just was curious if there was a nice way to do access it.
23:25:25 adriant mordred: also, a passthrough or shortcut for "OpenStackConfig(load_yaml_config=False).get_one_cloud" might not be a bad idea. Since we shouldn't assume clouds.yaml is the default. It's only the default from a cli/tools perspective, and we really should make use the SDK in services and such much easier. Or at least initialising it.
23:31:18 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Add some docs about not using Profile https://review.openstack.org/524021
23:32:00 mordred adriant: sorry - was writing that ^^
23:32:35 mordred adriant: SO - in the master branch currently you don't need do do a get_session_client at all - once you have a connectoin, each service's Proxy object is actually also an Adapter
23:32:46 mordred adriant: so you can do conn.compute.get('/foo')
23:35:11 mordred adriant: also, fwiw, the service proxy objects all all by service_type as per the service type registry, and there's also one for any known alias ... although still todo is to make sure old aliases like volumev2 actually construct appropriately versioned adapters
23:35:42 mordred adriant: AND - yes, definitely shortcuts such as what you say - there are several littered about, but we need to rationalize it all since the codebase merges
23:36:47 mordred adriant: so, like, in master, you can do "import openstack ; openstack.connect(**parameters_you_would_have_passed_to_get_one_cloud)"
23:37:45 SamYaple TheJulia: w00t! congrats on getting whiskey chicken. seems like a fair trade for that work (now you just have to hold monty to it)
23:39:23 mordred adriant: maybe we should make 2 factory functions - one for no-clouds.yaml and one for yes-clouds.yaml ... cause for utility scripts I gotta tell you I would never ever use not clouds.yaml and the extra typing to turn on loading the files would drive me batty
23:39:49 mordred adriant: *waves hands* ... fundamentally, totally agree with the things you're saying, we can work out what the shortcut things do pretty easily so we're all happy :)
#openstack-sdks - 2017-11-30
00:03:04 openstackgerrit Monty Taylor proposed openstack/cliff master: Remove tox_install.sh and just pass -c in tox https://review.openstack.org/523984
00:03:04 openstackgerrit Monty Taylor proposed openstack/cliff master: Move doc requirements to doc/requirements.txt https://review.openstack.org/522369
00:03:05 openstackgerrit Monty Taylor proposed openstack/cliff master: DNM testing docs and releasenotes jobs https://review.openstack.org/522370
00:40:21 adriant mordred: fantastic! :)
00:41:14 adriant mordred: and yeah, I don't think making clouds.yaml harder to use is right, just that, we should make it equally easy to use either. :)
00:42:24 adriant mordred: because for the longest time I always felt that the openstackSDK was tailored more towards clouds.yaml stuff rather than : "I'm an API service which needs to talk to another service." which is more my primary use case.
00:43:01 adriant while the python-<service>clients always felt far better to just throw a session at, and then use.
01:20:09 TheJulia SamYaple: Wait... is there a real whiskey chicken thing?!?
01:33:35 openstackgerrit Merged openstack-infra/shade master: de-client-ify baremetal get_machine https://review.openstack.org/499774
01:44:01 adriant mordred: I'm going to be opening a lot of issues for the sdk over the next while as I find things that don't quite work as expected, or are missing. Should I be doing that here: https://storyboard.openstack.org/#!/project/760 ?
01:44:30 adriant I mean, now that shade == openstacksdk that makes sense, but just curious.
01:44:49 adriant the launchpad page seem ancient so I assume bugs aren't tracked there.
01:48:54 openstackgerrit Merged openstack-infra/shade master: Catch attrbute error for other APIs https://review.openstack.org/499775
02:25:40 openstackgerrit Merged openstack/python-openstackclient master: Release note cleanup https://review.openstack.org/524004
03:33:01 openstackgerrit Hengqing Hu proposed openstack/python-openstacksdk master: Add pools attribute to load balancer heath monitor https://review.openstack.org/522962
04:57:14 SamYaple TheJulia: you can put whiskey in anything if you try hard enough
04:57:47 TheJulia SamYaple: Indeed, although I think most people will prefer a glass of some sort.
05:22:00 openstackgerrit Adrian Turjak proposed openstack/python-openstacksdk master: Stop osSDK mangling Swift metadata keys https://review.openstack.org/524066
05:25:53 adriant mordred: potentially awkward, but even with yaml=False, calling 'get_one_cloud' without a cloud name, it will look in local OS_* envvars. Which might be good to also disable/avoid as well as yaml checking.
05:26:49 adriant not that it really is likely to crop up much, but still a little weird because of how I was testing stuff. I didn't realise I had set any vars, and it seemed to pick them up and yell at me for providing domain name for token auth.
05:34:32 adriant hmmmm... anyone here who has played with swift support in the openstackSDK?
05:34:54 adriant I'm trying to do HEAD on a container to see if it exists before I try creating it.
05:35:28 adriant but well... `x = conn.object_store.get_container_metadata('not_a_real_container')` returns a valid resource object
05:35:40 adriant it's not a 'real' container
05:35:58 adriant but it has some attrs
05:36:19 adriant I can't for the life of me find a value to say "this is not a real container"
05:36:30 adriant which is broken
05:36:33 adriant like, pretty broken
05:36:55 adriant swiftclient sucks, but at least it throws a 404 type error
05:38:42 adriant `x = conn.compute.get_server("not a real server")` does throw an error
05:40:33 adriant `x = conn.object_store.get_object_metadata("not a real container", "fake object")` also returns a not none value :/
05:41:32 adriant while `x = conn.object_store.get_object("not a real container", "fake object")` returns: '<html><h1>Not Found</h1><p>The resource could not be found.</p></html>'
05:41:40 adriant wow, object_store in the SDK is broken
05:42:35 adriant I think this is a good time to call it quits for today...
08:35:01 ldl can I through the server id to get the server image snapshots?
10:43:46 openstackgerrit Édouard Thuleau proposed openstack/python-openstacksdk master: Set empty Tag list if Network Tag API extension not supported https://review.openstack.org/523834
14:31:07 mordred adriant: morning! so ...
14:31:35 mordred adriant: good point re: yaml=False - I think what we want is a "make me one of these and only use the kwargs I'm passing, not other stuff
14:37:03 mordred adriant: for the other thing, good to know - I know the shade layer for swift objects is good and also support for the 'use head to check if it exists' use case
14:38:00 mordred adriant: as we work through rationalizing the shade and sdk layers, we'll make sure to fix that SDK brokenness you just reported (thanks for that btw)
14:38:56 mordred adriant: hopefully we'll at least have a $something you can use related to the new-world-order soonish
14:42:08 mordred dhellmann: morning mr. "I know literally every python module that exists" ... I think I remember there being some sort of module that would let you execute example code from docs as part of tests ... do you know if that's a thing or if I'm just dreaming it?
14:44:41 cmurphy mordred: i've heard of that, i think it's this https://docs.python.org/2/library/doctest.html
14:45:55 mordred cmurphy: ah! yes - that's what I was thinking of, thanks!
14:46:20 mordred cmurphy: I wrote some example docs yesterday, but it turns out the examples I wrote don't work ... made me think testing examples in docs would be a good idea
14:47:13 mordred although that module isn't *quite* what I was thinking - maybe I could either add support for the usage pattern I was considering, or use it as inspiration to write a thing
14:48:27 mordred basically very similar to https://docs.python.org/2/library/doctest.html#simple-usage-checking-examples-in-a-text-file - but finding .. code-block: python sections in rst files, and not requiring them to be interactive sessions
14:49:06 mordred it's also possible that in this particular case it needs to be specific to shade/sdk codebase, as I probably want a requests_mock thing set up ...
14:53:24 mordred rcarrillocruz, Shrews, jlk: https://github.com/ansible/ansible/pull/20974 has been updated (the rebase on that SUCKED)
15:02:07 Shrews mordred: seems to have failed tests still
15:24:26 mordred yah - I'll fix that next
15:24:38 mordred Shrews: sake of argument, what do you think about this: https://gist.github.com/ce8b1c90cf972bfc104e3eb38ae74762 as an alternative approach?
15:34:59 Shrews mordred: hmm, i sort of like that
15:35:51 mordred Shrews: cool, me too ... jlk, rcarrillocruz, SamYaple: ^^ what about y'all ?
15:36:16 rcarrillocruz think latter is cleaner
15:52:45 thingee mordred: where are these roles defined again? https://review.openstack.org/#/c/523993
15:53:23 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: DNM: Test -tips variations https://review.openstack.org/523938
15:54:10 cdent api-sig meeting in #openstack-meeting-3 in a few minutes
15:59:12 mordred thingee: in that case, tox-siblings is/was defined in openstack-infra/zuul-jobs
15:59:21 thingee was?

Earlier   Later