| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-08-28 | |||
| 15:18:12 | openstackgerrit | Hamdy Khader proposed openstack/nova master: Adding NVMEoF for libvirt driver https://review.openstack.org/482640 | |
| 15:18:15 | dansmith | mriedem: so for this migration spec, do you want the spec to be for "use uuid to hold allocations during a migration" or just "migrations should have uuids because uuids are great" ? | |
| 15:18:38 | mriedem | the former | |
| 15:18:47 | dansmith | yeah okay | |
| 15:19:03 | mriedem | the problem statement is really related to the issues we found with move operations in pike with ocata computes overwriting n stuff | |
| 15:20:10 | mriedem | also have to think about at what point is the allocation dropped from the source node, and the migration consumer allocation 'transferred' to the instance consumer | |
| 15:20:33 | mriedem | maybe that's just a single PUT /resource_providers/uuid/allocations call? | |
| 15:21:13 | dansmith | um, say what? you mean on a revert? | |
| 15:23:11 | sdague | mriedem: yeh, can do, I guess that is done automatically after it's a landed patch? | |
| 15:23:18 | sdague | I was just using the gerrit ui | |
| 15:27:21 | mriedem | sdague: yeah for gerrit ui to do it the original has to be merged | |
| 15:27:40 | sdague | gotcha, yeh, I'll redo after the master merge. My bad | |
| 15:27:47 | mriedem | dansmith: doesn't the allocation on the dest node at some point have to transfer from the migration object to the instance object? | |
| 15:28:05 | mriedem | especially if we ever count usage using placement for quota stuff | |
| 15:28:35 | dansmith | mriedem: well I was going to do it the other way: | |
| 15:29:17 | dansmith | mriedem: replace the source allocation with one for the migration, then allocate for the instance on the destination, assuming that's the expected end state | |
| 15:29:55 | mriedem | and once the move is done, drop the migration allocation from the source node | |
| 15:29:57 | mriedem | that works too | |
| 15:30:01 | dansmith | right | |
| 15:30:30 | dansmith | there is a gotcha in there in that we need space on the source to claim again for the size of the instance, but we could fix that with another operation against placement letting us do an atomic swap | |
| 15:44:07 | openstackgerrit | Merged openstack/nova master: Don't warn on expected network-vif-unplugged events https://review.openstack.org/465794 | |
| 15:44:36 | openstackgerrit | Merged openstack/nova master: Add device tag support info in support matrix https://review.openstack.org/481478 | |
| 15:45:02 | openstackgerrit | Merged openstack/nova master: conf: Allow users to unset 'keymap' options https://review.openstack.org/496605 | |
| 15:45:23 | openstackgerrit | Merged openstack/nova master: Deprecate CONF.monkey_patch https://review.openstack.org/498113 | |
| 15:58:03 | efried | mriedem Not sure if you're caught up on the ML yet, but ^^ vs. https://review.openstack.org/#/c/494305/ -- one of 'em needs to be reverted, nah? | |
| 16:01:19 | openstackgerrit | Dan Smith proposed openstack/nova-specs master: WIP: Add migration-allocations spec https://review.openstack.org/498510 | |
| 16:01:50 | dansmith | mriedem: ^ | |
| 16:03:37 | mriedem | efried: thanks, commented on the puppet review | |
| 16:12:59 | openstackgerrit | Chris Dent proposed openstack/nova master: Optimize MiniDNS for fewer syscalls https://review.openstack.org/486829 | |
| 16:15:20 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Provide hints when nova-manage db sync fails to sync cell0 https://review.openstack.org/486660 | |
| 16:23:56 | openstackgerrit | Sean McCully proposed openstack/nova master: iso8601.is8601.Utc No Longer Exists https://review.openstack.org/498287 | |
| 16:40:32 | openstackgerrit | Merged openstack/nova master: Remove useless error handling in prep_resize https://review.openstack.org/497976 | |
| 16:51:35 | openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Return alternate allocation requests to scheduler https://review.openstack.org/471927 | |
| 16:53:32 | ssmith | Good morning. We're trying to prevent an instance from being deleted and have selected LOCK INSTANCE from the UI but people can still delete it. | |
| 16:53:48 | mriedem1 | people == admins? | |
| 16:54:35 | ssmith | mriedem1: Yes, they are admins | |
| 16:56:02 | mriedem | ok admins can do stuff to instances even if they are locked | |
| 16:56:08 | ssmith | mriedem: Yes, they are admins. I'm fine with them unlocking it and then deleting | |
| 16:56:22 | ssmith | Is that in a policy somewhere that can be disabled? | |
| 16:56:33 | mriedem | in this case it's hard-coded in code | |
| 16:56:46 | mriedem | https://github.com/openstack/nova/blob/16.0.0.0rc2/nova/compute/api.py#L2120 | |
| 16:56:56 | mriedem | https://github.com/openstack/nova/blob/16.0.0.0rc2/nova/compute/api.py#L198 | |
| 16:57:50 | mriedem | ssmith: https://review.openstack.org/#/c/492477/ is semi related | |
| 16:59:07 | ssmith | So remove " and not context.is_admin" from the code? | |
| 16:59:24 | mriedem | well, as i suggested in that patch, it could be a configurable policy check | |
| 16:59:59 | ssmith | I agree with the need for the patch | |
| 17:00:01 | mriedem | that patch is wrong in two ways: 1. it's not on the master branch and 2. it's config-driven API via nova.conf; configuring access to things in the API should be via policy rules, not nova.conf | |
| 17:00:29 | mriedem | i wasn't -2 on the thing it's trying to address, just the way it's triyng to do it | |
| 17:00:36 | mriedem | plus it's on stable/ocata which is wrong | |
| 17:00:56 | ssmith | We're running Newton | |
| 17:01:15 | mriedem | all changes start on master, so whatever the solution, it has to start on master | |
| 18:50:24 | openstackgerrit | Merged openstack/nova master: docs: Document the scheduler workflow https://review.openstack.org/475810 | |
| 18:50:49 | openstackgerrit | Merged openstack/nova master: Monkey patch the blockdiag extension https://review.openstack.org/476159 | |
| 18:51:25 | openstackgerrit | Merged openstack/nova master: Add formatting to scheduling activity diagram https://review.openstack.org/476204 | |
| 18:51:55 | openstackgerrit | Merged openstack/nova master: Update PCI passthrough doc for moved options https://review.openstack.org/498461 | |
| 18:52:18 | openstackgerrit | Merged openstack/nova master: VMware: Do not check if folder already exists in vCenter https://review.openstack.org/376387 | |
| 19:10:57 | edleafe | OK, nova objects experts: what sort of Field would I use to store a dict like this: http://paste.openstack.org/show/619670/ | |
| 19:12:50 | artom | edleafe, I think that depends on how you're going to be accessing it | |
| 19:13:44 | edleafe | artom: probably just obj.foo to get the dict | |
| 19:14:47 | cdent | edleafe: any of this help https://github.com/openstack/nova/blob/master/nova/objects/resource_provider.py#L2338-L2359 | |
| 19:14:56 | artom | edleafe, I would think it'd be just a Dict then | |
| 19:16:03 | cdent | edleafe: of https://github.com/openstack/nova/blob/master/nova/objects/resource_provider.py#L2437-L2455 | |
| 19:16:11 | cdent | basically reuse the objects use to create the data in the first place | |
| 19:16:34 | cdent | If people want structured objects... | |
| 19:16:51 | edleafe | artom: but I got the feeling that fields.Dict was not to be used anymore | |
| 19:16:54 | edleafe | artom: https://github.com/openstack/nova/blob/master/nova/objects/fields.py#L72 | |
| 19:17:27 | openstackgerrit | Merged openstack/nova master: VMware: Handle missing volume vmdk during detach https://review.openstack.org/484675 | |
| 19:17:30 | artom | edleafe, ah, true, didn't see the comment | |
| 19:17:35 | artom | dansmith's the expert | |
| 19:17:46 | edleafe | cdent: The object needs to have the hostname and the corresponding allocation | |
| 19:17:51 | edleafe | that's it | |
| 19:17:55 | openstackgerrit | Merged openstack/nova master: libvirt: Fix getting a wrong guest object https://review.openstack.org/496515 | |
| 19:17:55 | artom | cdent, I think the idea behind objects is a poor man's type system | |
| 19:18:16 | artom | So that we don't have random unversioned unknown undiscoverable dicts flying around | |
| 19:18:25 | openstackgerrit | Merged openstack/nova master: Tests: Add cleanup of 'instances' directory https://review.openstack.org/491589 | |
| 19:18:28 | cdent | artom: I know. Me and types have never felt all that friendly. | |
| 19:18:51 | artom | cdent, typist ;) | |
| 19:20:05 | edleafe | cdent: I was thinking about using AllocationList, but that has DB stuff. This should just be a blob to send back | |
| 19:20:43 | cdent | edleafe: the other thing to keep in mind is that we really don’t want to be using placement objects on the nova side of the equation if we can help it | |
| 19:20:53 | cdent | so a fresh object might be the way to go | |
| 19:20:55 | edleafe | cdent: yeah, that too | |
| 19:21:08 | mriedem | edleafe: https://github.com/openstack/nova/blob/master/nova/objects/fields.py#L72 doesn't mean you can't use them, | |
| 19:21:14 | mriedem | it means use the fields from ovo | |
| 19:21:15 | mriedem | rather than nova | |
| 19:21:16 | edleafe | but this allocations dict doesn't seem to fit any of the ovoo types | |
| 19:21:19 | mriedem | as we're moving stuff from nova to ova | |
| 19:21:21 | mriedem | *ovo | |
| 19:21:52 | mriedem | so you have an Allocation object | |
| 19:22:00 | mriedem | which has a resource_provider and resources field | |
| 19:22:02 | cdent | edleafe: so you’re idea at this point is an object that is effectively; ‘target’: String, ‘allocation’: Dict | |
| 19:22:29 | cdent | could be moultiple providers | |
| 19:22:45 | edleafe | cdent: pretty much | |
| 19:22:46 | cdent | is the idea to maintain the opacity of the allocation request bits? | |
| 19:23:16 | edleafe | IMO, the allocation dict is already too much placement internals to be passed around in Nova | |
| 19:23:24 | edleafe | But everyone else seems fine with it, so... | |
| 19:24:04 | edleafe | cdent: and the example I pasted contains multiple providers. It will only have one hostname, though | |
| 19:24:10 | edleafe | We're still in nova-land | |
| 19:24:21 | cdent | I was responding to mriedem on the rp thing | |
| 19:24:47 | cdent | I’m ambivalent about the allocation requests being passed around. sailed ship | |
| 19:25:17 | mriedem | passing allocation requests dicts around within the scheduler, as we are today, seems fine | |