| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-02-13 | |||
| 20:59:52 | mriedem | sdague: but we didn't specify a default before when we didn't even have this | |
| 20:59:54 | mriedem | right | |
| 21:00:24 | sdague | mriedem: we're going to need to be more dynamic on the ks_filter then, right? | |
| 21:00:35 | mriedem | sdague: passing interface=None is OK | |
| 21:01:07 | mriedem | i think, checking ksa again | |
| 21:01:26 | mriedem | this https://github.com/openstack/keystoneauth/blob/master/keystoneauth1/identity/base.py#L209 | |
| 21:02:16 | sdague | mriedem: yeh, that's some interesting selection logic | |
| 21:02:19 | sdague | but I think you are right | |
| 21:02:31 | cfriesen | where does the hex hash (or whatever it is) come from in the filenames in releasenotes/notes/ ? | |
| 21:02:40 | mriedem | cfriesen: reno | |
| 21:02:48 | mriedem | http://docs.openstack.org/developer/reno/usage.html | |
| 21:03:12 | cfriesen | mriedem: thx | |
| 21:03:34 | sdague | mriedem: ok, so you know what that means actually | |
| 21:04:04 | sdague | if osad only declared as single placement api service, and it was called internal, all this code would have still worked | |
| 21:04:36 | mriedem | osad? | |
| 21:04:44 | mriedem | openstack ansible? | |
| 21:05:22 | sdague | yeh | |
| 21:05:27 | mriedem | if you only have a single placement endpoint and it's interface is not "public" then you have to set this | |
| 21:05:36 | mriedem | and the "public" thing is only because of the default in ksa | |
| 21:06:02 | mriedem | i didn't really know how to express that in the config option, since we seem to have a weird catch-22 of not wanting to talk about how keystone/ksa encodes and defaults the endpoint interfaces | |
| 21:06:32 | mriedem | i can make the default value "public" if that helps | |
| 21:06:58 | mriedem | i mean, we have publicURL in our default for cinder.catalog_info | |
| 21:07:05 | mriedem | everyone loves that right? | |
| 21:07:25 | sdague | mriedem: right, because it's still assuming k2? | |
| 21:07:37 | mriedem | i guess | |
| 21:07:49 | sdague | mriedem: actually, no I think the existing code in nova works if you only have 1 endpoint, and it's call internal | |
| 21:07:51 | mriedem | public == publicURL as a keystone v3 vs v2 thing is news to me as of today | |
| 21:08:06 | sdague | the code before the patch landed that is | |
| 21:08:19 | sdague | https://github.com/openstack/keystoneauth/blob/9e010b995ba2f19fcc63b14cf5048cedd5ff1e36/keystoneauth1/session.py#L511-L513 | |
| 21:09:39 | mriedem | i'm not following what that has to do with 'internal' as a default | |
| 21:10:06 | sdague | mriedem: I'm trying to figure out if you send down None what it is going to do | |
| 21:10:29 | sdague | I guess it gets converted to public here - https://github.com/openstack/keystoneauth/blob/9e010b995ba2f19fcc63b14cf5048cedd5ff1e36/keystoneauth1/identity/base.py#L209-L210 | |
| 21:10:39 | mriedem | yeah that's what i pointed out above | |
| 21:11:08 | sdague | yeh, sorry, I was thinking that it jumped into service_catalog.py get_endpoints directly | |
| 21:11:20 | mriedem | well, there are certainly plenty of spaghetti hoops in here | |
| 21:11:24 | sdague | given that there are similar named functions in a bunch of modules here | |
| 21:11:24 | mriedem | and plugins and such | |
| 21:11:28 | mriedem | right | |
| 21:11:33 | mriedem | i could have totally broken everything | |
| 21:11:40 | mriedem | i was hoping devstack would tell me | |
| 21:15:12 | mriedem | cdent: also on that unit test thing, it constructs a new object each time, so i'm not sure why i had to move it into __init__ but i blame python | |
| 21:16:47 | cdent | mriedem: probably the joys of global conf getting your face | |
| 21:17:39 | cdent | but yeah, *shrug* | |
| 21:24:16 | openstackgerrit | STEW TY proposed openstack/nova master: Complete verification of servers-action-fixed-ip.inc https://review.openstack.org/338606 | |
| 21:30:16 | cdent | sigh, I'm clearly in the wrong time zone, miss all the cool shit. oh well. | |
| 21:37:35 | openstackgerrit | STEW TY proposed openstack/nova master: Complete verification of servers-action-fixed-ip.inc https://review.openstack.org/338606 | |
| 21:45:09 | openstack | Launchpad bug 1635008 in OpenStack Compute (nova) "rbd credentials in nova.conf overwrite connection_info from cinder" [Low,Fix released] - Assigned to Matt Riedemann (mriedem) | |
| 21:45:09 | knikolla | mriedem: hi. would it be possible to backport the fix for this to newton? https://bugs.launchpad.net/nova/+bug/1635008 | |
| 21:45:45 | mriedem | knikolla: no because it's a behavior change | |
| 21:46:23 | knikolla | mriedem: i see. | |
| 21:46:28 | knikolla | thanks | |
| 21:46:32 | mriedem | sorry, np | |
| 21:51:38 | jamielennox | mriedem, sdague: figure out your ksa issue? | |
| 21:52:12 | sdague | jamielennox: yeh, it was mostly just making sure we didn't encode too much keystone schema into nova. | |
| 21:53:05 | jamielennox | great | |
| 21:53:17 | jamielennox | although out of interest what is keystone schema in this cnotext? | |
| 21:53:37 | jamielennox | from highlights it seems to be mostly interface related | |
| 21:53:54 | mriedem | jamielennox: yeah, | |
| 21:54:05 | sdague | jamielennox: yeh, the interface list | |
| 21:54:09 | mriedem | there was a change to encode the admin/internal/public endpoint interfaces in nova.conf | |
| 21:54:16 | mriedem | ksa also allows 'auth', | |
| 21:54:30 | mriedem | and keystone v2 doesn't restrict the interface it looks like, but the v3 api does | |
| 21:54:51 | sdague | especially because I know when we were talking about service catalog improvements that list was very much an oddity for a lot of ops | |
| 21:55:17 | sdague | so I didn't want to force encode that beyond keystone, which would prevent it evolving in the future | |
| 21:55:55 | jamielennox | i *think* keystone v2 restricts that interface because in v2 'publicURL', 'adminURL' etc are actual fields where in v3 it's just like {'url': ..., 'interface': ...} | |
| 21:56:03 | jamielennox | but i would want to double check that | |
| 21:56:15 | jamielennox | but there's no validation of that in ksa | |
| 21:56:53 | jamielennox | the way cinder and others do this is with a : seperated string, which isn't great but more flexibility than most provide | |
| 21:57:41 | jamielennox | mriedem: i'm not sure what you mean with ksa allows 'auth' - auth is a plugin so something that will be loaded at a very different time | |
| 21:58:36 | mriedem | jamielennox: https://github.com/openstack/keystoneauth/blob/master/keystoneauth1/identity/base.py#L198 | |
| 21:59:11 | mriedem | oh i thought that meant it could be "auth" | |
| 21:59:15 | jamielennox | oh, right | |
| 21:59:19 | mriedem | which meant lookup the endpoint via auth_url config | |
| 21:59:33 | mriedem | it's not really doc'ed from what i can tell | |
| 21:59:43 | jamielennox | that's mostly a special case when you want to look back to the auth_url provided to the plugin and used for some discovery | |
| 21:59:57 | jamielennox | from nova and services perspective you would always expect to be actually looking at the catalog | |
| 22:00:40 | jamielennox | ie, i can't see a reason nova config would want to set that | |
| 22:00:59 | felipemonteiro_ | Is anyone familiar with the consoles API? When I create a console on a running server, then list all the consoles running on that server, I get an empty set. | |
| 22:08:14 | openstackgerrit | Dan Smith proposed openstack/nova master: Make server_external_events cells-aware https://review.openstack.org/433298 | |
| 22:08:23 | dansmith | mriedem: melwitt: with ^ I can _create_ an instance in another cell | |
| 22:10:19 | mriedem | dansmith: should we start a bp for tracking just random api compat things in pike? | |
| 22:10:30 | mriedem | multicell support or whatever | |
| 22:10:40 | dansmith | mriedem: well, I think we'll need a blueprint for all the "make XXX cells aware" stuff yeah | |
| 22:10:45 | dansmith | we know we have a bunch to do | |
| 22:10:57 | dansmith | it'll be a slow background burning type situation | |
| 22:10:58 | mriedem | yeah so let's group that into a bp | |
| 22:10:59 | mriedem | sure | |
| 22:11:05 | mriedem | you want to create that? | |
| 22:11:14 | dansmith | like when you pee and you know something's wrong but you avoid going to the doc for a while | |
| 22:11:21 | dansmith | sure | |
| 22:11:37 | mriedem | cranberry juice will help with that | |
| 22:11:47 | dansmith | won't help with cells unfortunately | |
| 22:11:51 | dansmith | beer might though | |
| 22:11:58 | mriedem | we don't have a cells-sized condom? | |
| 22:12:19 | mriedem | you don't have to answer that | |
| 22:13:10 | dansmith | https://blueprints.launchpad.net/nova/+spec/cells-aware-api | |
| 22:13:27 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Cleanup some issues with CONF.placement.os_interface https://review.openstack.org/433257 | |
| 22:23:34 | openstackgerrit | Pushkar Umaranikar proposed openstack/nova master: Placement api: set custom json_error_formatter in root https://review.openstack.org/433301 | |
| 22:39:37 | openstackgerrit | Anusha Unnam proposed openstack/nova master: Remove mox from nova/tests/unit/virt/vmwareapi/test_configdrive.py https://review.openstack.org/433315 | |
| 22:51:49 | mriedem | https://etherpad.openstack.org/p/nova-ptg-pike is relatively organized now | |