Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-03-08
20:13:23 cfriesen currently I'd expect the scheduler to trigger it to move in that case
20:16:13 cfriesen anyways, reported it as a bug: https://bugs.launchpad.net/nova/+bug/1819216
20:16:13 openstack Launchpad bug 1819216 in OpenStack Compute (nova) "in devstack, "nova migrate <uuid>" will try to migrate to the same host (and then fail)" [Undecided,New]
20:19:16 mriedem did someone say resize to same host problems?! https://review.openstack.org/#/c/638791/
20:19:31 mriedem if you have servers in an affinity group, you can *only* resize to the same ohst
20:19:32 mriedem *host
20:19:58 mriedem cfriesen: there is also a bug for the thing you hit, semi-related anyway
20:20:38 cfriesen mriedem: right, I knew about the affinity group
20:20:51 mriedem the api says resize to same host, sure, but the compute says 'no way'
20:22:04 mriedem https://bugs.launchpad.net/nova/+bug/1811235
20:22:05 openstack Launchpad bug 1811235 in OpenStack Compute (nova) "instance's vm-state becomes error when cold-migrate instance to same host failed" [Medium,In progress] - Assigned to Matt Riedemann (mriedem)
20:23:07 mriedem the api check is here https://github.com/openstack/nova/blob/master/nova/compute/api.py#L3626
20:23:43 mriedem the compute check is here https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L4287
20:24:04 mriedem which fails for everything except vmware
20:24:23 mriedem since vmware is a cluster and you can migrate within the same "host"
20:25:11 mriedem so allow_resize_to_same_host is true if you're doing a resize, but not a cold migration *unless* you've got vcenter
20:26:10 cfriesen arguably the scheduler should only allow cold migration to same host for vcenter
20:26:22 mriedem the scheduler doesn't know if it's vcenter
20:26:23 cfriesen I guess it's actually the conductor
20:26:27 mriedem or conductor
20:26:39 mriedem the control plane services are not supposed to know anything about the underlying compute driver
20:26:43 cfriesen can't we look at host capabilities or hypervisor type or something?
20:27:02 dansmith cfriesen: and make assumptions? what about vmware 2 which *can* do that thing.. it's a slippery slope
20:27:07 dansmith separation of responsibilities
20:27:11 cdent vmware even has a downstream patch which adds config for "always resize to same "host" "
20:28:19 mriedem cdent: that's kind of where https://review.openstack.org/#/c/638791/ is going
20:28:28 cfriesen how do we handle it for live migration?
20:28:50 mriedem we always ignore the source host
20:29:01 mriedem same for evac
20:29:18 cfriesen seems weird to make live migration and cold migration behave differently
20:31:47 mriedem so,
20:31:53 mriedem now that we report compute driver capabilities as traits,
20:31:56 mriedem and supports_migrate_to_same_host is a driver capability,
20:32:20 mriedem we could technically check the compute node provider traits for the source compute node from the api and determind if it supports cold migrate to the same host and filter the request spec appropriately
20:32:46 dansmith don't you need the opposite of that?
20:33:10 dansmith you need to know if it *needs* the same host for the vmware case, not whether not it *can* yeah?
20:33:57 mriedem yeah maybe, you can't cold migrate across clusters
20:33:59 mriedem right?
20:34:08 dansmith yeah
20:34:26 dansmith so you could exclude the source host unless can-haz-must-be-same-host
20:35:34 dansmith that means looking up traits for the current host and then deciding what to ask the scheduler to do
20:35:48 mriedem yeah that's what i was trying to get to
20:36:09 mriedem from the api, if supports_migrate_to_same_host is false, exclude the source host before calling the scheduler
20:36:16 mriedem if supports_migrate_to_same_host is true, then don't exclude the source host
20:36:17 mriedem right?
20:36:38 dansmith you want libvirt to say it *can't* ?
20:36:43 mriedem it already does
20:36:50 mriedem vmware is the only driver that says supports_migrate_to_same_host=true
20:37:09 dansmith oh okay I'm conflating the migrate and resize to same host cases I guess
20:37:21 dansmith in that case,
20:37:25 dansmith the flag is really backwards,
20:37:26 mriedem yeah we don't have a driver capability for resize to same host
20:37:39 dansmith because you don't want to allow the source if it can, you want to limit to the source if it can, else exclude
20:37:42 mriedem you mean supports_migrate_to_same_host should be "must_migrate_to_same_host"?
20:37:49 mriedem oh
20:37:50 dansmith mriedem: right we have the config flag tho
20:38:24 mriedem yeah again that's kind of where i was going with https://review.openstack.org/#/c/638791/1/nova/conf/workarounds.py
20:38:29 mriedem but it's fugly
20:39:40 mriedem i need to work on an alternative to that which doesn't bypass the scheduler filters
20:39:47 mriedem because that patch is really wild west at the moment
20:39:55 mriedem it's more like --force to the source host
20:40:02 mriedem which is like, -4.5
20:47:39 mriedem cfriesen: your test vm went to ERROR state right
20:47:39 mriedem ?
20:47:49 cfriesen mriedem: yep
20:48:09 mriedem https://review.openstack.org/#/q/topic:bug/1811235+(status:open+OR+status:merged) should fix that
21:16:55 mriedem dansmith: fried_rice: one of you want to hit https://review.openstack.org/#/c/641416/ to cut down the nova-next run time?
21:17:04 mriedem shaves 20-30 minutes
21:17:07 fried_rice ...
21:17:44 mriedem just noticed that the gate blocker fix for evacuate is still sitting on the nova-next job
21:20:32 fried_rice mriedem: done
21:20:50 fried_rice suggestion inline about further efficientizing
21:21:11 mriedem yeah i don't know what to do about tempest-slow
21:21:35 mriedem i've been pushing back lately on adding more stuff to tempest, which kind of sucks
23:57:51 openstackgerrit Matt Riedemann proposed openstack/nova master: Add finish_snapshot_based_resize_at_dest compute method https://review.openstack.org/635080
23:57:51 openstackgerrit Matt Riedemann proposed openstack/nova master: Add FinishResizeAtDestTask https://review.openstack.org/635646
23:57:52 openstackgerrit Matt Riedemann proposed openstack/nova master: Add Destination.allow_cross_cell_move field https://review.openstack.org/614035
23:57:52 openstackgerrit Matt Riedemann proposed openstack/nova master: Execute CrossCellMigrationTask from MigrationTask https://review.openstack.org/635668
23:57:53 openstackgerrit Matt Riedemann proposed openstack/nova master: Plumb allow_cross_cell_resize into compute API resize() https://review.openstack.org/635684
23:57:53 openstackgerrit Matt Riedemann proposed openstack/nova master: Filter duplicates from compute API get_migrations_sorted() https://review.openstack.org/636224
23:57:54 openstackgerrit Matt Riedemann proposed openstack/nova master: Change HostManager to allow scheduling to other cells https://review.openstack.org/614037
23:57:54 openstackgerrit Matt Riedemann proposed openstack/nova master: Add CrossCellWeigher https://review.openstack.org/614353
23:57:55 openstackgerrit Matt Riedemann proposed openstack/nova master: Start functional testing for cross-cell resize https://review.openstack.org/636253
23:57:55 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add confirm_snapshot_based_resize_at_source https://review.openstack.org/637058
23:57:56 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add ConfirmResizeTask https://review.openstack.org/637070
23:57:56 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add confirm_snapshot_based_resize conductor RPC method https://review.openstack.org/637075
23:57:57 openstackgerrit Matt Riedemann proposed openstack/nova master: Confirm cross-cell resize from the API https://review.openstack.org/637316
23:57:57 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add revert_snapshot_based_resize_at_dest compute method https://review.openstack.org/637630
23:57:58 openstackgerrit Matt Riedemann proposed openstack/nova master: Deal with cross-cell resize in _remove_deleted_instances_allocations https://review.openstack.org/639453
23:57:58 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add finish_revert_snapshot_based_resize_at_source compute method https://review.openstack.org/637647
23:57:59 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add RevertResizeTask https://review.openstack.org/638046
23:57:59 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add revert_snapshot_based_resize conductor RPC method https://review.openstack.org/638047
#openstack-nova - 2019-03-09
00:02:29 openstackgerrit Matt Riedemann proposed openstack/nova master: Add user_id field to InstanceMapping https://review.openstack.org/633350
00:02:30 openstackgerrit Matt Riedemann proposed openstack/nova master: Populate InstanceMapping.user_id during migrations and schedules https://review.openstack.org/638574
00:02:31 openstackgerrit Matt Riedemann proposed openstack/nova master: Add online data migration for populating user_id https://review.openstack.org/633351
00:32:48 openstackgerrit Merged openstack/nova master: update gate test for removal of force evacuate https://review.openstack.org/641986
02:56:01 openstackgerrit Eric Fried proposed openstack/nova master: pass endpoint interface to Ironic client https://review.openstack.org/640879
14:18:17 openstackgerrit Merged openstack/nova master: Do not run tempest.scenario.test_network* tests in nova-next https://review.openstack.org/641416
18:48:27 openstackgerrit Merged openstack/nova master: Add user_id field to InstanceMapping https://review.openstack.org/633350
#openstack-nova - 2019-03-10
09:23:07 openstackgerrit garyk proposed openstack/nova master: Better handle live migration abort https://review.openstack.org/635440

Earlier   Later