Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-10-10
14:50:16 openstackgerrit Merged openstack/nova master: Only query BDMs once in API during rebuild https://review.openstack.org/509574
15:05:53 abhi89 Hi all.. can any stable core please review https://review.openstack.org/#/c/505546/.. waiting for one more +2
15:34:11 openstackgerrit Merged openstack/nova master: Transform aggregate.add_host notification https://review.openstack.org/396210
15:38:32 dansmith jaypipes: when you're done posting your nrp set, maybe you could vet all the db fu in here: https://review.openstack.org/#/c/500410/12
15:38:40 dansmith it'd make it easier for me to just stamp it if you did
15:38:53 jaypipes ack
15:39:03 jaypipes I'll do it on the next test run
15:39:20 dansmith cool
15:40:34 thingee mriedem: re forum topics I'll send an update.
15:42:46 dansmith melwitt: are you respinning the bottom change for that notification target fix?
15:43:17 melwitt dansmith: yeah, figured it's best to have the comment explain what's going on since gibi figured it out
15:43:26 dansmith yep, just wanted to know if I should +W or not
15:43:31 melwitt instead of the "duh I dunno why this doesn't work"
15:43:47 melwitt yeah, k. one sec
15:44:22 openstackgerrit melanie witt proposed openstack/nova master: Target context for build notification in conductor https://review.openstack.org/509967
15:44:23 openstackgerrit melanie witt proposed openstack/nova master: Fix target_cell usage for scatter_gather_cells https://review.openstack.org/510691
15:44:23 openstackgerrit melanie witt proposed openstack/nova master: Elevate existing RequestContext to get bandwidth usage https://review.openstack.org/509968
15:44:24 openstackgerrit melanie witt proposed openstack/nova master: Make expected notifications output easier to read in tests https://review.openstack.org/510823
15:44:24 openstackgerrit melanie witt proposed openstack/nova master: Improve the CellDatabases test fixture and usage https://review.openstack.org/508432
15:44:52 openstackgerrit sean mooney proposed openstack/nova-specs master: Use neutron's new port binding API https://review.openstack.org/375580
15:47:42 mriedem johnthetubaguy: i also went through ^ https://review.openstack.org/#/c/375580/ yesterday, and i think when you originally wrote that it was before we had a plan for how to do the new style volume attachment stuff, but i think we'll want to do similar with the port bindings API
15:48:11 mriedem i.e. turn it on once everything is new enough to use it and then we can key move operations off whether or not the port was attached using the port binding API
15:52:21 jaypipes dansmith, cdent: remind me again why this sentence is true: "have an AllocationList from multiple project ids, something we want to be able to do with the new use of migration uuids."
15:52:51 dansmith jaypipes: it's not because of the migration change,
15:52:53 jaypipes dansmith, cdent: how does using a migration UUID as a consumer mean we need allocations with >1 user or project?
15:53:11 jaypipes oh, ok
15:53:13 dansmith jaypipes: we just said that if we're making this change, we might as well let you swap allocations atomically, not just within one user/project
15:53:19 jaypipes got it.
15:53:23 jaypipes ok, thanks
15:53:49 openstackgerrit Eric Fried proposed openstack/nova master: Use ksa adapter for placement conf & requests https://review.openstack.org/492247
15:54:18 dansmith this needs +W, which is just minor feedback rollup from the earlier set of patches: https://review.openstack.org/#/c/506416
15:54:27 dansmith in case anyone needs stats padding
15:55:40 dansmith melwitt: in case you haven't seen, these three patches further improve listing: https://review.openstack.org/#/c/508595/7
15:56:02 dansmith including BRs in the merge, so that they sort normally instead of the weird "always first" behavior
15:56:03 dansmith and
15:56:18 dansmith we only scatter to cells that you have instances in
15:57:31 jaypipes dansmith: https://review.openstack.org/#/c/506416/ +W'd
15:57:52 dansmith jaypipes: thanks
15:58:26 melwitt dansmith: a-ha, cool. thanks for the heads up, will review
16:08:19 openstackgerrit sean mooney proposed openstack/nova-specs master: Use neutron's new port binding API https://review.openstack.org/375580
16:14:54 openstackgerrit Jackie Truong proposed openstack/nova master: Add trusted_certs object https://review.openstack.org/489408
16:18:25 mriedem oh yay we regressed volume detach API semantics in pike
16:18:54 sean-k-mooney mriedem: oh how so?
16:18:59 mriedem we introduced a race
16:19:12 mriedem by detaching the volume in cinder before deleting the bdm in nova
16:19:22 mriedem it changes the behavior of the volume attachments API
16:19:31 mriedem and tempest relies on that behavior
16:19:42 mriedem i.e. tempest handles a 404 differently from a 400 when detaching a volume
16:20:03 sean-k-mooney so is that why my instaces that i booted from volume form image with delete volume on delete instance nolonger deletes the volume
16:20:05 mriedem arguably it could be a tempest bug or bad client expectations
16:20:24 mriedem that sounds different
16:20:39 mriedem that's a delete_on_termination thing
16:20:52 sean-k-mooney im not sure if its intended or not as i only noticed it happening when the instance goes to error state and i delete it
16:22:21 mriedem error state during scheduling?
16:22:24 mriedem or during building on the host?
16:22:24 sean-k-mooney mriedem: well tempest is ment to work across releases so if this changes across release it an api change
16:22:48 mriedem sean-k-mooney: well, in this case, tempest is requesting to detach a volume that is already detached
16:22:52 mriedem it's part of a cleanup routine
16:22:55 sean-k-mooney mriedem: durign building basically when the instance fails to boot due to port binding failed
16:23:02 mriedem arguably the client shouldn't be making that request
16:23:39 sean-k-mooney mriedem: ah ok ya that sound like a client bug as the precondition is not ment. e.g. it must be attached to detach it
16:23:41 mriedem sean-k-mooney: are you sure you're setting delete_on_termination=True for the bdm?
16:24:41 openstackgerrit Jackie Truong proposed openstack/nova master: Implement certificate_utils https://review.openstack.org/479949
16:24:48 efried mriedem Should removal of cinder v2 support also have removed volumev2 (and maybe also volume) from https://github.com/openstack/nova/blob/master/nova/context.py#L121 ?
16:24:50 sean-k-mooney i was booting from horizon and setting the Delete volumne on onstance delete option
16:25:01 sean-k-mooney so i assume that set it under the hood
16:25:58 mriedem efried: good point, probably yeah
16:26:05 mriedem did we remove volumev2 support already?
16:26:08 mriedem i think i pushed the change for it...
16:26:11 efried yeah
16:26:16 efried swhy I tagged ya
16:26:41 mriedem ah yeah https://github.com/openstack/nova/commit/eadbacbda628ecc969a980378faf55bc02f514bf#diff-d24869e36361bb430acd67658a83b0b5
16:26:47 mriedem yeah we should remove volume and volumev2 from context
16:27:11 efried mriedem But include block-storage
16:27:23 efried since that's the official service type name
16:27:38 efried though I don't think devstack is setting that up yet.
16:29:07 mriedem lyarwood: interesting fallout bug https://bugs.launchpad.net/tempest/+bug/1722577/comments/3
16:29:08 openstack Launchpad bug 1722577 in OpenStack Compute (nova) "test_list_get_volume_attachments failing with 400 error on teardown when detaching an already detached volume" [Undecided,Triaged]
16:29:09 mriedem mtreinish: ^
16:29:24 mriedem mtreinish: i'm inclined to fix ^ in tempest
16:30:42 mriedem lyarwood: result of this change https://review.openstack.org/#/c/440693/
16:45:32 lyarwood mriedem: huh, nice find, I guess I missed that detach_volume is cast at the compute layer so anything following the detach api call could race
16:45:43 lyarwood is a cast*
16:46:32 mriedem yeah, seemed innocent enough. i'm working a tempest patch.
16:46:50 mriedem cinder team is also considering just not returning a 400 if you ask to detach an already detached volume
16:47:54 sean-k-mooney mriedem: would they retrun a 200 instead in that case?
16:48:18 mriedem don't know
16:49:18 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/os-vif master: Add Virtual Ethernet device pair https://review.openstack.org/484726
16:49:59 openstackgerrit Balazs Gibizer proposed openstack/nova master: fix cleaning up evacuated instances https://review.openstack.org/510938
16:50:17 cfriesen sean-k-mooney: I think their normal success code is 202
16:51:09 sean-k-mooney cfriesen: 202 accepted would also be fine it they match the retrun code that would be retruned if it did detach then that would be fine
16:52:18 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/os-vif master: Add Open vSwitch patch port https://review.openstack.org/485228
16:52:39 mriedem the cinder detach api returns a 202 even though it's a synchronous api
16:52:47 mriedem it's an rpc call to the volume backend
16:52:48 openstackgerrit Balazs Gibizer proposed openstack/nova master: Reproduce bug 1721652 in the functional test env https://review.openstack.org/510176
16:52:49 openstack bug 1721652 in OpenStack Compute (nova) "Evacuate cleanup fails at _delete_allocation_for_moved_instance" [High,In progress] https://launchpad.net/bugs/1721652 - Assigned to Balazs Gibizer (balazs-gibizer)
16:52:49 openstackgerrit Balazs Gibizer proposed openstack/nova master: fix cleaning up evacuated instances https://review.openstack.org/510938
16:56:38 openstackgerrit Brianna Poulos proposed openstack/nova master: Add trusted_certs object https://review.openstack.org/489408
16:56:39 openstackgerrit Brianna Poulos proposed openstack/nova master: Add trusted_image_certificates to REST API https://review.openstack.org/486204
17:03:40 openstackgerrit Brianna Poulos proposed openstack/nova master: Implement certificate_utils https://review.openstack.org/479949

Earlier   Later