| *** edmondsw has joined #openstack-placement | 00:23 | |
| *** edmondsw has quit IRC | 00:28 | |
| *** tetsuro has joined #openstack-placement | 01:18 | |
| openstackgerrit | Brin Zhang proposed openstack/nova-specs master: Add support specify volume type when boot instance https://review.openstack.org/579520 | 01:31 |
|---|---|---|
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add policy to InstanceGroup object and api models. https://review.openstack.org/563375 | 02:38 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add policy field to ServerGroup notification object https://review.openstack.org/563401 | 03:06 |
| openstackgerrit | Brin Zhang proposed openstack/nova master: Add fault info when unshelve instance failed https://review.openstack.org/579747 | 03:12 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Change the anti-affinity Filter to adapt to new policy https://review.openstack.org/571166 | 03:31 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Change the anti-affinity Filter to adapt to new policy https://review.openstack.org/571166 | 03:32 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Adapt _validate_instance_group_policy to new policy model https://review.openstack.org/571465 | 03:32 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Microversion 2.64 - Use new format policy in server group https://review.openstack.org/567534 | 03:33 |
| openstackgerrit | Merged openstack/osc-placement master: Add nested resource providers (v1.14) https://review.openstack.org/546675 | 03:56 |
| openstackgerrit | Merged openstack/osc-placement master: Limit allocation candidates (v1.15, v1.16) https://review.openstack.org/548043 | 03:56 |
| openstackgerrit | Merged openstack/osc-placement master: Allocation candidates parameter: required (v1.17) https://review.openstack.org/548326 | 03:56 |
| openstackgerrit | Deepak Mourya proposed openstack/nova master: Resource_provider API handler does not return specific error codes https://review.openstack.org/579423 | 04:36 |
| *** takashin has joined #openstack-placement | 05:29 | |
| *** edmondsw has joined #openstack-placement | 05:48 | |
| *** edmondsw has quit IRC | 05:53 | |
| *** nicolasbock has joined #openstack-placement | 06:42 | |
| *** openstackgerrit has quit IRC | 06:49 | |
| *** openstackgerrit has joined #openstack-placement | 06:54 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/nova master: Imported Translations from Zanata https://review.openstack.org/578019 | 06:54 |
| *** peereb has joined #openstack-placement | 07:21 | |
| *** tssurya has joined #openstack-placement | 07:31 | |
| openstackgerrit | Tuan Do Anh proposed openstack/nova-specs master: fix tox python3 overrides https://review.openstack.org/579793 | 07:31 |
| *** edmondsw has joined #openstack-placement | 07:36 | |
| *** edmondsw has quit IRC | 07:41 | |
| openstackgerrit | Lei Zhang proposed openstack/nova master: Add method to get cpu traits https://review.openstack.org/560317 | 07:47 |
| *** avolkov has joined #openstack-placement | 07:51 | |
| openstackgerrit | Claudiu Belu proposed openstack/nova master: tests: autospecs all the mock.patch usages https://review.openstack.org/470775 | 07:56 |
| openstackgerrit | Claudiu Belu proposed openstack/nova master: hyper-v: autospec classes before they are instantiated https://review.openstack.org/342211 | 07:56 |
| openstackgerrit | Claudiu Belu proposed openstack/nova master: WIP: replace spec with autospec https://review.openstack.org/557299 | 08:01 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Prevent updating an RP's parent to form a loop https://review.openstack.org/579618 | 08:08 |
| openstackgerrit | Merged openstack/nova master: Merge server create schema for keypair extension https://review.openstack.org/578332 | 08:31 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add InstanceGroupPolicy object https://review.openstack.org/573628 | 08:33 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Refactor the policies to policy https://review.openstack.org/579113 | 08:33 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add policy to InstanceGroup object and api models. https://review.openstack.org/563375 | 08:33 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add policy field to ServerGroup notification object https://review.openstack.org/563401 | 08:33 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Change the anti-affinity Filter to adapt to new policy https://review.openstack.org/571166 | 08:33 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Adapt _validate_instance_group_policy to new policy model https://review.openstack.org/571465 | 08:34 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Microversion 2.64 - Use new format policy in server group https://review.openstack.org/567534 | 08:34 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Transform missing delete notifications https://review.openstack.org/410297 | 08:50 |
| *** cdent has joined #openstack-placement | 08:56 | |
| *** cdent_ has joined #openstack-placement | 09:04 | |
| *** cdent has quit IRC | 09:05 | |
| *** cdent_ is now known as cdent | 09:05 | |
| openstackgerrit | Takashi NATSUME proposed openstack/nova master: Transform instance-live_migration_post notification https://review.openstack.org/480119 | 09:06 |
| openstackgerrit | jiang wei proposed openstack/nova master: Add action initiator attribute to the instance payload https://review.openstack.org/536243 | 09:10 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Fix unbound local when saving an unchanged RequestSpec https://review.openstack.org/573768 | 09:12 |
| *** edmondsw has joined #openstack-placement | 09:25 | |
| openstackgerrit | Silvan Kaiser proposed openstack/nova master: Exec systemd-run with privileges in Quobyte driver https://review.openstack.org/554195 | 09:25 |
| openstackgerrit | yuqian proposed openstack/nova master: vmware does't support VirtualSriovEthernetCard https://review.openstack.org/579828 | 09:26 |
| *** edmondsw has quit IRC | 09:29 | |
| *** takashin has left #openstack-placement | 09:30 | |
| *** e0ne has joined #openstack-placement | 09:33 | |
| openstackgerrit | Brin Zhang proposed openstack/nova master: Add unshelve instance error info to fault table https://review.openstack.org/579747 | 09:45 |
| *** tetsuro has quit IRC | 09:49 | |
| *** nicolasbock has quit IRC | 09:55 | |
| *** tssurya has quit IRC | 10:06 | |
| *** cdent has quit IRC | 10:25 | |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Escalate UUID validation warning to error in test https://review.openstack.org/540386 | 11:21 |
| *** tssurya has joined #openstack-placement | 11:49 | |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Call generate_image_url only for legacy notification https://review.openstack.org/564528 | 11:50 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Call generate_image_url only for legacy notification https://review.openstack.org/564528 | 11:51 |
| *** rubasov has quit IRC | 12:13 | |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Enhance service restart in functional env https://review.openstack.org/512552 | 12:19 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Reproduce bug 1724172 in the functional test env https://review.openstack.org/512553 | 12:19 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: cleanup evacuated instances not on hypervisor https://review.openstack.org/512623 | 12:19 |
| openstack | bug 1724172 in OpenStack Compute (nova) "Allocation of an evacuated instance is not cleaned on the source host if instance is not defined on the hypervisor" [Medium,In progress] https://launchpad.net/bugs/1724172 - Assigned to Balazs Gibizer (balazs-gibizer) | 12:19 |
| *** rubasov has joined #openstack-placement | 12:20 | |
| openstackgerrit | jiang wei proposed openstack/nova master: Add action initiator attribute to the instance payload https://review.openstack.org/536243 | 12:49 |
| *** mriedem has joined #openstack-placement | 12:51 | |
| openstackgerrit | Merged openstack/nova master: Handle nested serialized json entries in assertJsonEqual https://review.openstack.org/579615 | 12:54 |
| openstackgerrit | Matt Riedemann proposed openstack/nova stable/queens: Use ironic-tempest-dsvm-ipa-wholedisk-bios-agent_ipmitool-tinyipa in tree https://review.openstack.org/579862 | 13:01 |
| *** edmondsw has joined #openstack-placement | 13:01 | |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Fix unbound local when saving an unchanged RequestSpec https://review.openstack.org/573768 | 13:02 |
| *** edmondsw has quit IRC | 13:06 | |
| mriedem | jaypipes or dansmith: can i get a review on the consumer heal allocations CLI change - has a +2 https://review.openstack.org/#/c/574488/ - then we can close out that blueprint | 13:06 |
| *** mriedem is now known as mriedem_afk | 13:09 | |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: cover migration cases with functional tests https://review.openstack.org/493865 | 13:18 |
| *** tssurya has quit IRC | 13:21 | |
| efried | jaypipes: I was trying to make sure that I could set it to whatever I want when I'm creating the child RPs for my devices. But it's not going to help me anyway: I've got no good way to correlate a UUID with a device without some kind of persistent storage. I have to do it via the name - which means my virt driver is going to have to talk to placement, because the allocation record I get only has the UUID of the RP. | 13:24 |
| *** bhagyashris has joined #openstack-placement | 13:26 | |
| openstackgerrit | Margarita Shakhova proposed openstack/nova master: Do not skip migrations in _destroy_evacuated_instances() https://review.openstack.org/563623 | 13:37 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Send soft_delete from context manager https://review.openstack.org/476459 | 13:43 |
| jaypipes | efried: wouldn't you get the provider tree which would allow you to look up the UUID from the name of the provider? | 13:43 |
| *** mriedem_afk is now known as mriedem | 13:44 | |
| efried | jaypipes: I'm trying to connect what I generate via update_provider_tree to how I need to correlate it back to a real device from spawn. | 13:44 |
| efried | jaypipes: In either case, yes, I'll need to look up the name from the UUID. | 13:45 |
| dansmith | mriedem: queued | 13:50 |
| efried | mriedem, dansmith: I'm working my way through it fwiw. | 13:51 |
| efried | nearly done | 13:51 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Wait for vif plugging during live migration job https://review.openstack.org/578551 | 13:52 |
| dansmith | ack | 13:56 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Stop setting glance_api_version in cinder.conf in nova-live-migration https://review.openstack.org/579871 | 13:56 |
| *** mriedem has quit IRC | 14:04 | |
| *** mriedem has joined #openstack-placement | 14:04 | |
| efried | dansmith: I left comments but no vote there, would like to see if you think the missing tests warrant a respin. (mriedem fyi) | 14:05 |
| dansmith | wow, mriedem was missing tests? | 14:05 |
| dansmith | kinda seems like the meat of the patch | 14:07 |
| efried | Maybe I'm crazy, maybe those scenarios were covered by existing tests, but I would find that odd considering the code paths in question are new. | 14:08 |
| *** tssurya has joined #openstack-placement | 14:08 | |
| mriedem | looking | 14:09 |
| mriedem | there are existing functional tests that hit this | 14:10 |
| mriedem | TestNovaManagePlacementHealAllocations | 14:11 |
| mriedem | or wait maybe i'm answering the wrong question | 14:11 |
| efried | mriedem: Tests that yield a nonempty response payload with an empty allocations member? | 14:12 |
| efried | mriedem: And that return a payload with the correct proj and user already set? | 14:12 |
| dansmith | yeah both are missing afaict | 14:12 |
| dansmith | since they're new behaviors, I don't think they're likely to be covered by existing tests | 14:12 |
| mriedem | when pulling allocatoins, placement will automatically fill in the sentinels from config | 14:12 |
| mriedem | so i can't functionally test that the project_id / user_id aren't set at all because the placement api sets those | 14:13 |
| efried | yup, not talking about that | 14:13 |
| efried | Talking about one scenario where the GET returns | 14:14 |
| efried | { proj: something, | 14:14 |
| efried | user: something, | 14:14 |
| efried | generation: something, | 14:14 |
| efried | allocations: {} | 14:14 |
| efried | } | 14:14 |
| efried | You should skip this because of the second part of the condition you added at the front. | 14:14 |
| efried | And a second scenario where the GET returns | 14:15 |
| efried | { proj: correct thing, | 14:15 |
| efried | user: correct thing, | 14:15 |
| efried | ... | 14:15 |
| efried | } | 14:15 |
| efried | You should skip this due to the new code where you check equality against the instance's proj/user | 14:15 |
| dansmith | right | 14:16 |
| * efried carefully folds gibi's cape and returns it to his foot locker. | 14:17 | |
| mriedem | replies inline | 14:18 |
| mriedem | for your 2 scenarios above, | 14:19 |
| mriedem | the latter is covered by the existing functional tests - change the continue to an exception, run the functional tests and see them blow up | 14:19 |
| mriedem | for the former, can we have empty allocations with project/user/generation? | 14:20 |
| mriedem | no we can't | 14:20 |
| mriedem | where project/user weren't provided | 14:20 |
| mriedem | oops | 14:20 |
| openstackgerrit | Takashi NATSUME proposed openstack/nova master: Transform instance-live_migration_post notification https://review.openstack.org/480119 | 14:21 |
| mriedem | https://github.com/openstack/nova/blob/master/nova/api/openstack/placement/handlers/allocation.py#L93 | 14:21 |
| mriedem | if there are no allocations, you don't get project/user/generation in the response | 14:21 |
| efried | ahyes, which is something jaypipes is fixing via https://review.openstack.org/#/c/579654/ but not yet merged. | 14:22 |
| mriedem | umm | 14:22 |
| mriedem | that would be a microversion if so | 14:22 |
| efried | I thought no, but edleafe agrees with you. | 14:22 |
| mriedem | "oops" doesn't mean you can't bypass the microversion laws | 14:23 |
| mriedem | *can | 14:23 |
| efried | My reasoning was along the lines of, this is broken, and you shouldn't have to opt out of broken behavior. I thought I'd heard you say that a time or two. | 14:23 |
| mriedem | for 500s yeah | 14:24 |
| efried | okay. | 14:24 |
| mriedem | or stuff that corrupts your data | 14:24 |
| mriedem | those are generally behavioral changes in the api, not response schema | 14:24 |
| jaypipes | efried, mriedem: actually in this case, https://bugs.launchpad.net/nova/+bug/1779717 is the more relevant (and blocking) bug. | 14:30 |
| openstack | Launchpad bug 1779717 in OpenStack Compute (nova) "No ability to update consumer's project and/or user external ID" [Medium,Triaged] - Assigned to Jay Pipes (jaypipes) | 14:30 |
| efried | oh riiight, duh, forgot all about that. | 14:31 |
| mriedem | jaypipes: is that also true of POST /allocations? https://developer.openstack.org/api-ref/placement/#manage-allocations | 14:36 |
| jaypipes | mriedem: yes | 14:37 |
| jaypipes | mriedem: both use the same util.ensure_consumer() calls. | 14:38 |
| jaypipes | mriedem: and that call doesn't update a consumer's information. | 14:38 |
| mriedem | alright, i'll work on a different functional test for my heal_allocations change in the meantime | 14:39 |
| jaypipes | mriedem: this is something that wasn't a problem when we were just jamming project and user ID into the allocations table, but now that we have a consumers table which stores this normalized data, I forgot to consider this, unfortunatyely. | 14:39 |
| jaypipes | mriedem: my apologies. I should have fixes up soon. | 14:39 |
| *** edmondsw has joined #openstack-placement | 14:51 | |
| *** edmondsw has quit IRC | 14:55 | |
| openstackgerrit | jiang wei proposed openstack/nova master: Add action initiator attribute to the instance payload https://review.openstack.org/536243 | 15:01 |
| openstackgerrit | Konstantinos Samaras-Tsakiris proposed openstack/nova master: Hide hypervisor id on windows guests https://review.openstack.org/579897 | 15:02 |
| *** peereb has quit IRC | 15:05 | |
| *** e0ne has quit IRC | 15:29 | |
| openstackgerrit | Merged openstack/nova master: Update xenapi_disable_agent config option usage in docs https://review.openstack.org/578848 | 15:36 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: libvirt: use dest host vif migrate details for live migration https://review.openstack.org/551370 | 15:43 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Annotate flows and handle PortBindingDeletionFailed in ComputeManager https://review.openstack.org/551371 | 15:43 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Port binding based on events during live migration https://review.openstack.org/434870 | 15:43 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: conductor: use port binding extended API in during live migrate https://review.openstack.org/522537 | 15:43 |
| *** tssurya has quit IRC | 15:47 | |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: network: Always retrieve network information if available https://review.openstack.org/564444 | 15:51 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: network: Retrieve tunneled status in '_get_physnet_info' https://review.openstack.org/564445 | 15:51 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Add NUMACell.network_metadata https://review.openstack.org/564439 | 15:51 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Add NUMATopologyLimits.networks https://review.openstack.org/575486 | 15:51 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: hardware: Start accounting for networks in NUMA placement https://review.openstack.org/564448 | 15:51 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Add RequestSpec.network_metadata https://review.openstack.org/564442 | 15:51 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: scheduler: Start utilizing RequestSpec.network_metadata https://review.openstack.org/564452 | 15:51 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Add '[neutron] physnets' and related options https://review.openstack.org/564440 | 15:51 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Start populating NUMACell.network_metadata field https://review.openstack.org/564441 | 15:51 |
| *** nicolasbock has joined #openstack-placement | 15:52 | |
| jaypipes | guh. one letter typo in a variable name stumped me for the last hour... :( | 16:45 |
| openstackgerrit | Jay Pipes proposed openstack/nova master: move lookup of provider from _new_allocations() https://review.openstack.org/579920 | 16:52 |
| openstackgerrit | Jay Pipes proposed openstack/nova master: placement: delete auto-created consumers on fail https://review.openstack.org/579921 | 16:52 |
| jaypipes | efried, mriedem: ^^ | 16:52 |
| jaypipes | mriedem: related to the bug about not being able to change user/project for a consumer. working on that one now... | 16:53 |
| efried | ack | 16:53 |
| openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Delete orphan nodes before updating resources https://review.openstack.org/579922 | 16:55 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Default embedded instance.flavor.disabled attribute https://review.openstack.org/579925 | 17:05 |
| *** e0ne has joined #openstack-placement | 17:45 | |
| openstackgerrit | Merged openstack/nova master: Merge server create schema for multiple create extension https://review.openstack.org/578338 | 18:11 |
| openstackgerrit | Merged openstack/nova master: Merge server create schema for scheduler hint extension https://review.openstack.org/578343 | 18:11 |
| *** cdent has joined #openstack-placement | 18:13 | |
| *** edmondsw has joined #openstack-placement | 18:27 | |
| *** mriedem has quit IRC | 18:31 | |
| *** edmondsw has quit IRC | 18:32 | |
| *** mriedem has joined #openstack-placement | 18:39 | |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Add action initiator attribute to the instance payload https://review.openstack.org/536243 | 18:45 |
| openstackgerrit | Merged openstack/nova master: conf: Correct documentation for '[pci] passthrough_whitelist' https://review.openstack.org/552874 | 18:59 |
| openstackgerrit | Merged openstack/nova master: [doc] enhance admin/configuration/api.rst https://review.openstack.org/566008 | 18:59 |
| openstackgerrit | Merged openstack/nova master: Remove mox in sec group test and functional tests https://review.openstack.org/576751 | 18:59 |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Time how long pre_live_migration() takes https://review.openstack.org/574481 | 19:00 |
| openstackgerrit | Merged openstack/nova master: conf: Resolve Sphinx errors https://review.openstack.org/579545 | 19:02 |
| *** tssurya has joined #openstack-placement | 19:15 | |
| openstackgerrit | Andreas Karis proposed openstack/nova master: Add documentation for emulator threads policy https://review.openstack.org/579956 | 19:18 |
| openstackgerrit | Andreas Karis proposed openstack/nova master: Add documentation for emulator threads policy https://review.openstack.org/579956 | 19:22 |
| openstackgerrit | Andreas Karis proposed openstack/nova master: Add documentation for emulator threads policy https://review.openstack.org/579956 | 19:24 |
| openstackgerrit | Andreas Karis proposed openstack/nova master: Add documentation for emulator threads policy https://review.openstack.org/579956 | 19:24 |
| *** e0ne has quit IRC | 19:40 | |
| efried | cdent: Let's move here so we're not interrupting. Whichever thing we're talking about, placement is heavily involved. | 19:42 |
| efried | cdent: One thing I'm trying to work on is https://bugs.launchpad.net/nova/+bug/1779931 which is the race exposed by https://review.openstack.org/#/c/556669/ which you've already looked at a week or two ago. | 19:43 |
| openstack | Launchpad bug 1779931 in OpenStack Compute (nova) "Provider update race between host aggregate sync and resource tracker" [Undecided,New] | 19:43 |
| * cdent nods | 19:43 | |
| efried | cdent: But the thing about rebuilds not handling sharing providers has to do with karimull's attempts to write functional tests per my comments at https://review.openstack.org/#/c/569498/9/nova/conductor/manager.py@1037 | 19:44 |
| efried | I suspect we're going to have to merge that patch without those tests, because there's probably no way to make those tests pass, because of the aforementioned not handling sharing providers in rebuild (and related) flows. | 19:45 |
| openstackgerrit | Andreas Karis proposed openstack/nova master: Add documentation for emulator threads policy https://review.openstack.org/579956 | 19:45 |
| * cdent nods at efried | 19:48 | |
| cdent | I guess we have rather a lot of auditing to do, just to make sure we know where the gaps are | 19:49 |
| openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Tighten up ReportClient use of generation https://review.openstack.org/556669 | 19:51 |
| efried | yeah | 19:52 |
| *** edmondsw has joined #openstack-placement | 20:16 | |
| *** edmondsw has quit IRC | 20:21 | |
| mriedem | jaypipes: comments inline https://review.openstack.org/#/c/579921/ | 20:49 |
| mriedem | mostly ok with it though | 20:49 |
| jaypipes | mriedem, edleafe, cdent, dansmith, efried: for https://review.openstack.org/#/c/579654/ and https://review.openstack.org/#/c/579163/, you've asked for microversions for both of those changes. Would you all be amenable to a single microversion bump that included the behaviour changes for both of those patches? (i.e. merging those patches into one and adding the microversion changes) | 21:35 |
| efried | jaypipes: Sight unseen, if this is fixing the holes we have in consumer reporting, I'm in favor in principle. Let me look closer, though.. | 21:36 |
| jaypipes | mriedem, edleafe, cdent, dansmith, efried: we do need to make a decision on whether consumer records should be kept even after all a consumer's allocation records are deleted. My personal preference is that, yes, they should. I understand there are differing opinions on that question. Shall we take a vote on it and move forward? | 21:36 |
| cdent | jaypipes: yeah, same as efried: without yet looking it makes sense to fix all the consumer stuff in one microversion if possible | 21:36 |
| dansmith | +1 for one microversion | 21:36 |
| dansmith | +1 for auto-delete consumers | 21:37 |
| efried | jaypipes: Yes, having looked, I would be in favor; BUT we can't move on the second patch until we've decided on the consumer deletion thing. | 21:37 |
| jaypipes | efried: well, there's the one thing (deleting the auto-created consumer) that can be fixed without a microversion bump. that's the separate https://review.openstack.org/#/c/579921/ patch. | 21:37 |
| cdent | jaypipes: I'd like to be able to vote, but I still don't understand your reasons for not wanting to delete them? | 21:37 |
| efried | jaypipes: yeah, putting that totally aside, that's different | 21:37 |
| efried | cdent: It sounds like he's in favor of auto-deleting. | 21:37 |
| efried | oh, sorry, backwards | 21:38 |
| efried | "should be kept" | 21:38 |
| cdent | jaypipes: did you see my questions/comments on https://bugs.launchpad.net/nova/+bug/1779717 ? | 21:38 |
| openstack | Launchpad bug 1779717 in OpenStack Compute (nova) "No ability to update consumer's project and/or user external ID" [Medium,Triaged] - Assigned to Jay Pipes (jaypipes) | 21:38 |
| efried | Yeah, I'm opposed to that. My vote is to auto-delete when last allocation goes away. | 21:38 |
| jaypipes | cdent: I view consumers similar to resource providers. we don't delete resource providers when they have all their inventory deleted. yes, the resource provider has a name attribute to it, but a consumer has a project and user attribute to it. I just don't see a reason to delete that data. | 21:38 |
| edleafe | No need for more than 1 microversion | 21:39 |
| cdent | yeah, my comments on the bug are sort of related to that | 21:39 |
| cdent | brb | 21:39 |
| efried | jaypipes: We have an API to delete resource providers. If we add an API to delete consumers then I'm totally on board with no-auto-delete. | 21:39 |
| dansmith | efried: that's what I said, but I also don't want us to have to do that cleanup from nova, | 21:39 |
| efried | jaypipes: If consumers are like resource providers, let's make 'em like resource providers. | 21:39 |
| dansmith | and I see RPs as a top-level object in placement, but consumers are not.. they're incidental, and thus can be deleted incidentally | 21:40 |
| efried | dansmith: The cleanup would be easy. When we delete allocations, we also delete consumers. And with a DELETE /consumer/{u} operation, it'd be trivial to write a manage script that cleans up oldymoldys. | 21:40 |
| mriedem | we delete allocations everywhere in nova | 21:40 |
| mriedem | doubling the calls there isn't something i want to do | 21:41 |
| dansmith | totally doable, but it's another call, and with anything non-atmonic, we could fail in the middle | 21:41 |
| mriedem | it's metadata about the allocation, just drop it when the allocs are gone | 21:41 |
| dansmith | which means we have to be able to reap them later | 21:41 |
| cdent | mriedem++ | 21:41 |
| efried | Given the relative amount of work, my preference is auto-delete. | 21:41 |
| dansmith | cdent: what was your vote? | 21:41 |
| mriedem | auto-delete for me | 21:41 |
| efried | edleafe? | 21:41 |
| dansmith | right so that's 3 for auto delete, 4 for keep (and presumably never delete or add a way to delete) | 21:41 |
| dansmith | eff | 21:41 |
| dansmith | 3 for auto-delete, 1 for keep | 21:41 |
| * efried isn't sure if the votes are weighted, though | 21:42 | |
| edleafe | sure, if it has no meaning, delete it | 21:42 |
| cdent | dansmith: like I said above I'm still trying to understand jay's reasoning, but if it is just "i view them the same as resource providers" then: I don't (I think they are incidental) and I think they should auto-delete | 21:42 |
| edleafe | efried: they always are | 21:42 |
| dansmith | okay that's 5 for auto-delete, 1 for keep | 21:42 |
| dansmith | anyone else we should poll? | 21:42 |
| mriedem | the tc | 21:43 |
| * dansmith asks his cat | 21:43 | |
| mriedem | i guess cdent is the proxy | 21:43 |
| cdent | naw mate, the tc is powerless, we all know that | 21:43 |
| mriedem | my cat has found the afternoon sun and wants to be left alone | 21:43 |
| jaypipes | if we shelve-offload an instance, nova will delete all the allocations for that instance. that will delete the consumer in placement, then. we will then have no record that the instance ever consumed anything in placement. is that ok with everyone? | 21:43 |
| cdent | it is with me | 21:43 |
| mriedem | i am | 21:43 |
| edleafe | when it's unshelved, a new consumer record would be created, no? | 21:43 |
| mriedem | until there is a use case for why we need that record, auto-delete | 21:44 |
| mriedem | edleafe: yes | 21:44 |
| dansmith | just to be clear, we wouldn't have a record of anything other than that it consumed something at some point | 21:44 |
| jaypipes | edleafe: for the same instance, yes. with a new generation and a new created_by, etc | 21:44 |
| dansmith | which seems meaningless to me | 21:44 |
| jaypipes | created_at... | 21:44 |
| edleafe | so it's cleaner than having to have a separate code path for unshelve | 21:44 |
| dansmith | placement is not an audit log right? | 21:44 |
| mriedem | if we want audit trails or something, add notifications to placement | 21:44 |
| efried | No record other than the fact that it's a shelved instance. We can infer from that not only that it once had allocations, but what they were. We don't need an allocation-less entry in placement to tell us (the least interesting) part of that. IMO. | 21:44 |
| edleafe | please no | 21:44 |
| dansmith | created_at on a consumer record tells us nothing about what it consumed, or even if it did | 21:44 |
| jaypipes | dansmith: true | 21:45 |
| mriedem | edleafe: please no to what? | 21:45 |
| cdent | efried: I don't understand "not only that it once had allocations, but what they were" | 21:45 |
| edleafe | adding notifications | 21:46 |
| cdent | how can you do the what they were part? | 21:46 |
| efried | cdent: Based on the flavor of the shelved instance, nah? | 21:46 |
| dansmith | cdent: you can infer from the instance | 21:46 |
| efried | cdent: Oh, I guess we can't tell which providers they came from. | 21:46 |
| efried | Unless we save the request spec or something. | 21:46 |
| mriedem | i thought there were approved blueprints at one point for adding notifications to placement... | 21:46 |
| cdent | in which case why use anything in placement for that at all, nova has all that info already | 21:46 |
| mriedem | things like blazar would probably want that | 21:46 |
| jaypipes | I predict we're going to want GET|PUT|POST|DELETE /consumers at some point in the future with the move to use placement for quota usage and accounting... but I'm fine auto-deleting consumer records when the last allocation is removed if that's what folks want. | 21:46 |
| efried | cause that has the allocations on it. | 21:46 |
| efried | jaypipes: I don't think the two are mutually exclusive. | 21:47 |
| cdent | efried: "cause that", what is "that"? | 21:47 |
| mriedem | the provider | 21:47 |
| mriedem | the provider which had the allocations | 21:47 |
| efried | cdent: sorry, I was saying the request spec has allocations on it. Is that true, after we claim? | 21:47 |
| efried | anyway... | 21:48 |
| mriedem | jaypipes: yeah i commented in https://review.openstack.org/#/c/579163/ and i can only think of a top level set of APIs for consumers in placement if we fill out that resource more, for things like consumer types | 21:48 |
| mriedem | efried: i don't know what you're talking about wrt request spec | 21:48 |
| efried | yeah, ignore me. | 21:48 |
| mriedem | heh | 21:48 |
| mriedem | request spec has the flavor | 21:48 |
| mriedem | that's it | 21:48 |
| mriedem | well, that's it* | 21:49 |
| mriedem | * all of the other crap | 21:49 |
| efried | jaypipes: In the same microversion where you introduce /consumers, you flip the allocations ops to not automatically do *anything* with consumers. So in that microversion, you *must* PUT /consumer/{u} first; and when you're done, you *may* DELETE /consumer/{u}. Folks wanting the auto-create/delete behavior use the old microversion. Done. | 21:49 |
| cdent | please don't do that | 21:49 |
| efried | mriedem: I thought we were stuffing the allocation object on the request spec before we sent it to spawn. | 21:49 |
| mriedem | you added that code to compute right? | 21:50 |
| mriedem | i never looked at it | 21:50 |
| cdent | we still haven't sufficiently worked out the meaning of consumers _outside of nova_ to be creating them as first class entities in the api | 21:50 |
| mriedem | but allocations are sent to spawn() as a separate parameter from what i remember | 21:50 |
| openstackgerrit | Andreas Karis proposed openstack/nova master: Add documentation for emulator threads policy https://review.openstack.org/579956 | 21:50 |
| edleafe | wait - there's a world outside of nova?? | 21:50 |
| efried | mriedem: Yeah, would have to look again to see exactly where that plays. Anyway, I don't think it's important enough to worry about in this context. | 21:50 |
| dansmith | edleafe: no | 21:50 |
| mriedem | ovirt | 21:50 |
| mriedem | it's in the ML thread | 21:50 |
| efried | cdent: I'm just saying that, if we wind up eventually wanting GET|PUT|POST|DELETE /consumers as jaypipes suggests, it would make sense to go the whole way in that microversion. | 21:51 |
| efried | cdent: When we try to go partway, we end up in situations like this. | 21:51 |
| cdent | efried: I agree, which is why I'm certain we don't want to touch it now :) | 21:51 |
| mriedem | efried: allocations = self.reportclient.get_allocations_for_consumer( | 21:51 |
| mriedem | context, instance.uuid) | 21:51 |
| mriedem | we get them from placement and pass them to driver.spawn() | 21:52 |
| efried | mriedem: We also stuff 'em into `resources['allocations']` in the _build_resources context mgr in _build_and_run_instance. | 21:53 |
| mriedem | that's different | 21:53 |
| mriedem | that's just a yield trick | 21:53 |
| mriedem | same as bdms and network_inf | 21:53 |
| mriedem | *network_info | 21:53 |
| mriedem | anywho | 21:54 |
| cdent | anyway, can someone please respond or react to my comment on https://bugs.launchpad.net/nova/+bug/1779717 because I think that's key to including me in any future discussion on this stuff. until I can get past that, I'll be stuck | 21:54 |
| openstack | Launchpad bug 1779717 in OpenStack Compute (nova) "No ability to update consumer's project and/or user external ID" [Medium,Triaged] - Assigned to Jay Pipes (jaypipes) | 21:54 |
| mriedem | sounds like the majority has spoken right | 21:54 |
| cdent | yup, but it is ack'd as a "some issues still remain to figure out" | 21:54 |
| efried | oh, which we then throw away. | 21:54 |
| jaypipes | cdent: responding now | 21:54 |
| cdent | awesome, thanks jaypipes | 21:54 |
| mriedem | cdent: i would agree that we should only allow setting new project/user if the existing project/user is None or sentinels | 21:55 |
| mriedem | otherwise that's like changing ownership on stuff and we don't do that anywhere else in openstack | 21:55 |
| mriedem | despite people wanting that in the past | 21:55 |
| mriedem | i.e. transfer instance foo from project A to project B | 21:55 |
| efried | Where sentinels means the default values of the conf options? | 21:56 |
| jaypipes | cdent: responded. | 21:56 |
| efried | Cause it can't be the actual values of the conf options. | 21:56 |
| efried | cause we can't get at those from this side of the API. | 21:57 |
| cdent | jaypipes: cool, thanks, will read in just a sec, pushing out the tc report | 21:57 |
| jaypipes | mriedem: lol, jinx, that's pretty much what I just wrote on the bug :) | 21:57 |
| efried | I guess placement can tell. Assuming we didn't set those records at one point and then change the conf values later. | 21:57 |
| jaypipes | efried: we can whilst placement is part of nova. :) | 21:57 |
| *** tssurya has quit IRC | 21:58 | |
| cdent | why does the nova side need to know? | 21:58 |
| jaypipes | efried: or at least, we can fudge it and hope. | 21:58 |
| jaypipes | cdent: nova-manage heal_allocations. | 21:58 |
| jaypipes | I presume. | 21:58 |
| efried | jaypipes: Still not, because we're still on different sides of an API boundary, likely on different machines. | 21:58 |
| jaypipes | efried: like I said, fudge it and hope. ;) | 21:58 |
| mriedem | heal_allocations doesn't know about the sentinels, | 21:59 |
| efried | What would be the reason we shouldn't *always* change them to the instance's proj/user if not already? | 21:59 |
| mriedem | it just knows that the allocations project/user for this instance's allocations don't match the instance project/user | 21:59 |
| edleafe | for transfer, wouldn't it be simpler to DELETE allocs for A and then PUT allocs for B? | 21:59 |
| jaypipes | mriedem: ah, right, sorry. please ignore me. | 21:59 |
| edleafe | conceptually, that is | 21:59 |
| mriedem | https://review.openstack.org/#/c/574488/5/nova/cmd/manage.py@1843 | 21:59 |
| efried | edleafe: Yeah, that'd be a way of getting around the limitation, without a microversion. Just POST both of those payloads, done. | 21:59 |
| *** nicolasbock has quit IRC | 22:00 | |
| mriedem | the one worry i have is what efried said - if the config values change, then we're stuck | 22:00 |
| jaypipes | efried: of course, that won't work until https://bugs.launchpad.net/nova/+bug/1779717 is fixed, but that's besides the point. | 22:00 |
| openstack | Launchpad bug 1779717 in OpenStack Compute (nova) "No ability to update consumer's project and/or user external ID" [Medium,Triaged] - Assigned to Jay Pipes (jaypipes) | 22:00 |
| efried | Personally I don't see a problem from a *placement* perspective allowing edit of a consumer's proj/user. That may not fly in nova, but why not in placement? | 22:00 |
| edleafe | efried: nothing to do with microversion avoidance. Consumer A is no longer consuming, so delete its allocs (and itself). Then allocate (and create) consumer B | 22:00 |
| jaypipes | efried: it's the same consumer. it's just that consumer's ownership is being transferred. | 22:01 |
| efried | edleafe: In one step, though, so we don't race and wind up unallocated. | 22:01 |
| jaypipes | edleafe: sorry, that last one was directed at you. | 22:01 |
| jaypipes | edleafe: last one I directed at efried that is.. | 22:01 |
| mriedem | i think edleafe is saying you can workaround the bug by just deleting the allocations (with the sentinel) and re-creating them (with the instance info) - but it's not atomic | 22:02 |
| jaypipes | edleafe: in other words, the instance isn't changing its consumption of resources. it's that the ownership of the instance is being transferred to another project. | 22:02 |
| jaypipes | mriedem: no, it would be atomic (if using POST /allocations). it's just a lot of effort to change the user/project of a consumer :) | 22:02 |
| efried | mriedem: it's atomic if you use POST allocations | 22:03 |
| *** avolkov has quit IRC | 22:03 | |
| efried | yeah | 22:03 |
| efried | not so much effort. | 22:03 |
| mriedem | ok then yeah POST /allocations doesn't work until that bug is fixed | 22:03 |
| efried | Just wrapping the existing payload you've created in [{}, {here}] | 22:04 |
| jaypipes | efried: well... it would be atomic with regards to the allocation... since consumer objects are created outside of the AllocationList.create_all() transaction currently, there's still some holes (like the ones being plugged up with the auto-created consumer deletion) | 22:04 |
| mriedem | maybe we want that 1.13 osc-placement patch after all... | 22:04 |
| edleafe | We already have the atomic switcheroo in place; we just need to adjust it for handling migrations to handling ownership changes | 22:04 |
| openstackgerrit | Andreas Karis proposed openstack/nova master: Add documentation for emulator threads policy https://review.openstack.org/579956 | 22:04 |
| *** edmondsw has joined #openstack-placement | 22:05 | |
| mriedem | coincidentally, atomic switcheroo was my band in high school | 22:05 |
| efried | never mind, POST doesn't help you, cause it's keyed by consumer ID. | 22:05 |
| edleafe | I have all your CDs | 22:05 |
| jaypipes | my high school band was called Slow Children at Play. true story. | 22:06 |
| cdent | I was never in a band. I bet this surprises no one. | 22:07 |
| edleafe | I was, but much more boring. Mine was called Mono | 22:07 |
| cdent | Can I go to bed, or are you all going to make decisions tonight? | 22:07 |
| cdent | same note. over and over. | 22:07 |
| edleafe | It's after beer o'clock on the day before a holiday. I'm outta here | 22:07 |
| jaypipes | cdent: I believe the decision has been made? auto-delete consumer records when no allocs. | 22:08 |
| cdent | I meant additional decisions ... | 22:08 |
| cdent | but yeah, if we've reached that decision I think I can safely retire | 22:09 |
| jaypipes | ++ ciao | 22:09 |
| *** edmondsw has quit IRC | 22:09 | |
| efried | Are we in favor of combining that auto-deletion with the 404 behavior? | 22:11 |
| mriedem | i'm on a call from 6:30-8 so i'll be around making major decisions | 22:11 |
| mriedem | nvm, just cancelled, smell you jerks later | 22:14 |
| *** mriedem is now known as mriedem_explosio | 22:14 | |
| *** mriedem_explosio is now known as mriedem_burnmark | 22:14 | |
| *** cdent has quit IRC | 22:23 | |
| openstackgerrit | Jay Pipes proposed openstack/nova master: move lookup of provider from _new_allocations() https://review.openstack.org/579920 | 22:41 |
| openstackgerrit | Jay Pipes proposed openstack/nova master: placement: delete auto-created consumers on fail https://review.openstack.org/579921 | 22:41 |
| *** mriedem_burnmark has quit IRC | 23:30 | |
| *** edmondsw has joined #openstack-placement | 23:53 | |
| *** edmondsw has quit IRC | 23:57 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!