Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2018-02-07
17:11:48 mordred rods_: we just need to land the three patches in front of it
17:12:25 mordred rods_: although I kinda think we should wait for https://review.openstack.org/#/c/540556/ so we can verify we don't break osc ...
17:12:33 mordred rods_: lemme rebase that stack on the jobs
17:14:20 rods_ mordred +2 on https://review.openstack.org/#/c/541802/1
17:32:56 dtroyer mordred: yes, aggregates are an occasional race-like issue
17:52:01 mordred dtroyer: cool. I've rechecked. it should be green this time
18:04:45 johnsom dtroyer Have a minute to give the Octavia team some input on a new OSC command?
18:04:56 dtroyer johnsom: sure
18:05:31 johnsom dtroyer Thank you sir. We have a need to expose some "driver" specific commands targeted to admins.
18:05:36 johnsom dtroyer https://etherpad.openstack.org/p/octavia-drivers-osc
18:06:21 johnsom dtroyer The team likes "openstack octavia amphora failover" where octavia is the project name, amphora is the "load balancing element (service vm currently).
18:07:25 johnsom dtroyer I warned them that we tend to stay away from project code names in the commands. In this case the "octavia amphora ..." is all driver dependent and will return "Not implemented" if the driver isn't loaded into octavia, etc.
18:07:50 johnsom dtroyer We seek you CLI wisdom... grin
18:08:19 dtroyer yes, I really try to discourage using project names in CLIs, the resources should be named descriptively, but generically
18:08:28 mordred johnsom: SO ... I just discovered something that i'm going to bug you about during the rocky cycle
18:08:41 mordred johnsom: octavia does not have a version discovery document
18:09:26 johnsom mordred Ha, ok. Yes, this is on the todo list to make sure we have that figured out. We dump some version info based on the API-wg doc at the time, but need to understand that better for sure.
18:10:16 mordred johnsom: I discovered it because of an sdk patch that's failing functional tests on octavia things (so yay for testing!)
18:10:16 johnsom mordred Can you put a story in for us with a pointer of the "right way", maybe a project example?
18:10:33 mordred johnsom: I will TOTALLY put in a story for you and work to make sure it's all done right
18:10:45 mordred johnsom: (and I can work around it sdk-side for now)
18:10:59 johnsom mordred Thank you. I can probably hammer it out pretty quickly with the right guidence
18:11:17 mordred yah- it should be super simple
18:11:44 dtroyer johnsom: so "amphora" is the name of the type of service vm?
18:11:48 johnsom dtroyer So, my other worrry is: "openstack octavia", "openstack f5", "openstack radware", etc.
18:12:16 johnsom dtroyer https://docs.openstack.org/octavia/latest/reference/glossary.html
18:12:34 dtroyer at some point we break this overloading of the "openstack" command. maybe this is approaching one of those places? IDK
18:12:45 johnsom dtroyer it's an element of the octavia driver that does the load balancing. Could be service vm, container, baremetal, etc.
18:13:50 johnsom dtroyer Yeah, it's a bit interesting. The need is there for admin commands for the plugin drivers.
18:14:11 dtroyer my first (sime-uninformed) reaction is to suggest just using "amphora" as the top-level resource. It is closer to a project name in style than a resource description, but might make snese in this case
18:14:23 dtroyer s/sime/semi/
18:14:31 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Shift config exceptions to openstack.exceptions https://review.openstack.org/539396
18:14:31 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Cleanup the BaseProxy docs a bit https://review.openstack.org/539270
18:14:32 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Rename BaseProxy to Proxy https://review.openstack.org/539271
18:14:32 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Generate proxy methods from resource objects https://review.openstack.org/540169
18:14:33 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Calculate name in CloudRegion https://review.openstack.org/541059
18:14:33 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Use get_session_client in Connection https://review.openstack.org/541060
18:14:47 johnsom dtroyer This is why I was thinking making a namespace under the overall service, "openstack loadbalancer" might be the best idea. Then the plugins can just expand that as needed.
18:14:49 mordred rods_: ^^ rebased that stack on the OSC tips enablement patch
18:14:57 dtroyer johnsom: the big win with that is "amphora" is likely to be unique
18:15:24 dtroyer I really try to not use the word "namespace" as it is subtly different from the way OSC names resources
18:15:25 johnsom dtroyer The downside is only the "octavia" driver has "amphora". F5 doesn't for example.
18:15:35 rods_ mordred cool, I'll give a look at all of them in a few minutes
18:15:50 mordred dtroyer: yah - I think in this case it'sa driver-specific function for admins who have deployed that driver, so having them do 'openstack amphora blah' doesn't seem super bad to me
18:16:06 dtroyer johnsom: so maybe that is why we use it bare? it names the resource directly. Is there a parallel for f5 or other drivers?
18:16:56 johnsom dtroyer Speaking out of line since I don't work for F5, maybe "bigIP"????
18:17:07 johnsom dtroyer or "LTM"
18:18:03 dtroyer the point being that other lb drivers may have a similar concept, if they do, is using amphora for the octavia driver setting a good or bad precedent?
18:18:04 johnsom dtroyer So we would be ok with third party OSC plugins that add top level commands?
18:19:08 dtroyer johnsom: mostly, yes. we do ask for plugins to play nice and register resource names with OSC, but can not enforce that
18:19:59 dtroyer and when we get into areas where there are multiple implementations of a thing, we either need to hide that from the user, or if the user is expected to know and care about the difference, make it logical for them
18:20:50 johnsom dtroyer Ok, so "openstack octavia amphora failover" = "openstack <driver> <element><command>"
18:20:52 dtroyer is the use of LB drivers "one-per-cloud"? or could a deployment have multiple? it does sound like the user of this CLI will need to kow in any case
18:21:13 johnsom dtroyer "openstack amphora failover" = "openstack <element> <action>"
18:21:17 dtroyer think this way: openstack <resource-name> <action>
18:21:32 johnsom dtroyer most deployments use two or more load balancing drivers
18:21:36 dtroyer your suggestion is resource-name == "octavia amphora", mine is == "amphora"
18:23:37 dtroyer johnsom: ok, for the other drivers, how will users think about driver-specific stuff? I'm looking for the name of the driver-specific resources, they might begin with something parallel to amphora for octavia. this is a space I don't have much experience with
18:23:49 johnsom dtroyer Ok, I think we have good guidance to bring to our meeting today. The meeting is at 20:00 UTC today in #openstack-lbaas if you want to join the conversation.
18:24:37 johnsom dtroyer For non-admin commands, they have little knowledge of the driver. They simply do "openstack loadbalancer create --provider <driver name>"
18:25:38 dtroyer johnsom: I'll be on the phone most of the afternoon… ok, so using a name that identifies the driver as part of the admin driver-specific resource nmae is appropriate I think
18:28:23 dtroyer johnsom: I left a couple of comments in that etherpad
18:28:30 johnsom dtroyer Thank you
19:19:51 openstackgerrit Monty Taylor proposed openstack/keystoneauth master: WIP Handle unversioned endpoints with no discovery document https://review.openstack.org/541844
19:23:36 mordred dtroyer: https://review.openstack.org/#/c/540554/ <-- that's green now! dhellmann if you're around too
19:24:05 mordred dtroyer, dhellmann: this is the fix for the gate job issue that let us release sdk with changes that broke osc
19:25:01 mordred dtroyer, dhellmann: and https://review.openstack.org/#/c/540556 is the corresponding SDK change that's going through recheck right now
19:25:06 dhellmann dtroyer beat me to it
19:25:10 dtroyer mordred: +W and thank you!
19:25:50 mordred woot!
19:30:56 openstackgerrit Monty Taylor proposed openstack/python-openstackclient master: Make Profile fallback go bye-bye https://review.openstack.org/540541
19:31:25 mordred dtroyer: there's the followup for post-queens to remove the profile fallback code
20:01:49 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Calculate name in CloudRegion https://review.openstack.org/541059
20:05:47 openstackgerrit Monty Taylor proposed openstack/python-openstacksdk master: Use get_session_client in Connection https://review.openstack.org/541060
20:25:48 openstackgerrit Merged openstack-infra/shade master: Use openstacksdk for most transitive depends https://review.openstack.org/541796
20:26:17 openstackgerrit Monty Taylor proposed openstack/os-service-types master: Update from service-types-authority https://review.openstack.org/541864
20:51:02 EmilienM hi
20:54:19 mordred EmilienM: please don't tell me that the latest openstacksdk release has broken tripleo in new and creative ways
20:54:38 mordred EmilienM: I think the release team might kill me if I have to release another point release during freeze :)
20:54:58 mordred EmilienM: (I'm kidding, obviously tell us if something broke)
20:55:55 mordred dtroyer: if you have a sec, wanna pull the trigger on sdk gating on osc tips? https://review.openstack.org/#/c/540556/
20:56:37 stevebaker mordred: we're just talking optimal default behaviour when someone ctrl-c's a command
20:56:56 mordred oh good. that sounds like an osc question :)
20:57:39 dtroyer mordred: +W
20:57:56 mordred dtroyer: thanks! I'll be much happier with that landed
20:58:27 openstackgerrit Merged openstack/python-openstackclient master: Make osc-functional-devstack-tips actually use tips https://review.openstack.org/540554
20:59:15 mordred \o/
21:01:33 mordred dtroyer: tell me about it
21:17:16 openstackgerrit Merged openstack/cliff master: Updated from global requirements https://review.openstack.org/534911
21:27:14 EmilienM mordred: which makes me think I always complain!
21:28:16 mordred EmilienM: :) nah - I just had a fun week last week of discovering sdk release breaks due to missing tests
21:28:46 EmilienM but if you ask, I can certainly find something that broke us recently :P
21:31:43 mordred EmilienM: hehe. I'm sure :)
21:34:53 openstackgerrit Merged openstack/python-openstacksdk master: fix misspelling of 'volume' https://review.openstack.org/541095
22:02:43 rods_ mordred probably not important but you have small typo here https://review.openstack.org/#/c/539271/6
22:25:34 openstackgerrit Merged openstack/python-openstacksdk master: Gate on osc-functional-tips https://review.openstack.org/540556
#openstack-sdks - 2018-02-08
00:07:03 openstackgerrit OpenStack Proposal Bot proposed openstack-infra/shade master: Updated from global requirements https://review.openstack.org/541943
00:36:46 openstackgerrit zhangzhaoshan proposed openstack/keystoneauth master: Fix a spelling error https://review.openstack.org/541949
00:38:34 openstackgerrit Merged openstack/python-openstackclient master: Zuul: Remove project name https://review.openstack.org/541004
06:59:12 openstackgerrit Merged openstack/osc-lib master: Zuul: Remove project name https://review.openstack.org/540080

Earlier   Later