| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-05-03 | |||
| 16:41:32 | mdbooth | Oh, weird. | |
| 16:41:36 | mdbooth | I’ll put it back | |
| 16:42:07 | imacdonn | I see it appearing now ;) | |
| 16:48:48 | openstackgerrit | garyk proposed openstack/nova master: Improve metadata performance https://review.opendev.org/615435 | |
| 16:51:03 | wwriverrat | Likely asked and answered... (sorry)... Is there a new channel out there for placement? | |
| 16:51:33 | wwriverrat | Or do the questions still fall here? | |
| 16:51:48 | gibi_cape | wwriverrat: #openstack-placement | |
| 16:52:43 | mriedem | dansmith: mdbooth: question on that libvirt image type patch - https://review.opendev.org/#/c/653454/ - i need a decoder ring for the ternary config option of doom | |
| 17:08:08 | mriedem | tssurya: i see this is incomplete for ironic https://github.com/openstack/nova/blob/b9bcbab86b8314fbaaeb2d2af6282d4a612aeb8d/nova/api/openstack/compute/services.py#L270 | |
| 17:09:45 | mriedem | tssurya: because it only deletes the provider for the first compute node | |
| 17:09:50 | mriedem | but there could be 1000 | |
| 17:10:27 | mriedem | https://github.com/openstack/nova/blob/b9bcbab86b8314fbaaeb2d2af6282d4a612aeb8d/nova/objects/service.py#L313 | |
| 17:12:28 | mriedem | jaypipes: i'm assuming the answer is yes here, but will this model_query soft_delete delete all records if there is more than one compute node per service (ironic): https://github.com/openstack/nova/blob/b9bcbab86b8314fbaaeb2d2af6282d4a612aeb8d/nova/db/sqlalchemy/api.py#L393 | |
| 17:46:52 | mriedem | efried: there is an open runway slot but the next in the queue is gibi's bug for the port requested resources heal allocation stuff, so i'm not sure if you want to allow that since it's not technically a blueprint, or what | |
| 17:47:29 | mriedem | efried: also, i nixed the "Fixing OSC's live migrate interface" entry from the saturday agenda since the recap on that is in the ML - just looking for owners for the identified work items | |
| 17:47:33 | mriedem | artom already signed up for one | |
| 17:48:26 | efried | mriedem: I've never been one to be strict about must-be-blueprint-to-be-in-runway, so gibi_cape's thing would be fine by me. | |
| 17:55:17 | openstackgerrit | Dan Smith proposed openstack/nova master: Make libvirt expose supported image types https://review.opendev.org/653454 | |
| 17:55:18 | openstackgerrit | Dan Smith proposed openstack/nova master: Add vmware driver image type capabilities https://review.opendev.org/655730 | |
| 17:55:18 | openstackgerrit | Dan Smith proposed openstack/nova master: Add ironic driver image type capabilities https://review.opendev.org/655729 | |
| 17:55:19 | openstackgerrit | Dan Smith proposed openstack/nova master: Add xenapi driver image type capabilities https://review.opendev.org/655731 | |
| 17:55:19 | openstackgerrit | Dan Smith proposed openstack/nova master: Add zvm driver image type capabilities https://review.opendev.org/655732 | |
| 17:55:20 | openstackgerrit | Dan Smith proposed openstack/nova master: Enable image type query support in nova-next https://review.opendev.org/656903 | |
| 17:55:20 | openstackgerrit | Dan Smith proposed openstack/nova master: Add image type request filter https://review.opendev.org/656413 | |
| 18:00:33 | openstackgerrit | Dan Smith proposed openstack/nova master: WIP: Make libvirt able to trigger a backend image copy when needed https://review.opendev.org/656998 | |
| 18:31:23 | openstackgerrit | Merged openstack/os-traits master: Dropping the py35 testing https://review.opendev.org/654651 | |
| 19:28:56 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Delete resource providers for all nodes when deleting compute service https://review.opendev.org/657016 | |
| 19:31:47 | mriedem | tssurya: ^ | |
| 19:33:14 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Delete resource providers for all nodes when deleting compute service https://review.opendev.org/657016 | |
| 19:35:34 | tssurya | mriedem: yea forgot we didn't handle the 1:N case for ironic | |
| 19:42:30 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Avoid unnecessary joins in delete_resource_provider https://review.opendev.org/657021 | |
| 19:48:06 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Change the default of notification_format to unversioned https://review.opendev.org/603079 | |
| 19:51:33 | cdent | mriedem: do you know which provider is limestone-regionone? | |
| 19:53:25 | mriedem | cdent: limestone? https://www.limestonenetworks.com/wordpress/tag/openstack-cloud/ | |
| 19:53:37 | mriedem | https://www.limestonenetworks.com/blog/2018/12/10/openstack-public-cloud-at-limestone-networks/ | |
| 19:53:54 | cdent | ah, how incredibly obvious. I blame that giant sugar soaked cookie at lunch | |
| 19:54:13 | cdent | they've timed out some jobs on the placement deletion | |
| 19:55:47 | cdent | i seem to recall that region coming up fairly often | |
| 19:56:23 | mriedem | gibi_cape: what's the change of heart in https://review.opendev.org/#/c/603079/ ? | |
| 19:59:34 | gibi_cape | mriedem: I guess I accepted that we cannot make our notification consumers to change to versioned. | |
| 20:00:07 | gibi_cape | mriedem: and emiting both notification by default is not logical as most of the consumers consume either legacy or versioned but noth both | |
| 20:06:07 | openstackgerrit | Dan Smith proposed openstack/nova master: Add docs for image type support request filter https://review.opendev.org/657025 | |
| 20:08:29 | mriedem | gibi_cape: ah, did that come up during your talk? | |
| 20:08:39 | mriedem | i haven't seen the recording yet (probably not posted) | |
| 20:13:09 | tssurya | gibi_cape: for the record I don't think you should do this upstream as a default, if deployments are annoyed with it they can just set it right ? | |
| 20:13:51 | mriedem | tssurya: most projects consuming nova notifications aren't consuming versioned notifications, right? | |
| 20:14:05 | mriedem | so if the default is versioned, most deployments have to change, | |
| 20:14:13 | tssurya | mriedem: yea they are not | |
| 20:14:21 | mriedem | if the default is both (current), most deployments have extra traffic/performance hit of generating an unused versioned notification | |
| 20:14:29 | mriedem | hence the bug | |
| 20:14:44 | tssurya | but if we have put in a lot of work to get it done we have it available, if they want they can set it in the config | |
| 20:14:52 | tssurya | yea i know the bug :) was there was for his presentation | |
| 20:15:20 | mriedem | i'd argue that we (nova) should work on getting versioned notification support in other projects in openstack | |
| 20:15:24 | mriedem | if we want people using it | |
| 20:15:35 | mriedem | once that is done, then change the default | |
| 20:16:10 | bauzas | uh, my cellsv2 skills are rusty | |
| 20:16:12 | tssurya | sure :) that effort has been going on I guess. I agree, so I think we are saying the same thing right ? keep the default as both for now ? | |
| 20:16:33 | bauzas | I wonder whether the migration.execute() code is run under the super-cond or the cond | |
| 20:16:37 | bauzas | I guess the latter | |
| 20:16:45 | bauzas | since we used the cell MQ | |
| 20:16:54 | mriedem | bauzas: what migration.execute() code? | |
| 20:17:03 | mriedem | tssurya: no i'm not saying that | |
| 20:17:05 | bauzas | mriedem: the conductor task | |
| 20:17:16 | mriedem | "if the default is both (current), most deployments have extra traffic/performance hit of generating an unused versioned notification" | |
| 20:17:26 | tssurya | okay so you want the default to be unversioned and then move when ll projects have moved to both ? | |
| 20:17:32 | bauzas | oh shit | |
| 20:17:34 | bauzas | nevermind | |
| 20:17:39 | bauzas | it calls the scheduler | |
| 20:17:42 | bauzas | so it's super-cond | |
| 20:17:45 | bauzas | silly question it is | |
| 20:17:49 | tssurya | all* | |
| 20:18:02 | mriedem | bauzas: conductor task will be targeted to the source cell | |
| 20:18:07 | mriedem | the context that is | |
| 20:18:18 | bauzas | mmm ? | |
| 20:18:20 | mriedem | i.e. @targes_cell | |
| 20:18:26 | mriedem | *@targets_cell | |
| 20:18:26 | bauzas | oh that yeah | |
| 20:20:09 | bauzas | just to make it clear, this section is run by super-cond (IMO) https://github.com/openstack/nova/blob/master/nova/conductor/tasks/migrate.py#L187 | |
| 20:20:16 | bauzas | we don't target the cell yet | |
| 20:21:05 | mriedem | um | |
| 20:21:16 | mriedem | well resize can reschedule and run on a conductor in the cell | |
| 20:22:10 | mriedem | bauzas: whether or not that works depends on your cell conductor config being able to hit the api db https://bugs.launchpad.net/nova/+bug/1781286 | |
| 20:22:11 | openstack | Launchpad bug 1781286 in OpenStack Compute (nova) "CantStartEngineError in cell conductor during reschedule - get_host_availability_zone up-call" [Medium,Triaged] | |
| 20:22:12 | bauzas | I'm confused then, but nevermind, I'll just take my devstack and try | |
| 20:22:29 | mriedem | bauzas: devstack will fail on ^ during a resize reschedule | |
| 20:22:41 | mriedem | if there is no reschedule, then yes it's just super-conductor | |
| 20:23:12 | bauzas | ok, the context is just a resize | |
| 20:23:25 | bauzas | and yeah, gotcha | |
| 20:26:44 | gibi_cape | mriedem, tssurya: I don't think we have manpower to transform ceilometer to versioned notifications. | |
| 20:29:45 | tssurya | gibi_cape: agreed i'm just saying changing this to unversioned by default basically means no one will take efforts to move maybe, I do understand the bug scenario and perf issues which is why we have the workaround option for the deployments to set it to unversioned, I wasn't just sure if we should go out and make that the default | |
| 20:30:20 | tssurya | but yea its your call :) | |
| 20:30:31 | openstackgerrit | Merged openstack/nova master: Trivial: use default value in next() func https://review.opendev.org/653362 | |
| 20:33:31 | gibi_cape | tssurya: for me it is OK to keep it 'both' and keep the perf bug open. What I cannot do is to promise I will change ceilometer over to versioned | |
| 20:33:56 | tssurya | gibi_cape: fair enough | |
| 20:34:07 | mriedem | it'd be interesting to know if deployers (1) don't even notice the change or (2) noticed and changed to unversioned | |
| 20:34:12 | mriedem | i know godaddy found out and changed it | |
| 20:34:21 | mriedem | since they are heavy on notifications for internal stuff | |
| 20:34:37 | mriedem | by "the change" i mean to 'both' long ago | |
| 20:34:51 | mriedem | there is a guy 2 over from me from godaddy so he'd probably know better | |
| 20:43:34 | openstackgerrit | Merged openstack/nova stable/stein: Fix help for ironic.peer_list config https://review.opendev.org/656632 | |