Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2017-11-30
17:16:09 cdent yeah, good, thanks edleafe
17:17:09 cdent away
17:17:17 cdent as in, thar she blows
17:17:41 cdent dtantsur|afk: you should see your +2 (if I’ve done it right)
17:23:27 openstackgerrit Merged openstack-infra/shade master: Avoid tox_install.sh for constraints support https://review.openstack.org/508061
17:30:08 openstackgerrit Merged openstack/python-openstacksdk master: Remove use of tox-siblings role https://review.openstack.org/523993
17:45:55 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: DNM: Test -tips variations https://review.openstack.org/523938
17:47:48 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: DNM: Test -tips variations https://review.openstack.org/523938
17:48:52 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: DNM: Test -tips variations https://review.openstack.org/523938
18:05:35 thingee with the merging of shade and python-openstacksdk, are extensions still going to be supported https://review.openstack.org/#/c/523834/ - shade normalized things is why I'm asking
18:29:22 mordred thingee: we support some openstack extensions in shade already, so I thnk so - the main thing froma shade pov is in whether or not the extension in question can be exposed/presented in a way that makes sense or degrades in a reasonable way
18:30:34 mordred thingee: like, if an extension adds a field to a resource, then we can always add that field to our returned data model and just fill in None if the cloud doens't have the extension ... or if the same info is given by different extensions slightly differently and we can rationalize it - then we can try to do that
18:31:06 mordred thingee: but some extensions are things that don't make much sense in a general context - like something that exposed vendor-specific attributes - and those are ones we'd never add to the data model
18:32:45 mordred thingee: long-winded way of saying "it depends/case-by-case" for the shade layer ... the sdk object layer is somewhat similar to shade's data model, but it's also closer to exposing the details of the cloud in question - so I could see an argument for some variation to be exposed, along with discovery flags, at thatlayer
18:34:06 mordred thingee: in the case of https://review.openstack.org/#/c/523834/ - that particular case seems reasonable for the responses - tags are a neutron feature that a deployer may or may not have enabled, so defaulting them to [] seems maybe reasonable?
18:34:16 thingee yeah got it, thanks
18:34:28 thingee and yes I agree it's reasonable
18:34:33 mordred otoh - I'm not sure what the experience is going to be for a user that tries to add a tag to a resource on a cloud that doesn't support it - and I'm not sure that patch addresses that
18:35:02 mordred (it's also possible that's already handled in the existing code, I need to go look)
18:54:39 thingee mordred: pretty sure that's just going give a 404 http://git.openstack.org/cgit/openstack/python-openstacksdk/tree/openstack/network/v2/tag.py#n27
18:55:50 thingee just being a keystone session request.
18:59:09 thingee mordred: I'll check around for other instances, but this should probably have the _has_neutron_extension('tags') check and raise like the other place shade does it
18:59:11 thingee yuk
20:38:34 thingee installing the latest python-openstacksdk, if I call openstack.OpenStackCloud() I get http://paste.openstack.org/show/627928/
20:39:25 thingee ah I guess get_one_cloud() doesn't just get one without something passed in kwargs
20:43:19 mordred thingee: yah - you need to give it at leat a cloud=
20:44:05 mordred dtroyer: I've got 3 different keystoneauth patches I need to rebase/fix and get landed - would you like me to take a stab at taking the timing wrapper you have in osc and putting it directly on ksa.session?
20:44:14 mordred dtroyer: since I'll be in there anyway?
20:45:48 openstackgerrit Andreas Jaeger proposed openstack/os-api-ref master: void tox_install.sh for constraints support https://review.openstack.org/524344
21:00:13 openstackgerrit Julia Kreger proposed openstack-infra/shade master: De-clientify baremetal create/delete https://review.openstack.org/500119
21:02:03 openstackgerrit Julia Kreger proposed openstack-infra/shade master: De-client-ify baremetal machine port list https://review.openstack.org/500120
21:02:04 openstackgerrit Julia Kreger proposed openstack-infra/shade master: De-client-ify machine patch operations https://review.openstack.org/500121
21:04:21 dtroyer mordred: sure, I don't mind. I think we'll want to land a compat check into osc-lib for that, especially if you don't put it into Session
21:04:49 dtroyer the version in osc itself appears to be unused and I never got around to removing it
21:05:22 thingee mordred: so there are two interfaces to python-openstacksdk? I can openstack.OpenStackCloud.update_qos_policy() or use openstack.network.v2.qos_policy
21:06:15 thingee mordred: Is there anywhere written of what the plan is there?
21:06:48 openstackgerrit Andreas Jaeger proposed openstack/os-api-ref master: Avoid tox_install.sh for constraints support https://review.openstack.org/524344
21:09:15 openstackgerrit Monty Taylor proposed openstack/os-api-ref master: void tox_install.sh for constraints support https://review.openstack.org/524344
21:09:16 openstackgerrit Monty Taylor proposed openstack/os-api-ref master: Add os-api-ref-nova-src job https://review.openstack.org/524353
21:10:46 mordred thingee: yes! there are two interfaces, and there is a plan written down ... (and also there is a plan to make the two interfaces be less completely different)
21:11:49 mordred thingee: oh - I lied - there is a document, but it does not contain words about this - lemme fix
21:12:00 briancurtin thingee: hello
21:15:47 openstackgerrit Andreas Jaeger proposed openstack/os-api-ref master: Add os-api-ref-nova-src job https://review.openstack.org/524353
21:16:22 thingee briancurtin: hi
21:18:21 openstackgerrit Andreas Jaeger proposed openstack/os-api-ref master: Add os-api-ref-nova-src job https://review.openstack.org/524353
21:18:21 openstackgerrit Andreas Jaeger proposed openstack/os-api-ref master: Avoid tox_install.sh for constraints support https://review.openstack.org/524344
21:18:59 briancurtin thingee: i ran by Sizzle Pie the other day. just thought you should know
21:19:38 thingee briancurtin: and you didn't go in?
21:19:48 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Update the shade-merge document https://review.openstack.org/524391
21:20:33 mordred thingee: ^^ ok - I updated that doc a bit- it should probably be fundamentally rewritten and moved somewhere more discoverable
21:20:45 briancurtin thingee: unfortunately not. a mid-run pizza didn’t sound great at the time, but actually sounds great now
21:22:09 mordred briancurtin: the concept of a mid-run pizza might almost be enough to convince me to run
21:22:39 thingee mordred: thanks
21:22:40 mordred although I'd probably just end up skipping the first part of the run, going straight for pizza, then skipping the second part because I'm full
21:23:03 mordred thingee: oh - it occurs to me there's another thing to put in there ...
21:26:23 adriant mordred: I honestly don't mind working on bugs in the SDK, so potentially I'll get distracted and submit some patches for the Swift issues I've found.
21:26:52 mordred adriant: woot!
21:27:01 adriant mordred: and yes a function of flag for "a connection with just these args thanks" would be amazing.
21:27:10 adriant or* flag
21:28:04 openstackgerrit Julia Kreger proposed openstack-infra/shade master: Remove version arg from updated ironic calls https://review.openstack.org/524393
21:28:11 adriant mordred: I submitted one patch already for swift related nonsense: https://review.openstack.org/#/c/524066/
21:30:19 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Update the shade-merge document https://review.openstack.org/524391
21:35:04 openstackgerrit Krzysztof Klimonda proposed openstack-infra/shade master: Prioritize external networks for assigning FIP to instance https://review.openstack.org/524396
21:40:27 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: Remove a bunch of things we promised to remove in 2H 2017 https://review.openstack.org/524399
21:45:52 openstackgerrit Monty Taylor proposed openstack/os-api-ref master: Add os-api-ref-nova-src job https://review.openstack.org/524353
21:45:52 openstackgerrit Monty Taylor proposed openstack/os-api-ref master: Avoid tox_install.sh for constraints support https://review.openstack.org/524344
22:00:14 TheJulia mordred: so hit a minor issue. python-ironicclient has retry logic baked in for handling nodes locked by the conductor. Any thoughts? I'm thinking the least invasive would be just to double check for the locks before operations that could be blocked by them and wait for the lock to clear, which would clear until ironic likely needs to go re-check the power state.
22:04:38 mordred TheJulia: ah- interesting ... and yes, that seems like a good plan ... altough I suppose there's a race condition there and that something could still lock between the check and the operation - so we still probably want to handle geting a node-locked error
22:05:36 TheJulia Yeah...
22:09:25 adriant mordred: omfg. my notfound 404 issue. it's simple. no one stuck `exceptions.raise_from_response(resp)` in the _head function
22:09:35 adriant although I'm curious if that breaks anything else
22:09:47 adriant will submit a patch and we can test/review
22:10:09 mordred adriant: woot!
22:12:05 TheJulia mordred: I guess it only really matters on actions that are actually trying to make a change that requires a lock in the first place
22:16:25 dtroyer that sounds like it involved neutron…. where I think the only atomic operation is "halt"
22:19:22 TheJulia I guess there could be code someplace that requires the task object to be re-read, which could make things... fun if someone deleted ports while a task was running in ironic
22:33:08 openstackgerrit Merged openstack-infra/shade master: De-clientify baremetal create/delete https://review.openstack.org/500119
22:33:09 openstackgerrit Merged openstack-infra/shade master: De-client-ify baremetal machine port list https://review.openstack.org/500120
22:33:10 openstackgerrit Merged openstack-infra/shade master: De-client-ify machine patch operations https://review.openstack.org/500121
22:45:30 openstackgerrit Lance Bragstad proposed openstack/python-openstackclient master: Add system role functionality https://review.openstack.org/524416
22:51:28 openstackgerrit Adrian Turjak proposed openstack/python-openstacksdk master: Swift 404s to raise NotFound correctly https://review.openstack.org/524418
22:52:04 adriant mordred: this fixes my current swift issues with 404: https://review.openstack.org/#/c/524418/
22:52:16 adriant potentially may break other APIs that use HEAD differently, but I doubt it
22:52:40 adriant I think someone just forgot to add that function all to resource.head
22:52:51 adriant function call*
22:57:04 openstackgerrit Merged openstack-infra/shade master: Remove version arg from updated ironic calls https://review.openstack.org/524393
22:57:29 openstackgerrit Brian Haley proposed openstack/python-openstackclient master: Display IPv6 addresses better https://review.openstack.org/524420
23:36:04 thingee adriant: curious why the key would be in the values?
23:36:29 thingee for https://review.openstack.org/#/c/524066/1
23:37:11 thingee adriant: nevermind, found the storyboard ticket with the info
23:38:30 mordred Shrews, briancurtin, dtroyer, SamYaple, thingee, TheJulia: I have discovered that pypy unittests are basically unpossible in the gate atm (the version of cryptography in global-requirements is not compatible with the version of pypy in xenial)
23:39:04 TheJulia ugh
23:39:14 mordred there is a ppa that has newer pypy, but that'll take a decent amount of consideration to deal with broadly - so I leaning toward just disabling the pypy tox env for now
23:39:34 mordred until we can get a job set up that will *actually* test it
23:40:08 mordred (it's been green so far because of a different bug that has caused it to not actually use pypy - fixing that causes the job to start (correctly) failing)
23:42:22 dtroyer mordred: ++ on disabling for now… I'd say make it experimental but we know it's never going to actually work without tasks elsewhere first
23:42:37 mordred yah, that was my thinking
23:42:54 dtroyer mordred: fwiw, I'm testing a fix for OSC's -tips job… I finally duplicated it here
23:42:57 mordred HOWEVER - I just had another idea I'm going to try out REAL quick ... we could tell it to run on fedora26 nodes instead of xenial nodes ...
23:43:02 mordred dtroyer: woot!

Earlier   Later