Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-10-23
15:20:29 mriedem dansmith: are you ok with this before i propose a backport https://review.openstack.org/#/c/513001/
15:20:54 dansmith omg no
15:20:55 dansmith wtf
15:21:10 dansmith ugh
15:21:21 efried So... not okay with it
15:21:36 efried Just reading between the lines here.
15:21:38 mriedem i brought this up when it was initially being reviewed, pretty sure anyway - but there is no api schema restriction in the baremetal api about the resource class name
15:22:07 dansmith I very specifically think this is completely broken
15:22:11 dansmith I know and I fought hard to NOT do this ambiguous thing
15:22:51 openstackgerrit Matt Riedemann proposed openstack/nova master: Only add CUSTOM_ prefix if required https://review.openstack.org/513001
15:22:56 mriedem pulled it out of the gate
15:22:59 mriedem johnthetubaguy: ^
15:24:17 johnthetubaguy so VCPU becomes CUSTOM_VCPU, just CUSTOM_FOO stays CUSTOM_FOO
15:24:33 johnthetubaguy its a pain to setup the system with CUSTOM_ in only half the places
15:24:58 mriedem VCPU is a standared resource class
15:25:41 johnthetubaguy yep, that's where it gets funky, that code keeps it as always CUSTOM_ at the start, it just avoids the double encoding
15:25:43 mriedem if you set the ironic node.resource_class as vcpu, then we'd translate that to CUSTOM_VCPU yeah
15:25:56 dansmith mriedem: thanks
15:25:57 johnthetubaguy that's what I mean, I should have added a test for that really
15:26:23 dansmith I just commented on that patch,
15:26:24 mriedem the ironic docs are at least pretty clear about this https://docs.openstack.org/ironic/pike/install/configure-nova-flavors.html#scheduling-based-on-resource-classes
15:26:31 dansmith but I'm super against us adding this ambiguity
15:27:28 dansmith anything else that wants to look at placement-related data for ironic nodes would have to implement the same "is it already right, or not" behavior when taking the ironic class and turning it into a placement class
15:27:44 dansmith instead of just "apply the following set of rules to get from ironic to placement"
15:27:46 openstackgerrit Stephen Finucane proposed openstack/nova master: console: introduce framework for RFB authentication https://review.openstack.org/345397
15:27:46 openstackgerrit Stephen Finucane proposed openstack/nova master: console: introduce the VeNCrypt RFB authentication scheme https://review.openstack.org/345398
15:27:47 openstackgerrit Stephen Finucane proposed openstack/nova master: console: provide an RFB security proxy implementation https://review.openstack.org/345399
15:27:47 openstackgerrit Stephen Finucane proposed openstack/nova master: doc: Document TLS security setup for noVNC proxy https://review.openstack.org/500544
15:27:49 dansmith as documented in the ironic docs there
15:28:19 johnthetubaguy I get that, just as an operator, I need to remember if its Nova or Ironic I am talking to when setting different names for things
15:28:54 dansmith but they're two totally different things
15:29:01 dansmith ironic's resource class predates our use of it,
15:29:13 dansmith and can be set to completely incompatible things
15:29:37 mriedem yeah, note the baremetal.with-GPU example in the ironic docs,
15:29:46 mriedem nova replaces special characters like the '.' with underscores
15:30:05 mriedem so the CUSTOM_ prefix isn't the only thing that will be different
15:30:48 johnthetubaguy dansmith: the spec says it was added for placement: https://specs.openstack.org/openstack/ironic-specs/specs/not-implemented/node-resource-class.html
15:31:13 johnthetubaguy my thinking was more simplistic really... can't I just set all the strings to the same thing?
15:31:16 dansmith johnthetubaguy: but like two releases before we actually had that stuff, and with zero restriction :)
15:31:28 openstackgerrit Stephen Finucane proposed openstack/nova master: Move the idmapshift binary into privsep. https://review.openstack.org/495541
15:31:28 openstackgerrit Stephen Finucane proposed openstack/nova master: Move loopback setup and removal to privsep. https://review.openstack.org/495664
15:31:29 openstackgerrit Stephen Finucane proposed openstack/nova master: Move nbd commands to privsep. https://review.openstack.org/500351
15:31:29 openstackgerrit Stephen Finucane proposed openstack/nova master: Move kpartx calls to privsep. https://review.openstack.org/500354
15:31:30 openstackgerrit Stephen Finucane proposed openstack/nova master: Move blkid calls to privsep. https://review.openstack.org/500398
15:31:31 johnthetubaguy dansmith: yep, agreed.
15:31:36 dansmith johnthetubaguy: yeah I know, but it's not equivalent
15:32:01 dansmith johnthetubaguy: it'd be like us using the uuid, as CUSTOM_$UUID and then people wanting to set their uuid to CUSTOM_$UUID... they're not the same field
15:33:55 johnthetubaguy it just seems to me like 80% of folks are setting the resource classes to something they can request in a flavor, they care little about the translation, even if they probably should
15:34:36 johnthetubaguy but I guess its dangerous to assume they are the same thing, when you use them, because they may not be
15:34:41 dansmith and integers are like floats, but usually without the decimal point?
15:36:21 johnthetubaguy I was thinking about following up with a new microversion that forces all new resource classes to follow the placement patterns, etc.
15:36:54 johnthetubaguy but that is kinda aweful in its own way
15:36:59 dansmith you have to migrate all their node names currently
15:37:57 dansmith and will you enforce CUSTOM_ as a prefix? you kinda have to, but then you have chicken and egg...
15:38:26 dansmith the ironic docs are exceedingly clear on this point
15:39:07 edleafe With hindsight, maybe prefixing standard classes with STD_ would have been simpler
15:39:41 dansmith I don't think that changes this particular thing, but I agree it'd be nice to have consistent namespacing instead of one magic one
15:39:43 openstackgerrit priyaduggirala proposed openstack/nova master: Rename parameters in call() of nova/image/glance.py https://review.openstack.org/508533
15:42:22 dansmith another argument for operator confusion would be:
15:42:33 dansmith I named one class CUSTOM_GOLD and another SILVER
15:42:48 dansmith both show up in placement as CUSTOM_$thing, yet one tool can never find the silver ones for some reason
15:42:49 dansmith or
15:42:58 dansmith I have CUSTOM_SILVER and CSUTOM_GOLD
15:43:10 dansmith and the gold one shows up in placement as CUSTOM_CSUTOM_GOLD for some reason
15:51:48 openstackgerrit John Garbutt proposed openstack/nova master: Add test so we remember why CUSTOM_ prefix added https://review.openstack.org/513001
15:52:56 johnthetubaguy dansmith: I turned the change upside down ^
15:53:48 dansmith johnthetubaguy: +2 thanks
15:55:40 openstackgerrit Murali Annamneni proposed openstack/nova master: Enables MySQL Cluster Support for Nova https://review.openstack.org/446643
16:14:59 openstackgerrit Eric Berglund proposed openstack/nova master: PowerVM Driver: config drive https://review.openstack.org/409404
16:15:17 openstackgerrit Brianna Poulos proposed openstack/nova master: Add trusted_image_certificates to REST API https://review.openstack.org/486204
16:19:18 openstackgerrit Matt Riedemann proposed openstack/nova master: db: add flavors.description column https://review.openstack.org/514371
16:27:11 openstackgerrit Stephen Finucane proposed openstack/nova master: xenapi: cached images should be cleaned up by time https://review.openstack.org/465954
16:29:17 openstackgerrit Stephen Finucane proposed openstack/nova master: xenapi: cached images should be cleaned up by time https://review.openstack.org/465954
16:36:05 openstackgerrit Murali Annamneni proposed openstack/nova master: Enables MySQL Cluster Support for Nova https://review.openstack.org/446643
16:46:12 mriedem huh, it's odd that we allow flavors.name = None in the object but name is not nullable in the db
16:57:10 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/nova master: Change 'InstancePCIRequest' spec field https://review.openstack.org/449257
17:07:11 efried mriedem How come https://review.openstack.org/#/c/510244/ isn't showing up in https://specs.openstack.org/openstack/nova-specs/specs/queens/ ?
17:08:40 mriedem busted docs job?
17:09:12 mriedem looks like there is no publish job
17:09:46 efried Is this a post-zuulv3 thing?
17:09:53 mriedem i believe so
17:10:10 efried Other specs are in that list, but I haven't dug in to see if any were ones that merged after the zuulv3 cutover
17:10:44 efried Should I follow up in -infra? Or -docs?
17:10:55 efried or you got this?
17:11:09 mriedem find out where "openstack-specs-jobs" is defined
17:11:42 mriedem looks like it's in openstack-zuul-jobs
17:11:57 mriedem there is a post job "publish-openstack-specs"
17:12:01 mriedem that should publish the latest
17:12:05 mriedem not sure if that's broken or what
17:12:49 mriedem efried: you should follow up in -infra
17:12:56 efried mriedem ack
17:19:48 mriedem dansmith: want to push this through the specs repo to trigger the publish job? https://review.openstack.org/#/c/456756/
17:28:29 dansmith mriedem: done
17:40:36 openstackgerrit Merged openstack/nova-specs master: Add a new section: "Upgrade impact" to the template https://review.openstack.org/456756
18:07:06 mriedem efried: ^ should have flushed a post job
18:07:21 efried mriedem Not yet https://specs.openstack.org/openstack/nova-specs/specs/queens/index.html
18:07:25 efried been watching.
18:28:16 mriedem dansmith: remember this test? https://github.com/openstack/nova/blob/master/nova/tests/unit/db/test_sqlalchemy_migration.py#L379
18:28:50 mriedem it's causing me some problems with the flavor.description stuff because it doesn't run the schema change that adds the description column, but it creates a flavor using the object, which has a flavor attribute now,
18:29:00 mriedem so trying to think of non terrible ways to resolve that in the test

Earlier   Later