Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-04-08
13:23:04 bauzas ack thanks
13:24:05 sean-k-mooney stephenfin: bauzas lyarwood since ye are all here care to review the final few patches for cyborg. the first two are real staight forward and dont require much knoladge of cyborg
13:24:23 bauzas sean-k-mooney: I can take a look today
13:24:30 sean-k-mooney https://review.opendev.org/#/c/716185/6 https://review.opendev.org/#/c/716186/6 and https://review.opendev.org/#/c/715326/7
13:24:32 lyarwood sean-k-mooney: yup I can
13:24:55 sean-k-mooney i think sundar is nolonger working on this stuff so if there are any issue ill respin the patches
13:25:17 bauzas I'm mostly done on reviewing lyarwood's stable rescue thingies, so either I go with pinned/non-pinned stephenfin's patches or cyborg ones
13:25:32 bauzas but I think cyborg is moar important upstream
13:25:50 bauzas stephenfin: amirite ?
13:26:19 sean-k-mooney they are shorter. the first one is the release note/api docs the second is a minor fix and the last is rebuild/evacuate
13:26:34 sean-k-mooney we should land the first two definetly.
13:26:53 bauzas FWIW, I'd be glad if we could merge my own stuff which is small too :p
13:26:56 sean-k-mooney the last need proper review so if you cover the first two you could always swap back to stephens
13:27:24 bauzas kk
13:27:52 bauzas finishing lyarwood's patch review on this new trait first and then going on cyborg
13:28:10 sean-k-mooney yes im getting coffee but after that im proably goign to go review the rest of your series then lyarwood's or stephenfin's ones
13:31:04 openstackgerrit Lee Yarwood proposed openstack/nova master: fup: Combine SUPPORTED_DEVICE_BUS and SUPPORTED_STORAGE_BUSES https://review.opendev.org/718448
13:36:30 openstackgerrit Lee Yarwood proposed openstack/nova master: fup: Remove the use of the term `unstable rescue` INFO logs https://review.opendev.org/718451
13:43:02 gibi stephenfin: does 'trait:CUSTOM_FOO' : 'foo' is only passing validation if the CUSTOM trait exists or not at all?
13:43:14 gibi sorry 'trait:CUSTOM_FOO' : 'required'
13:43:46 stephenfin Nope, it's a wildcard
13:43:55 stephenfin trait:CUSTOM_*
13:44:25 gibi then I think we have a bug
13:44:46 gibi this fails
13:44:47 gibi body = {'extra_specs': {
13:44:47 gibi def test_create_trait_spec(self):
13:44:48 gibi }}
13:44:48 gibi 'trait:CUSTOM_FOO': 'required',
13:44:50 gibi self.admin_api.post_extra_spec(self.flavor_id, body)
13:46:00 stephenfin gibi: Cool, let me look
13:47:03 openstackgerrit Stephen Finucane proposed openstack/nova master: Add new default roles in server password policies https://review.opendev.org/717417
13:47:50 stephenfin gibi: Wait - are custom traits a thing?
13:48:02 stephenfin I didn't think they were. That's why I've nothing handling them
13:48:28 stephenfin whereas I do for resources https://review.opendev.org/#/c/704643/23/nova/api/validation/extra_specs/resources.py@41
13:48:48 gibi stephenfin: they are https://docs.openstack.org/api-ref/placement/?expanded=update-traits-detail#list-traits
13:49:12 gibi qos using them (but now view flavor extra_spec)
13:49:13 stephenfin gdi, of course they are. Good spot. Will fix
13:49:28 gibi (but not via flavor extra_spec)
13:51:57 sean-k-mooney gibi: we can use custome traits via flavor extra specs
13:52:04 sean-k-mooney or via images
13:52:51 gibi sean-k-mooney: ack. I meant that qos is using CUSTOM traits through the port not the flavor
13:52:52 sean-k-mooney oh you ment qos does not use them via flavor extra specs
13:52:56 gibi yepp
13:53:14 gibi I just type something crazy first :D
13:53:19 gibi typed
13:54:13 sean-k-mooney looking at the validator i dont see wher CUSTOM_ traits are being blocked
13:55:24 sean-k-mooney https://review.opendev.org/#/c/718357/5/nova/api/validation/extra_specs/traits.py is not blocking it anyway
13:55:40 bauzas sean-k-mooney: gibi: I have a functional test that proves it works :p
13:55:54 bauzas (adding custom traits to a flavor)
13:56:39 sean-k-mooney i think we have a couple actully
13:57:07 sean-k-mooney bauzas: dont we have one for the compute status trait
13:57:20 sean-k-mooney we used used a custom one breifly
13:57:21 bauzas sean-k-mooney: indeed, and we always ask for it
13:57:33 bauzas but that's not a custom one
13:57:41 sean-k-mooney not anymore
13:57:48 bauzas root_required=!COMPUTE_DISABLED or something like this
13:58:07 bauzas but that's 'root' required
13:58:16 sean-k-mooney matt used a custome one for dev i can recall if we merged it after the standard one was added and release in os-traits or not
13:58:21 sean-k-mooney we proably waited
13:58:30 bauzas I was afraid of custom traits on a child RP not being bobbled up to the tree for verification but it does work
13:58:30 openstackgerrit Stephen Finucane proposed openstack/nova master: api: Allow custom traits https://review.opendev.org/718461
13:58:39 stephenfin gibi: done ^
13:59:15 bauzas I was able to tag CUSTOM_NVIDIA_XX on a child RP and just request this trait on a flavor which was getting me the right RP as a result
13:59:19 sean-k-mooney bauzas: yes if you add them to the unumbered group i think we check the tree
13:59:34 bauzas sean-k-mooney: that's what the api-ref says at least :)
14:00:06 bauzas https://docs.openstack.org/api-ref/placement/?expanded=list-allocation-candidates-detail#list-allocation-candidates
14:00:25 bauzas "Allocation requests in the response will be for resource providers that have capacity for all requested resources and the set of those resource providers will collectively contain all of the required traits. These traits may be satisfied by any provider in the same non-sharing tree or associated via aggregate. "
14:00:44 bauzas custom traits are just treated equally
14:01:04 sean-k-mooney yep
14:01:45 gibi stephenfin: ack, still someting feels wrong
14:01:46 gibi 'resource:CUSTOM_FOO': 'boo',
14:01:46 gibi body = {'extra_specs': {
14:01:46 gibi def test_create_invalid_resource_spec(self):
14:01:47 gibi }}
14:01:49 gibi self.admin_api.post_extra_spec(self.flavor_id, body)
14:02:05 gibi this should fail as 'boo' is not int, but it is accepted
14:02:14 gibi haven't checked the latest ps
14:03:05 gibi pulling it now
14:03:22 stephenfin gimme a few
14:05:16 openstackgerrit Lee Yarwood proposed openstack/nova master: fup: Add missing docstrings from get_rescue_device|bus diskinfo funcs https://review.opendev.org/718464
14:06:07 sean-k-mooney gibi: are you useing the correct microversion
14:06:29 gibi sean-k-mooney: I'm using the functional test env
14:06:47 gibi sean-k-mooney: and that sets it to 2.86
14:06:49 sean-k-mooney right well it might not be using the latest microverions
14:06:53 sean-k-mooney ah ok
14:06:56 sean-k-mooney https://review.opendev.org/#/c/704643/23/nova/api/validation/extra_specs/resources.py
14:07:03 sean-k-mooney so the valu is defiend as an int
14:07:16 gibi there are a bunch of tests alreay there that works as expected
14:07:36 gibi something is wrong with the parameterized namespaced keys
14:07:47 sean-k-mooney its possible that boo is being treated as hex
14:08:08 sean-k-mooney actully proably not
14:08:51 gibi yepp, it accepts not-an-int as well
14:09:06 stephenfin gibi: s/resource/resources/
14:09:29 gibi /o\
14:09:55 sean-k-mooney stephenfin: should your typo chekcing not catch that :P
14:10:11 stephenfin nope, it's an unknown namespace :)
14:10:25 gibi stephenfin: you are right. resources works as expected
14:10:30 sean-k-mooney ... oh because we are droping ?validation
14:10:31 openstackgerrit Stephen Finucane proposed openstack/nova master: api: Allow custom traits https://review.opendev.org/718461
14:12:37 openstackgerrit Lee Yarwood proposed openstack/nova master: DNM - Test stable device rescue tests with BFV instances https://review.opendev.org/710050

Earlier   Later