Posted	Nick	Remark
#openstack-nova - 2018-05-31
09:11:17	bauzas	naichuans: replied to the thread with a proposal, lemme give you the link
09:12:21	naichuans	bauzas: OK
09:12:51	bauzas	naichuans: there it is http://lists.openstack.org/pipermail/openstack-dev/2018-May/130955.html
09:13:23	bauzas	naichuans: does Xen support vGPU live migration ?
09:13:34	naichuans	bauzas: got it~
09:13:58	bauzas	because QEMU folks say it's a very difficult thing to implement because there are a huge number of corner cases to consider
09:22:16	kashyap	bauzas: Which part is this?
09:22:29	kashyap	Ah, Xen
09:22:36	bauzas	kashyap: which part of what ?
09:22:49	kashyap	Difficulty in implementing in QEMU
09:23:12	bauzas	maybe the wording is inappropriate
09:23:40	bauzas	it's not "difficult", it's just quite having a lot of cases to consider
09:30:41	naichuans	bauzas: sorry, just saw your message about VGPU live migration. XenServer do support live migration
09:44:14	naichuans	bauzas: just finish reading your conversation history about n-rp in vGPU. You are using DB online migration to avoid compute service restart, right?
09:49:49	openstackgerrit	Lee Yarwood proposed openstack/nova stable/queens: libvirt: handle DiskNotFound during update_available_resource https://review.openstack.org/571424
09:49:50	openstackgerrit	Lee Yarwood proposed openstack/nova stable/queens: libvirt: Skip fetching the virtual size of block devices https://review.openstack.org/571425
09:50:39	openstackgerrit	Lee Yarwood proposed openstack/nova stable/pike: libvirt: handle DiskNotFound during update_available_resource https://review.openstack.org/571426
09:50:40	openstackgerrit	Lee Yarwood proposed openstack/nova stable/pike: libvirt: Skip fetching the virtual size of block devices https://review.openstack.org/571427
09:56:36	openstackgerrit	Balazs Gibizer proposed openstack/nova master: Reject interface attach with QoS aware port https://review.openstack.org/570078
09:56:37	openstackgerrit	Balazs Gibizer proposed openstack/nova master: Reject networks with QoS policy https://review.openstack.org/570079
09:58:21	openstackgerrit	wanghongtao proposed openstack/nova master: Fix the metadata re to match the unicode https://review.openstack.org/536236
10:07:31	openstackgerrit	Merged openstack/nova master: pci: don't consider case when match tags specs https://review.openstack.org/565808
10:16:40	openstackgerrit	Lee Yarwood proposed openstack/nova stable/ocata: libvirt: handle DiskNotFound during update_available_resource https://review.openstack.org/571432
10:16:41	openstackgerrit	Lee Yarwood proposed openstack/nova stable/ocata: libvirt: Skip fetching the virtual size of block devices https://review.openstack.org/571433
10:20:27	sahid	artom: thanks for what you did on the vf trusted serie :)
10:21:00	artom	sahid, oh, hah, I'd just apologized on the downstream channel
10:21:47	sahid	artom: yes but, please don't, i have appreciated thanks again
10:22:33	artom	sahid, cool, you're welcome :)
10:22:44	artom	We *really* need to unfsck the metadata API
10:23:15	artom	The fact that we're not using templates, and adding/removing stuff based on ifs...
10:23:21	artom	*Versioned* templates, I should say
10:23:38	sahid	artom: i see, yes that make sense to have such mech
10:24:08	artom	sahid, well the "real" API uses those, I believe
10:27:00	sahid	artom: i did not get time to look at your serie about the live migration, i saw that it's still in WIP - my understanding is i could wwait a bit more before to start reviewing it
10:27:11	sahid	or do you think we should start to make some reviews?
10:27:17	artom	sahid, yeah, no worries, there's not much there yet
10:27:23	artom	Well
10:28:07	artom	https://review.openstack.org/#/c/566398/7 and https://review.openstack.org/#/c/566723/8 and ready
10:28:23	artom	The last one is marked WIP because the one below it is
10:28:35	artom	It's like a WIP sandwich
11:27:56	openstackgerrit	Hironori Shiina proposed openstack/nova master: virt/ironic: Implement rescue and unrescue https://review.openstack.org/416487
11:51:24	openstackgerrit	Hironori Shiina proposed openstack/nova master: WIP: virt/ironic: Implement rescue and unrescue https://review.openstack.org/571450
11:57:43	openstackgerrit	Jan Gutter proposed openstack/os-vif master: Add virtio_multiqueue field to InstanceInfo https://review.openstack.org/570959
12:34:01	openstackgerrit	Zhenyu Zheng proposed openstack/nova master: WIP abort live migration in queue. https://review.openstack.org/568542
12:36:55	openstackgerrit	Zhenyu Zheng proposed openstack/nova master: WIP abort live migration in queue. https://review.openstack.org/568542
12:37:33	openstackgerrit	Jan Gutter proposed openstack/nova master: Convert vrouter legacy plugging to os-vif https://review.openstack.org/571325
12:37:34	openstackgerrit	Jan Gutter proposed openstack/nova master: Use vif.vif_name in _set_config_VIFGeneric https://review.openstack.org/571461
12:37:35	openstackgerrit	Jan Gutter proposed openstack/nova master: Pass virtio multiqueue info to os-vif plugins https://review.openstack.org/571462
12:51:44	openstackgerrit	Yikun Jiang (Kero) proposed openstack/nova master: WIP: Adapt _validate_instance_group_policy to new policy model https://review.openstack.org/571465
13:13:47	openstackgerrit	Surya Seetharaman proposed openstack/nova master: Delete orphaned RS and IM records of archived instances https://review.openstack.org/560042
13:14:19	openstackgerrit	Chen proposed openstack/nova master: WIP Closes-Bug: #1449221 https://review.openstack.org/571472
13:14:20	openstack	bug 1449221 in OpenStack Compute (nova) "Nova doesn't allow cleanup of volumes stuck in 'attaching' or 'detaching' status" [High,Confirmed] https://launchpad.net/bugs/1449221
13:50:29	melwitt	nova meeting in 10 minutes
13:53:00	mriedem	dansmith: can you get this queens backport? https://review.openstack.org/#/c/571425/
13:54:13	mriedem	lyarwood: looks like you have something to fix in this one https://review.openstack.org/#/c/571433/
13:57:02	dansmith	hmm, not sure why that depends on the update resource patch
13:57:59	lyarwood	mriedem: yeah, couldn't run tox locally for Ocata for some reason, I'll get to it later today
13:58:44	lyarwood	dansmith: it just keeps the backport clean, I was already looking to backport the resource patch anyway so...
13:59:24	dansmith	hmm, okay
14:00:55	openstackgerrit	sahid proposed openstack/os-vif master: Add virtio_multiqueue field to InstanceInfo https://review.openstack.org/570959
14:04:13	openstackgerrit	Jay Pipes proposed openstack/nova master: placement: always create consumer records https://review.openstack.org/567678
14:04:14	openstackgerrit	Jay Pipes proposed openstack/nova master: add consumers generation field https://review.openstack.org/557958
14:04:15	openstackgerrit	Jay Pipes proposed openstack/nova master: placement: Allocation.consumer field https://review.openstack.org/565405
14:04:16	openstackgerrit	Jay Pipes proposed openstack/nova master: rework allocation handler _allocations_dict() https://review.openstack.org/565407
14:04:17	openstackgerrit	Jay Pipes proposed openstack/nova master: Add a microversion for consumer generation support https://review.openstack.org/565604
14:09:00	jaypipes	btw, dansmith, fixed the silly pep8 error in ^
14:09:16	dansmith	ack
14:09:18	jaypipes	dansmith: let me know if you'd like further tests added for ensure_consumer()
14:09:27	dansmith	you know I will
14:09:30	jaypipes	:)
14:10:07	mriedem	i had some random thoughts about a nova-status upgrade check for the missing consumer stuff,
14:10:14	mriedem	but haven't thought about it much
14:10:25	mriedem	basically query the api db for allocations w/o a consumer
14:10:36	mriedem	but not sure what to do if we find them
14:10:42	jaypipes	mriedem: the bottom patch on that series adds an online db migration for "incomplete" consumers.
14:10:46	dansmith	you want nova-status to connect directly to the placement database?
14:11:03	mriedem	nova-status already connects directly to the nova_api db, which is the placement db today
14:11:09	mriedem	jaypipes: yeah i forgot about that
14:11:17	dansmith	right, but it muddies the water for a split
14:12:16	mriedem	yeah if we have an online data migration i'm not sure we need an upgrade status check thing
14:13:34	mriedem	i think i was thinking,
14:13:53	mriedem	somehow healing the allocations / creating missing consumers based on the actual instance for the allocation, rather than the config options
14:13:59	mriedem	which is what jay's online data migration is using it looks like
14:14:51	dansmith	hmm
14:15:17	dansmith	I mean, I guess we can do that now, but we won't be able to do things like that past the split, so I would think we should stick to being cleanly separated even before
14:15:42	jaypipes	mriedem: your heal_allocations() work would already do that (once the always create consumer patch is merged, of course)
14:16:00	mriedem	jaypipes: the heal_allocations CLI won't do anything if the instance already has allocations
14:16:06	mriedem	but it doesn't check if the project/user is missing
14:16:17	jaypipes	mriedem: ah, yes, true nuf
14:16:41	mriedem	https://review.openstack.org/#/c/565886/9/nova/cmd/manage.py@1794
14:21:14	sahid	jangutter: i have question for you in the osvif patch
14:21:27	jangutter	shoot!
14:21:50	mriedem	jaypipes: having said that, we could make heal_allocations look to see if the existing allocations are missing consumer/project and re-post the allocations w/ the consumer/project if so
14:22:40	sahid	jangutter: is that could help whether we have that new virtio_multiqueu attribute an integer, so we can pass the exact number of queue enabled?
14:23:35	jangutter	sahid: I dunno, this is weird territory for me... literally the only code I could find used iproute2 to enable "multiqueue" and that's it.
14:23:52	jaypipes	mriedem: IIRC, that was my original suggestion that I could do that in the online DB migration
14:24:14	jaypipes	mriedem: but it was decided that the single CONF option was a simpler solution for the majority case.
14:24:37	sahid	jangutter: it's not something that will be used by vrouter so?
14:24:51	jaypipes	mriedem: part of the issue with that is that I would have to import nova objects and DB into the placement object files in order to grab instance information.
14:24:59	jangutter	sahid: It seems to be a bit of an aberration, almost all of the other VIF types can learn the multiqueue configs from the hypervisor, just TAP seem to need something on the plug side.

Earlier   Later