Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2017-04-12
12:30:33 jamielennox when nova_client creates the adapter it should pass client_name and client_version
12:31:01 jamielennox when heat (or whoever) creates the session it passes to novaclient it passes app_name and app_version
12:31:21 jamielennox so when it talked to multiple clients you would get the correct heat/X novaclient/Y keystoneauth/Z ...
12:31:55 mordred cool - did we do those?
12:32:04 jamielennox in a couple of places...
12:32:07 jamielennox but largely no
12:32:24 mordred oh - but session and adapter have them - so I can use them it seems perhaps
12:32:49 jamielennox there was then a session.additional_user_agent list that was for things like os-c-c that fit somewhere in the middle
12:33:13 mordred jamielennox: awesome. I will rework this patch to make use of that system
12:33:45 mordred so for example, occ willbe additional_user_agent - shade will be client and nodepool will be app
12:34:11 mordred (at least, once all the python-*clients are removed)
12:34:12 jamielennox mordred: yea, that was the intent but it was one of those things that the person who was pushing a lot to have it in lost interest once it was
12:34:18 jamielennox yep
12:35:03 jamielennox i feel like additional on session is going to be weird so let me know if you want that tweaked
12:35:43 mordred kk.
12:35:51 openstackgerrit Merged openstack/python-openstacksdk master: Updated from global requirements https://review.openstack.org/456004
12:50:23 openstackgerrit Merged openstack/python-openstacksdk master: Add 'project_id' to Server query parameters https://review.openstack.org/454241
13:45:53 openstackgerrit yuval shalev proposed openstack/python-openstacksdk master: Add support for volume attachments in compute v2 https://review.openstack.org/455335
13:56:03 openstackgerrit yuval shalev proposed openstack/python-openstacksdk master: Add support for volume attachments in compute v2 https://review.openstack.org/455335
13:57:46 openstackgerrit yuval shalev proposed openstack/python-openstacksdk master: Add support for volume attachments in compute v2 https://review.openstack.org/455335
14:19:57 openstackgerrit yuval shalev proposed openstack/python-openstacksdk master: Add support for volume attachments in compute v2 https://review.openstack.org/455335
14:46:49 openstackgerrit Monty Taylor proposed openstack/os-client-config master: Add ability to pass in user_agent https://review.openstack.org/452550
14:47:04 mordred jamielennox: ^^ ok. I think that does things right using the app_name/client_name bits
14:57:59 mordred jamielennox: also - client_name,client_version are not getting added to the user-agent in keystoneauth
14:58:25 mordred so setting them on the adapter is possible, but it doesn't change anyting (setting additional_user_agent in the Session _does_ work)
15:01:57 mordred jamielennox: oh - I see - it happens at request time - so I'm breaking that in shade - I can fix that
15:33:18 openstackgerrit Monty Taylor proposed openstack/keystoneauth master: Shift additional_user_agent in the stack https://review.openstack.org/456259
15:33:19 openstackgerrit Monty Taylor proposed openstack/keystoneauth master: Allow setting client name on the Session https://review.openstack.org/456260
16:00:39 reedip_ o/
21:21:22 mordred edleafe, cdent: https://review.openstack.org/456394 I just made a neutron API spec that I'd love API-WG thoughts on
21:22:25 cdent mordred: ah, cool, I look forward to reading that tomorrow. i have zero brains left for today
21:23:07 mordred cdent: braaaaaaains
21:23:17 mordred cdent: I'm fairly certain it's all wrong
21:23:25 cdent best way to get started
21:23:27 mordred cdent: I mean, except for the part where it's an essential feature
21:23:37 mordred that I am certain I am correct about
21:23:40 mordred but everythign else is wrong
21:39:33 edleafe mordred: from an API perspective it seems fine. Maybe split the key type from the "key" key
21:39:45 edleafe heh - a key named 'key'
21:40:17 edleafe so the 'keys' list would contain dicts of "comment", "type", and "key"
21:40:36 edleafe just to make it easier to filter client-side
21:41:10 mordred edleafe: that'sa great idea
23:24:21 jamielennox mordred: with the additional shift why do you want it there?
23:24:37 jamielennox mordred: not that anyone else uses this, but my thought would be you would want to see like
23:24:52 jamielennox nova/A os-c-c/B glanceclient/C
23:25:03 jamielennox because os-c-c is like creating the glanceclient
23:28:48 jamielennox mordred: yea, i'd like to have a bit more discussion on what you're trying to do there
#openstack-sdks - 2017-04-13
00:00:36 mordred jamielennox: ah- I see what you mean - so I'll try to make this extra complex ...
00:01:40 mordred but first - I mostly have stopped caring about python-*client - but let's keep them for now because they make the discussion more interesting (harder)
00:02:57 mordred so - for the shade case, we have nodepool uses shade which uses os-client-config which uses keystoneauth which uses requests (at least that's how I think of it in my head)
00:04:11 mordred then for novaclient, it would be nodepool -> shade -> os-client-config -> novaclient -> keystoneauth ... or is it?
00:04:34 mordred maybe for the novaclient/glanceclient path the occ part is not meaningful
00:04:46 jamielennox sounds right
00:04:52 mordred in fact, I wonder if occ is important in the user agent at all?
00:05:14 jamielennox i think this is what dtroyer came to as well, telling people osc-lib version wasn't that important - but it's there if you want it
00:05:15 mordred it's setting up objects, but it's not really involved (atm) in making actual calls
00:05:20 mordred nod
00:09:05 mordred jamielennox: so maybe focusing on interior ordering isn't important - nodepool "the app" being first in the list seems important, and ksa/requests being at the end seems right
00:09:21 mordred other than that, I can make all sorts of arguments about why one sequence makes more sense than the other
00:09:28 mordred but I'm not sure any of them actually make sense :)
00:09:56 jamielennox mordred: yea, it took us a while to get to the layout we have because of these sorts of things - you can get really picky about it
00:10:10 mordred ++
00:10:12 jamielennox realistically the best thing to do would be to have the python clients actually insert their name into the sequence
00:10:43 mordred yah. although hopefully 'soon' I won't be using them for anything anymore and that can be someone else's problem
00:12:17 mordred jamielennox: for the other patch - the main thing was that without adding client_name to session and only on adater, we lose it when ksa makes discovery calls on behalf of the user
00:12:38 mordred so I figured just setting it on session would be eaiser than plumbing client_name to wherever it needed to be for discovery calls
00:12:46 mordred but again, it's possible I'm making it too difficult :)
00:12:52 jamielennox oh, because discovery is internal to ksa?
00:13:30 jamielennox i mean we can plumb it through - but maybe it doesn't matter
00:13:36 mordred yah - and ksa doens't use an adapter for it
00:13:42 jamielennox things done internal to ksa don't care what invoked the request
00:13:57 jamielennox i mean it would be useful to know what client started the call
00:14:42 mordred right - but when I think of user-agents I think of ops looking at logs to figure out why a particuar traffic pattern came in - and if it was shade that asked ksa for something, I feel like I'd be missing part of the story?
00:14:42 jamielennox mordred: i don't mean to argue this too much, but it's two new params in an already complicated list for fairly small benefit
00:14:51 mordred _but_.... I don't have an actual use case
00:15:16 mordred jamielennox: totally understand - I appreciate you curating the library well :)
00:15:40 jamielennox mordred: is shade creating the session?
00:16:09 jamielennox oh right, this is why shade is out of order with the python-*client in additional
00:16:21 mordred yah
00:16:50 mordred actually - shade does not create the session - occ creates the session for it
00:18:51 mordred in general, it 'should' be occ creating session, adapater and/or python-client - but shade directly instantiates Adapter because it wraps it in a wrapper class
00:19:44 mordred and I gave up on giving occ a "set_adapter_factory" feature because that was just too much :)
00:23:40 jamielennox so i mean shade could add itself to additional at that point
00:25:55 mordred jamielennox: hrm. maybe that's not a bad idea
04:08:25 openstackgerrit Rui Chen proposed openstack/python-openstackclient master: Make block-device-mapping more stable and clear https://review.openstack.org/438141
06:19:00 openstackgerrit Rui Chen proposed openstack/cliff master: Add smart help formatter for command parser https://review.openstack.org/454613
07:39:34 openstackgerrit Reedip proposed openstack/osc-lib master: Add MultiKeyValueCommaAction to osc-lib https://review.openstack.org/425515
07:39:46 reedip RuiChen : Done ^^
11:21:21 openstackgerrit Reedip proposed openstack/python-openstackclient master: Add extra dhcp option to 'port create/set/unset' https://review.openstack.org/356263
12:07:19 openstackgerrit Merged openstack/python-openstacksdk master: Add support for volume attachments in compute v2 https://review.openstack.org/455335
12:28:05 openstackgerrit Brian Curtin proposed openstack/python-openstacksdk master: Add is_profile_only to Cluster resource https://review.openstack.org/455211
13:15:21 elmiko edleafe: heads up, good chance i will be a few minutes late to the meeting today
13:15:50 edleafe elmiko: ok, we'll gossip about you until then
13:16:53 elmiko haha, i would expect nothing less ;)
13:43:16 openstackgerrit Brian Curtin proposed openstack/python-openstacksdk master: Add docs for volume_attachment compute methods https://review.openstack.org/456620
14:27:58 openstackgerrit Merged openstack/python-openstacksdk master: Add docs for volume_attachment compute methods https://review.openstack.org/456620
14:49:43 openstackgerrit Brian Curtin proposed openstack/python-openstacksdk master: Add is_profile_only to Cluster resource https://review.openstack.org/455211
14:50:27 openstackgerrit Brian Curtin proposed openstack/python-openstacksdk master: Enable warnings_as_errors in doc enforcer https://review.openstack.org/453676
15:19:32 openstackgerrit Merged openstack/python-openstacksdk master: Enable warnings_as_errors in doc enforcer https://review.openstack.org/453676
15:39:44 openstackgerrit Monty Taylor proposed openstack/os-client-config master: Add ability to pass in user_agent https://review.openstack.org/452550

Earlier   Later