| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-04-26 | |||
| 12:51:33 | openstackgerrit | Matt Riedemann proposed openstack/nova master: conductor: use port binding extended API in during live migrate https://review.openstack.org/522537 | |
| 13:18:04 | Tahvok | Hey guys! Is it possible to request a release for ocata branch? | |
| 13:18:24 | mriedem | Tahvok: i think that's already up | |
| 13:18:35 | mriedem | https://review.openstack.org/#/c/564044/ | |
| 13:19:40 | Tahvok | Oh, nice! I was waiting for some fix to be realesed for a couple of months now :) | |
| 13:19:44 | Tahvok | Thanks! | |
| 13:21:14 | openstackgerrit | Kashyap Chamarthy proposed openstack/nova master: libvirt: Lift the restriction of choices for `cpu_model_extra_flags` https://review.openstack.org/563926 | |
| 13:26:20 | mordred | mriedem, jaypipes: how hard/expensive would it be to return a total count of objects for paginated collections? | |
| 13:27:32 | mordred | (just got a question about being able to get a count of items without having to loop over the iterator in the sdk - but we can't do it because pagination... thought I'd check to see how insane it would be to get that added | |
| 13:29:30 | cdent | mordred: you're after a 'total' bit of metadata? | |
| 13:30:07 | mordred | cdent: yah. atleast, maybe I'm after a total bit of metadata | |
| 13:30:42 | mordred | cdent: I'm honestly not sure how muchit's a thing I want - but cost of implementation probably impacts how much I want it :) | |
| 13:31:15 | cdent | I know how that can be. I'm afrait that I'm not familiar with the pagination handling in nova's api, so can't provide any quick and dirty guess | |
| 13:34:03 | jaypipes | mordred: it would be a total refactoring of the pagination system. | |
| 13:34:19 | mordred | jaypipes: awesome.then I don't want it | |
| 13:34:23 | jaypipes | ha | |
| 13:36:36 | openstackgerrit | Jianghua Wang proposed openstack/nova master: XenAPI: define a new image handler to use vdi streaming https://review.openstack.org/486475 | |
| 13:36:37 | openstackgerrit | Jianghua Wang proposed openstack/nova master: XenAPI: document change for supporting image handler https://review.openstack.org/563077 | |
| 13:37:31 | openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add host/hostId to instance action events API https://review.openstack.org/557145 | |
| 13:38:09 | openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Support inherit api samples in ApiSampleTestBase https://review.openstack.org/562547 | |
| 13:46:11 | openstackgerrit | Jianghua Wang proposed openstack/nova master: XenAPI: document change for supporting image handler https://review.openstack.org/563077 | |
| 13:51:45 | mriedem | right we don't return a total in the links response today, | |
| 13:51:56 | mriedem | and...we also limit the responses by default to 1000 | |
| 13:52:09 | mriedem | so based on that, the total we give might not be the actual total number of that resource in the db | |
| 13:52:22 | mriedem | well, you can pass in limit | |
| 13:56:02 | openstackgerrit | Merged openstack/nova master: ServerActionsSampleJsonTest refactor https://review.openstack.org/561819 | |
| 13:58:56 | openstackgerrit | Merged openstack/nova stable/queens: libvirt: fix hard reboot issue with mdevs https://review.openstack.org/564454 | |
| 14:13:34 | mriedem | kashyap: some small things to cleanup in https://review.openstack.org/#/q/topic:bump_min_libvirt_and_qemu+status:open | |
| 14:13:48 | kashyap | mriedem: Will look | |
| 14:14:17 | kashyap | mriedem: On that other patch of `cpu_model_extra_flags` -- while context is fresh, a functional change comment: | |
| 14:14:52 | kashyap | mriedem: I was just talking to QEMU / libvirt folks, and I was reminded that allowing / removing extra flags even for 'host-passthrough' mode makes sense. Why? An example: | |
| 14:15:23 | kashyap | (a) An Operator might want to explicitly disable nested virt for a Nova guest; then they can remove 'vmx' CPU flag | |
| 14:16:22 | kashyap | (b) QEMU may disable some features unless you explicitly ask for them (like: "invtsc" CPU flag) -- in that case it makes sense to explicitly specify the flag | |
| 14:16:29 | kashyap | I'll add a note in the code for a future patch. | |
| 14:16:55 | kashyap | Or after some thinking ... I might add it in the current patch. Haven't decided yet | |
| 14:17:06 | openstackgerrit | Eric Fried proposed openstack/nova master: Add host/hostId to instance action events API https://review.openstack.org/557145 | |
| 14:17:50 | efried | mriedem: This should be a quick +A (trivial changes since your prior +2) ^ | |
| 14:19:16 | mriedem | kashyap: if you're going to drop the model restrictions when dropping the choices restriction, it'd be best to do that in the same change so you're not having to dump multiple feature release notes on people for the same config option | |
| 14:19:41 | mriedem | i will have to take your word on the host-passthrough thing | |
| 14:20:11 | kashyap | Yeah, exactly -- I too don't like find it appealing to make the feature release notes muddly | |
| 14:20:19 | openstackgerrit | Jay Pipes proposed openstack/nova master: support multiple member_of qparams https://review.openstack.org/561315 | |
| 14:20:53 | kashyap | mriedem: You don't have to; I can provide a minimal functional test case you can try yourself on a linux laptop | |
| 14:21:06 | kashyap | With bare QEMU. That demonstrates it. | |
| 14:22:51 | kashyap | mriedem: Sorry about this one: https://review.openstack.org/#/c/564013/3/nova/virt/libvirt/driver.py | |
| 14:22:59 | openstackgerrit | Dan Smith proposed openstack/nova master: Make service all-cells min version helper use scatter-gather https://review.openstack.org/539590 | |
| 14:23:00 | openstackgerrit | Dan Smith proposed openstack/nova master: Make host_manager use scatter-gather and ignore down cells https://review.openstack.org/539617 | |
| 14:23:05 | kashyap | mriedem: Mixed up versions there; it's for Stein | |
| 14:23:13 | dansmith | mriedem: ^ | |
| 14:24:00 | mriedem | kashyap: ok, s/have to// | |
| 14:31:14 | kashyap | mriedem: Here's the first piece of evidence on point (b) I noted above: | |
| 14:31:17 | kashyap | Starting a guest with <cpu mode='host-passthrough' and feature policy='require' name='invtsc' does not provide cpu flag nonstop_tsc in the guest. | |
| 14:31:23 | kashyap | Source: https://bugzilla.redhat.com/show_bug.cgi?id=1384124 | |
| 14:31:23 | openstack | bugzilla.redhat.com bug 1384124 in qemu-kvm-rhev "cpu flag nonstop_tsc is not present in guest with host-passthrough and feature policy require invtsc" [High,Closed: errata] - Assigned to ehabkost | |
| 14:33:55 | kashyap | Where "invtsc" == Invariant TSC (Time Stamp Counter) -- there's a Wikipedia entry about it | |
| 14:33:56 | openstackgerrit | Jay Pipes proposed openstack/nova master: support multiple member_of qparams https://review.openstack.org/561315 | |
| 14:36:39 | openstackgerrit | Ed Leafe proposed openstack/nova master: Add the Consumer object https://review.openstack.org/561406 | |
| 14:38:05 | openstackgerrit | Chris Dent proposed openstack/nova master: Use nova.db.api directly https://review.openstack.org/543262 | |
| 14:58:11 | kashyap | mriedem: I made a test case, just for you: https://kashyapc.fedorapeople.org/CPU-flags-and-host-passthrough/readme.txt | |
| 14:58:43 | kashyap | Translation: libvirt's "host-passthrough" in QEMU paralance is '-cpu host' | |
| 14:59:17 | kashyap | If you have a CirrOS image called 'base.qcow2', then you can run the two scripts in the top-level directory and see the difference | |
| 15:01:58 | kashyap | (The above is a proper function test, BTW.) | |
| 15:02:44 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Call generate_image_url only for legacy notification https://review.openstack.org/564528 | |
| 15:08:48 | jaypipes | cdent: you may want to have a gander at https://review.openstack.org/#/c/524425/ | |
| 15:09:46 | cdent | jaypipes: yeah, I'm aware of that | |
| 15:11:32 | cdent | jaypipes: I still have the same question as ever on that stuff: what policy? | |
| 15:12:40 | jaypipes | cdent: sorry, I'm not following you... | |
| 15:14:40 | cdent | jaypipes: I don't see the use case for granular authZ in placement (yet). | |
| 15:18:33 | mriedem | i'm working on that right now, hence my questions about request context in -placement | |
| 15:18:54 | mriedem | cdent: talk to mgagne, he wants granular policy for placement | |
| 15:19:42 | cdent | mriedem: I know, and the use cases that they've described do not provide any real protections | |
| 15:20:43 | cdent | if you have access to modify inventories, you have access to breaking the whole thing | |
| 15:39:47 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: libvirt: Configure multiple vGPU types https://review.openstack.org/564214 | |
| 15:39:48 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: libvirt: mdevs returning parent and vendor PCI info https://review.openstack.org/562304 | |
| 15:39:49 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: WIP: libvirt: Get inventories per pGPU https://review.openstack.org/564215 | |
| 15:46:00 | dansmith | mriedem: ugh, I think we have a race for service version checking somewhere in unit tests | |
| 15:46:03 | dansmith | a bad mock or something | |
| 15:46:13 | dansmith | I dunno why it's showing up in this patch | |
| 15:51:43 | mriedem | cdent: you don't have to give access to modify inventories | |
| 15:51:53 | mriedem | you can give read-only access to non-admins for things like listing providers and showing inventory | |
| 15:51:59 | mriedem | and leave POST and PUT to admin | |
| 15:52:16 | cdent | mriedem: I believe the ask was to be able to control allocation ratios, but maybe I'm not remembering right | |
| 15:53:06 | wolverineav | sahid: stephenfin - just trying to move the conversation here for faster convergence. if you're both online here. there's 2 open queries. (1) get_config_ivs() being removed. i'm not sure where its called or used. (2) #try os-vif codepath works or not - this is true and verified. | |
| 15:55:00 | stephenfin | wolverineav: It's being called here https://github.com/openstack/nova/blob/master/nova/virt/libvirt/vif.py#L653-L654 | |
| 15:55:25 | sahid | wolverineav: you have to convert this function to be something like _set_config_VIF | |
| 15:55:28 | stephenfin | (Yes, that _is_ really non-obvious and I plan to fix it in the future) | |
| 15:56:14 | sahid | so when osvif object is going to be generate, nova will call it based on the vif type | |
| 15:56:21 | sahid | that to configure domain xml | |
| 15:56:43 | sahid | os-vif never configure domain xml | |
| 15:56:50 | sahid | it's nova responsbility | |
| 15:56:52 | stephenfin | wolverineav: Yeah, what sahid said. The path it will be taking is from [1] to [2] | |
| 15:56:56 | stephenfin | [1] https://github.com/openstack/nova/blob/master/nova/virt/libvirt/vif.py#L648 | |
| 15:57:08 | stephenfin | [2] https://github.com/openstack/nova/blob/master/nova/virt/libvirt/vif.py#L616-L622 | |
| 15:57:36 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: DNM: Showing some strange behavior https://review.openstack.org/564541 | |
| 15:58:08 | stephenfin | wolverineav, sahid: We should probably document this once in os-vif we're done. It's certainly a little confusing | |
| 16:00:48 | sahid | stephenfin: yes it is | |
| 16:04:57 | wolverineav | sahid: stephenfin - got it. so the get_config_* methods i should just simply remove, since they're for legacy path. that fixes (1). and for (2) - i installed the os-vif-bigswitch plugin locally and did some regression to bring up a VM. so that satisfies (2). makes sense? | |
| 16:05:22 | wolverineav | i mean get_config_ivs related methods to be removed. | |
| 16:05:43 | stephenfin | wolverineav: I think you need to remove the get_config_ivs methods and, if necessary, add some '_set_config_ivs' functions | |
| 16:06:03 | stephenfin | Assuming you need to modify the libvirt XML for those interfaces | |
| 16:06:10 | stephenfin | If not, you're good | |