| *** tetsuro has joined #openstack-placement | 00:36 | |
| *** takashin has joined #openstack-placement | 00:41 | |
| *** tssurya has joined #openstack-placement | 00:42 | |
| *** tssurya has quit IRC | 00:47 | |
| openstackgerrit | huanhongda proposed openstack/nova stable/ocata: [Stable Only] Remove soft-deleted instances from quota_usages https://review.openstack.org/570653 | 01:02 |
|---|---|---|
| *** gjayavelu has quit IRC | 01:13 | |
| *** tssurya has joined #openstack-placement | 02:05 | |
| *** tssurya has quit IRC | 02:09 | |
| *** bhagyashris has quit IRC | 02:14 | |
| *** bhagyashris has joined #openstack-placement | 02:15 | |
| openstackgerrit | Chen proposed openstack/nova master: Specify default update_resources_interval https://review.openstack.org/576864 | 02:21 |
| openstackgerrit | Dinesh Bhor proposed openstack/nova master: PCPU: Define numa dedicated CPU resource class https://review.openstack.org/561770 | 02:24 |
| openstackgerrit | Dinesh Bhor proposed openstack/nova master: PCPU: Add respective conf options https://review.openstack.org/561771 | 02:24 |
| openstackgerrit | Dinesh Bhor proposed openstack/nova master: NUMACell, InstanceNUMACell: Adopt 'PCPU' changes https://review.openstack.org/576021 | 02:24 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add instance.lock notification https://review.openstack.org/576702 | 02:27 |
| rpifan | hi | 02:29 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add instance.lock notification https://review.openstack.org/576702 | 02:30 |
| openstackgerrit | Chason Chan proposed openstack/nova master: Fix the duplicated config options of api_database and placement_database https://review.openstack.org/577023 | 02:41 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add instance.unlock notification https://review.openstack.org/576703 | 03:01 |
| *** gjayavelu has joined #openstack-placement | 03:17 | |
| *** tssurya has joined #openstack-placement | 03:28 | |
| *** tssurya has quit IRC | 03:33 | |
| *** rpifan has quit IRC | 03:50 | |
| *** gjayavelu has quit IRC | 04:04 | |
| *** gjayavelu has joined #openstack-placement | 04:11 | |
| openstackgerrit | Dinesh Bhor proposed openstack/nova master: Report PCPU to placement https://review.openstack.org/577038 | 04:17 |
| openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: Add microversion for nested allocation candidate https://review.openstack.org/565487 | 04:23 |
| openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: Fix nits in placement-return-all-resources series https://review.openstack.org/573604 | 04:28 |
| *** dansmith has quit IRC | 04:29 | |
| *** gjayavelu has quit IRC | 04:49 | |
| *** tssurya has joined #openstack-placement | 04:52 | |
| *** gjayavelu has joined #openstack-placement | 04:56 | |
| *** tssurya has quit IRC | 04:57 | |
| openstackgerrit | hyunsik Yang proposed openstack/nova-specs master: Specification for support pxe/IPMI boot for instance https://review.openstack.org/577048 | 05:08 |
| *** e0ne has joined #openstack-placement | 05:13 | |
| *** e0ne has quit IRC | 05:14 | |
| *** dansmith has joined #openstack-placement | 05:31 | |
| *** dansmith is now known as Guest88320 | 05:32 | |
| *** tssurya has joined #openstack-placement | 05:53 | |
| *** alex_xu has quit IRC | 05:57 | |
| *** tssurya has quit IRC | 05:58 | |
| *** nicolasbock has joined #openstack-placement | 06:00 | |
| *** alex_xu has joined #openstack-placement | 06:03 | |
| openstackgerrit | hyunsik Yang proposed openstack/nova-specs master: Specification for support pxe/IPMI boot for instance https://review.openstack.org/577048 | 06:15 |
| openstackgerrit | Chen proposed openstack/osc-placement master: Fix docstring for delete allocation method https://review.openstack.org/577075 | 06:17 |
| openstackgerrit | garyk proposed openstack/nova master: VMware: populate datastore refs at init https://review.openstack.org/574688 | 06:24 |
| openstackgerrit | Chen proposed openstack/nova master: Specify default update_resources_interval https://review.openstack.org/576864 | 06:29 |
| openstackgerrit | hyunsik Yang proposed openstack/nova-specs master: Specification for support pxe/IPMI boot for instance https://review.openstack.org/577048 | 06:51 |
| *** tssurya has joined #openstack-placement | 07:08 | |
| *** peereb has joined #openstack-placement | 07:15 | |
| *** tssurya_ has joined #openstack-placement | 07:16 | |
| *** tssurya_ has quit IRC | 07:21 | |
| *** efried has quit IRC | 07:28 | |
| *** efried has joined #openstack-placement | 07:28 | |
| *** e0ne has joined #openstack-placement | 07:40 | |
| *** ttsiouts has joined #openstack-placement | 07:48 | |
| *** belmoreira has joined #openstack-placement | 07:48 | |
| *** gjayavelu has quit IRC | 07:49 | |
| *** gjayavelu has joined #openstack-placement | 07:50 | |
| *** ttsiouts has quit IRC | 07:52 | |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add InstanceGroupPolicy object https://review.openstack.org/573628 | 07:54 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add policy to InstanceGroup object and api models. https://review.openstack.org/563375 | 07:54 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Add policy field to ServerGroup notification object https://review.openstack.org/563401 | 07:54 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Change the anti-affinity Filter to adapt to new policy https://review.openstack.org/571166 | 07:54 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Adapt _validate_instance_group_policy to new policy model https://review.openstack.org/571465 | 07:54 |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Microversion 2.64 - Use new format policy in server group https://review.openstack.org/567534 | 07:54 |
| *** ttsiouts has joined #openstack-placement | 08:23 | |
| *** tssurya_ has joined #openstack-placement | 08:39 | |
| *** tssurya_ has quit IRC | 08:44 | |
| openstackgerrit | Vladyslav Drok proposed openstack/nova master: ironic: Report resources as reserved when needed https://review.openstack.org/517921 | 08:49 |
| openstackgerrit | Gergely Csatari proposed openstack/nova master: Adding oslo.service to test-requirements https://review.openstack.org/577095 | 08:51 |
| *** e0ne has quit IRC | 09:01 | |
| *** gjayavelu has quit IRC | 09:08 | |
| *** tetsuro has quit IRC | 09:10 | |
| *** tetsuro has joined #openstack-placement | 09:10 | |
| *** tetsuro has quit IRC | 09:10 | |
| *** tetsuro has joined #openstack-placement | 09:13 | |
| openstackgerrit | Chris Dent proposed openstack/nova master: Cleanup nits in placement database changes https://review.openstack.org/577107 | 09:27 |
| *** takashin has left #openstack-placement | 09:30 | |
| openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Fix all invalid obj_make_compatible test case https://review.openstack.org/574240 | 09:31 |
| openstackgerrit | Vishakha Agarwal proposed openstack/nova master: No requirement of –all-tenants while listing servers https://review.openstack.org/577111 | 09:41 |
| *** e0ne has joined #openstack-placement | 09:55 | |
| *** cdent has joined #openstack-placement | 10:00 | |
| openstackgerrit | Vishakha Agarwal proposed openstack/nova master: No requirement of –all-tenants while listing servers https://review.openstack.org/577111 | 10:10 |
| *** e0ne has quit IRC | 10:10 | |
| *** e0ne has joined #openstack-placement | 10:12 | |
| *** ttsiouts has quit IRC | 10:15 | |
| *** e0ne has quit IRC | 10:16 | |
| *** ttsiouts has joined #openstack-placement | 11:10 | |
| *** tetsuro has quit IRC | 11:29 | |
| *** e0ne has joined #openstack-placement | 11:58 | |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Merge extended availability zone controller https://review.openstack.org/502859 | 12:25 |
| *** ttsiouts has quit IRC | 12:29 | |
| *** ttsiouts has joined #openstack-placement | 12:31 | |
| *** edmondsw has joined #openstack-placement | 12:58 | |
| *** ttsiouts has quit IRC | 12:59 | |
| openstackgerrit | jiang wei proposed openstack/nova master: Add action initiator attribute to the update payload https://review.openstack.org/536243 | 13:08 |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Merge server create schema for legacy BDM extension https://review.openstack.org/577152 | 13:09 |
| *** ttsiouts has joined #openstack-placement | 13:12 | |
| openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Tighten up ReportClient use of generation https://review.openstack.org/556669 | 13:14 |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Merge extended availability zone controller https://review.openstack.org/502859 | 13:15 |
| belmoreira | edleafe: efried: I'm having issues when trying to use a custom resource class for a GPU | 13:21 |
| *** jaypipes has joined #openstack-placement | 13:21 | |
| efried | belmoreira: tell me. | 13:21 |
| belmoreira | the resource class is defined but after adding it to the resource_provider inventory it gets removed by the resource tracker update | 13:22 |
| efried | belmoreira: You add it to the inventory how? Via CLI? | 13:22 |
| cdent | anybody noodling summit submissions? | 13:22 |
| belmoreira | efried: yes | 13:23 |
| efried | cdent: I just saw that email, haven't started thinking about it yet. | 13:23 |
| * cdent nods | 13:23 | |
| efried | belmoreira: Okay, so the virt driver is reporting the inventory it thinks it has - which doesn't include your GPU. | 13:23 |
| efried | belmoreira: and on RT update, we flush what the virt driver reports back to placement. | 13:23 |
| belmoreira | but being a custom resource class it shouldn't care | 13:24 |
| efried | That's an interesting way to look at it. It's not how it works at the moment. | 13:24 |
| efried | We don't just let you add random inventory to the compute from the CLI. | 13:24 |
| cdent | it's something we've discussed, but yeah, it's not currently supported | 13:25 |
| efried | belmoreira: Are you trying to do passthrough of physical GPU device? | 13:25 |
| belmoreira | efried: thanks. I was understanding custom to "anything" that we would like to define: colour, ... | 13:26 |
| efried | You can certainly *define* anything you want. But we're still pretty restricted in terms of how you can use it. | 13:26 |
| belmoreira | efried: yes, some are for passthrough but we also have the use case for vgpus with P100 | 13:27 |
| efried | belmoreira: vGPU stuff is partially supported and making progress - bauzas is on the bleeding edge of that effort. | 13:27 |
| efried | belmoreira: For traditional passthrough, you should still be able to use the legacy [pci]passthrough_whitelist BS. | 13:28 |
| cdent | for later: I reckon we probably need a way to figure out how to apply non-virt-driver managed resource classes. We've planned it for traits, right? | 13:28 |
| efried | cdent: Yes to the latter. For the former, the plan is to allow non-virt-driver to manage providers elsewhere in the tree. | 13:29 |
| belmoreira | efried: in our test infrastructure we were using the legacy [pci]passthrough_whitelist. But in that case we also need to enable the pci filter | 13:29 |
| efried | cdent: In fact, as soon as nrp-in-alloc-cands lands, you could probably get away with doing that via the CLI. | 13:30 |
| cdent | efried: i'm unable to translate what you just said. What I'm trying to say is "I think there ought to be a way for belmoreira to do what's being described: out of band custom resource classing that isn't happening on the nova-compute itself" | 13:30 |
| efried | including belmoreira's use case. | 13:30 |
| cdent | "just said being ^^, not ^" | 13:30 |
| belmoreira | the issue is that we only have few nodes with GPUs isolated in one cell and I would like to avoid to enable the pci filter in the entire cloud because with cellsV2 scheduler is global | 13:30 |
| efried | cdent: Yes, and I'm saying we don't plan to allow out-of-band custom resource classes (or any inventory changes) to be injected into providers that are *owned by compute*. Virt will always blow those away, as belmoreira is seeing here. | 13:31 |
| cdent | okay, I'm saying we should allow htat | 13:31 |
| efried | cdent: But for e.g. neutron & cyborg the plan is to allow nested providers. | 13:31 |
| cdent | It seems sane to be able to say my computer resource provider has stuff | 13:32 |
| cdent | doing it just by nested is fine, I guess, but I'm pathalogically meh about nested :) | 13:32 |
| efried | cdent: So once we get nrp-in-alloc-cands, that starts to be possible - but there's nothing special about neutron & cyborg that you couldn't just do the same via the CLI. | 13:32 |
| efried | belmoreira: How willing are you to run with unmerged code? :) | 13:32 |
| belmoreira | efried: ahahah | 13:33 |
| efried | belmoreira: I think we have proposed code that will allow you to get that GPU *allocated* - but note that said code still won't actually *attach* the GPU to your VM. | 13:33 |
| efried | ...which is presumably what you actually want. | 13:34 |
| efried | cdent: I don't agree that we should allow entity X to mess with inventory on a provider owned by entity Y. That way lies madness. | 13:35 |
| cdent | the thing that owns all the entities is the human running the installation | 13:35 |
| efried | (If you recall, I also didn't want entity X messing with *traits* on a provider owned by entity Y, but I got overruled on that one.) | 13:36 |
| belmoreira | efried: what I was expecting was to have something similar to ironic. A resource class is defined and the corresponding flavor. When using the flavor placement will provide only those RP to scheduler | 13:36 |
| efried | belmoreira: Ironic's custom resource classes are pushed by the virt driver. | 13:36 |
| efried | and also represent an entire compute node, IIUC. | 13:37 |
| belmoreira | efried: true | 13:37 |
| edleafe | so why isn't the virt driver able to "see" the GPU? | 13:38 |
| efried | edleafe: Cause it's not programmed to. | 13:38 |
| edleafe | Is that what bauzas is working on? | 13:39 |
| bauzas | mmm ? | 13:39 |
| efried | no, bauzas is working on *v*GPU | 13:39 |
| bauzas | are you discussing about GPU passthrough ? | 13:39 |
| efried | belmoreira: As it happens, I care very much about this subject. You could even say my paycheck arrives twice a month because I've promised to get generic GPU passthrough working for Power. So watch this space - it's coming soon. | 13:40 |
| edleafe | I thought that the virt driver is able to understand all its hardware, including how they get virtualized | 13:40 |
| efried | edleafe: In theory, sure. But at the moment if it reports hardware through get_available_resource, the only way it can get allocated is if it's a) in one of the special resource classes (VCPU, MEMORY_MB, DISK_GB), or b) via traditional PCI subsystem craziness. | 13:41 |
| openstackgerrit | Lee Yarwood proposed openstack/nova master: libvirt: Log breadcrumb for known encryption bug https://review.openstack.org/577164 | 13:42 |
| efried | edleafe: Stuff it reports through get_inventory/update_provider_tree gets allocated generically, but at the moment only libvirt is reporting anything other than those three, and I believe it's only reporting VGPU. | 13:42 |
| belmoreira | so for know the only workaround that I see is to play with placement aggregates like we are doing for cells and have one aggregate per GPU type | 13:42 |
| edleafe | efried: thanks for the explanation. So yeah, I would much rather see the virt driver be the only thing that messes with compute node inventory. | 13:43 |
| efried | belmoreira: You mean so that you can use PCI passthrough without enabling it across the whole cloud? | 13:43 |
| belmoreira | efried: yes | 13:44 |
| efried | belmoreira: I suppose. | 13:44 |
| belmoreira | efried: edleafe: what I don't understand is why is not a good idea to define directly a RC in a RP | 13:45 |
| belmoreira | I though that was the point of custom RC | 13:45 |
| edleafe | belmoreira: a custom RC is just something that isn't standard across all OpenStack deployments | 13:46 |
| efried | belmoreira: Okay, let's say hypothetically that you did that. | 13:46 |
| edleafe | An ironic node type, for example | 13:46 |
| efried | You defined CUSTOM_GPU and gave your compute node inventory of 2 units of that. | 13:46 |
| edleafe | The thing is that the virt driver is the authority on what is on a compute node | 13:46 |
| efried | Now you create a VM and ask for resources:CUSTOM_GPU=1 | 13:47 |
| edleafe | If it can't report it, it essentially isn't there | 13:47 |
| efried | And verily, we claim that and land you on that host and build your VM. | 13:47 |
| efried | Something has to know what to *do* with that CUSTOM_GPU=1 in the allocation. | 13:47 |
| efried | That something is... the virt driver. | 13:47 |
| cdent | not necessarily | 13:47 |
| efried | And nobody ever told the virt driver what to do with that thing. | 13:47 |
| cdent | there could be a third party which cares about that. the only thing that matters is counting them up and down by claiming and releasing allocations | 13:48 |
| efried | cdent: What, you want the user to manually attach the GPU after the fact? | 13:48 |
| cdent | i think gpu is a bad example because it is a piece of hardware | 13:48 |
| efried | bad example? It's the use case we're trying to solve right now. | 13:48 |
| cdent | but there could be lots of things that are inventory that is cared about by the compute node but do not need to be attached to the vm | 13:48 |
| cdent | the general question is "why can't I have a custom rc, right?" | 13:49 |
| cdent | for gpu's yes, they need to be managed by the virt driver | 13:49 |
| cdent | let's say, for example, that you wanted to limit the total number of absolute instances on a host | 13:50 |
| edleafe | cdent: then what would be an example of the sort on non-virt-driver-managed thing you have in mind? | 13:50 |
| efried | or by cyborg, or whatever. | 13:50 |
| cdent | CUSTOM_INSTANCES: total 45 | 13:50 |
| cdent | when we get to zero you can't use this host any more | 13:50 |
| cdent | when an instance goes away we make room for one more | 13:50 |
| edleafe | ugh - implementing scheduler filters? | 13:50 |
| cdent | originally the entire idea was to get rid of all scheduler filters | 13:51 |
| cdent | I'm not saying the above is a good idea, but is an example of the concept | 13:51 |
| cdent | (first thing that came to mind in the inevitable painful rush of irc conversation) | 13:51 |
| edleafe | so... you're saying that the virt driver should only be authoritative for *some* RCs, and not others? | 13:52 |
| efried | well, to be fair, that's what we've said for traits. | 13:52 |
| efried | which, once again, I think sucks. | 13:52 |
| efried | though I see why it's necessary (sort of). | 13:53 |
| cdent | edleafe: I'm not making a categorical statement one way or another. I'm merely supporting the idea that counting things is what resource classes provide, and some people may want to count in ways that we're not currently thinking of and I'm trying to think it through rather than dismiss it. | 13:54 |
| cdent | it could be a horrible way to do things, but I'm not currently sure | 13:54 |
| edleafe | cdent: sure; I'mm trying to think it through too | 13:54 |
| belmoreira | in my case I was just trying to avoid enable a filter because few nodes | 13:55 |
| efried | In my case I want the whole PCI subsystem to die in a fire. | 13:55 |
| edleafe | I can certainly support things like "CUSTOM_AVAILABLE_VMS" as a resource in theory | 13:56 |
| edleafe | ... except that if you don't want to have a limit, you'll have to set it to float(inf) :) | 13:56 |
| edleafe | efried: we all do | 13:57 |
| efried | edleafe, cdent: For such things, ironic has been using the model of having the user define them in such a way that the virt driver picks them up. | 13:57 |
| cdent | yes, but that works because of a special relationship between ironic and nova | 13:57 |
| efried | like there's some kind of ironic API or CLI where you have to set custom traits so that they get picked up and *reported* by virt. | 13:57 |
| edleafe | efried: yeah - the node has to have its resource_class defined | 13:57 |
| efried | I was talking about the traits thing, not the custom resource classes for compute nodes. | 13:58 |
| cdent | we have, in the past, talked about a compute node being able to define itself (it's traits and classes) at startup | 13:58 |
| edleafe | that was made mandatory in Ocata, I think | 13:58 |
| edleafe | but yeah, that is a one-off solution between ironic and nova | 14:00 |
| efried | The traits thing isn't | 14:00 |
| efried | it was designed to be more generic. | 14:00 |
| * edleafe goes for more coffee | 14:01 | |
| cdent | to distract us, I have a question about consumer generation conflict error messages. If you send a numeral when null/None is expected you get "expected None but got 1" | 14:02 |
| cdent | None is python-speak, not JSON speak, does that matter? | 14:02 |
| efried | I guess I would prefer if it said null. That would require a special case in the code, right? | 14:02 |
| efried | "expected %s but got %s" % ('null' if expected is None else expected, 'null' if actual is None else actual) | 14:03 |
| efried | kind of thing | 14:03 |
| efried | which is ew, but technically more correct. | 14:04 |
| cdent | it's already special cased: https://github.com/openstack/nova/blob/master/nova/api/openstack/placement/util.py#L643 | 14:05 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Add request_spec.RequestGroup versioned object https://review.openstack.org/568840 | 14:07 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Add requested_resources field to RequestSpec https://review.openstack.org/567267 | 14:07 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Add bandwidth related standard resource classes https://review.openstack.org/570847 | 14:07 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Transfer port.resource_request to the scheduler https://review.openstack.org/567268 | 14:07 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Send resource allocations in the port binding https://review.openstack.org/569459 | 14:07 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Test boot with more ports with bandwidth request https://review.openstack.org/573317 | 14:07 |
| cdent | efried: unless you've already pushed that on to your magic stack, I'll go ahead and make a reminder bug so I don't switch out of what I'm up to now | 14:09 |
| efried | cdent: Oh, didn't realize it was that easy. I can push up a patch. | 14:10 |
| efried | cdent: Doesn't even look like we have gabbits testing for it, so it'll be a one word change. | 14:11 |
| cdent | since i'm nearly done with the bug anyway, I'll continue, then you have something to link to | 14:11 |
| efried | cdent: Unless you want me to write a gabbi, which prolly wouldn't be the worst idea. | 14:11 |
| belmoreira | thanks all for the help | 14:12 |
| cdent | a gabbi thing is probably nice, becuase why not? https://bugs.launchpad.net/nova/+bug/1778071 | 14:12 |
| openstack | Launchpad bug 1778071 in OpenStack Compute (nova) "The placement consumer generation conflict error message can be misleading" [Low,Triaged] | 14:12 |
| efried | cdent: ack, on it. | 14:12 |
| cdent | belmoreira: are you any further along than when you started or did we go off into abstraction too much? | 14:13 |
| efried | cdent, edleafe: btw, real example of that "traits owned outside of virt" thing happened yesterday - see https://review.openstack.org/#/c/576877/ and its predecessor (linked in the commit message) | 14:15 |
| belmoreira | cdent: the abstraction was good because I was expecting to use custom RC as I wish. | 14:15 |
| cdent | belmoreira: cool, good | 14:15 |
| belmoreira | but still don't understand why we have a CLI to add custom RC to an inventory if they need to be set by the libvirt | 14:17 |
| openstackgerrit | Eric Fried proposed openstack/nova master: placement: s/None/null/ in consumer conflict msg https://review.openstack.org/577171 | 14:18 |
| efried | cdent: ^ | 14:18 |
| cdent | belmoreira: I think that discussion didn't reach its finish. There's more to talk about. | 14:18 |
| efried | belmoreira: Because Nova is (theoretically) not the only thing that can use placement. | 14:18 |
| efried | belmoreira: In the case of Nova, it was decided that the virt driver needs to be the source of truth for the compute node's inventory. | 14:19 |
| efried | belmoreira: With nested providers, we're edging toward a world where entities outside of Nova are allowed to be the source of truth for providers *underneath* the compute node provider. | 14:19 |
| efried | but we're not quite there yet. | 14:20 |
| belmoreira | efried: ok, thanks a lot | 14:20 |
| efried | belmoreira: In this case, the fact remains that you would still need some way to attach the allocated GPU to the instance. If you wanted to do that by hand, we could actually make this thing work. | 14:20 |
| efried | Because you *can* attach custom *traits* to an instance. So you could stuff CUSTOM_HAS_GPU on your compute node, just to make sure your instance landed there, and then do your manual attach. | 14:21 |
| efried | course, then you don't know when you've run out. | 14:21 |
| belmoreira | efried: yes | 14:23 |
| efried | belmoreira: if you're attaching GPUs by hand anyway, then you'll know when you've used the last of 'em, and you can remove the trait from the compute node :) | 14:27 |
| efried | hackisher and hackisher... | 14:27 |
| *** ttsiouts has quit IRC | 14:32 | |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Merge server create schema for BDM v2 extension https://review.openstack.org/577185 | 14:38 |
| openstackgerrit | Ghanshyam Mann proposed openstack/nova master: Merge server create schema for BDM v2 extension https://review.openstack.org/577185 | 14:40 |
| openstackgerrit | Chris Dent proposed openstack/nova master: WIP: [placement] Add /reshaper handler for POST https://review.openstack.org/576927 | 14:49 |
| openstackgerrit | Chris Dent proposed openstack/nova master: [placement] Extract create_allocation_list https://review.openstack.org/577200 | 14:49 |
| cdent | efried: the bottom of that ^ is some method extraction/refactoring for the thing above it. Your input appreciated on how to make further cleanups in the least annoying way. The inventory handling in reshaping duplicates a bit of code. The allocation handling doesn't need to. | 14:50 |
| cdent | edleafe: if you're around for real today, you too | 14:50 |
| efried | ack | 14:51 |
| *** peereb has quit IRC | 14:51 | |
| *** ttsiouts has joined #openstack-placement | 14:55 | |
| *** ttsiouts has quit IRC | 15:03 | |
| *** ttsiouts has joined #openstack-placement | 15:09 | |
| *** gjayavelu has joined #openstack-placement | 15:14 | |
| edleafe | cdent: about 50% today. I'll try to get to it. | 15:22 |
| cdent | thanks | 15:22 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Send resource allocations in the port binding https://review.openstack.org/569459 | 15:24 |
| openstackgerrit | Balazs Gibizer proposed openstack/nova master: Test boot with more ports with bandwidth request https://review.openstack.org/573317 | 15:24 |
| *** ttsiouts has quit IRC | 15:28 | |
| *** ttsiouts has joined #openstack-placement | 15:29 | |
| *** ttsiouts has quit IRC | 15:33 | |
| *** Guest88320 is now known as dansmith | 15:59 | |
| openstackgerrit | Eric Fried proposed openstack/nova master: Address nits from consumer generation https://review.openstack.org/577227 | 16:11 |
| openstackgerrit | Eric Fried proposed openstack/nova master: WIP: deduplicate {placement|api}_database help text https://review.openstack.org/577229 | 16:28 |
| *** e0ne has quit IRC | 16:30 | |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Remove NUMATopologyLimits.obj_from_db_obj https://review.openstack.org/537412 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Remove legacy '_to_dict' functions https://review.openstack.org/537413 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: network: Unchain '_get_phynet_info' from '_get_port_vnic_info' https://review.openstack.org/564443 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: network: Rename 'create_pci_requests_for_sriov_ports' https://review.openstack.org/575488 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: network: Always retrieve network information if available https://review.openstack.org/564444 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: network: Retrieve tunneled status in '_get_network_info' https://review.openstack.org/564445 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Add NUMACell.network_info https://review.openstack.org/564439 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Add NUMATopologyLimits.networks https://review.openstack.org/575486 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: hardware: Start accounting for networks in NUMA placement https://review.openstack.org/564448 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Add RequestSpec.numa_networks https://review.openstack.org/564442 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: scheduler: Start utilizing RequestSpec.numa_networks https://review.openstack.org/564452 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: conf: Add '[neutron] physnets' and related options https://review.openstack.org/564440 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Start populating NUMACell.network_info field https://review.openstack.org/564441 | 16:38 |
| openstackgerrit | Stephen Finucane proposed openstack/nova master: network: Rename '_get_phynet_info' to '_get_network_info' https://review.openstack.org/577233 | 16:38 |
| openstackgerrit | Eric Fried proposed openstack/nova master: WIP: deduplicate {placement|api}_database help text https://review.openstack.org/577229 | 16:45 |
| openstackgerrit | Zack Cornelius proposed openstack/nova master: Fix nits from change I676291ec0faa1dea0bd5050ef8e3426d171de4c6 https://review.openstack.org/576985 | 16:45 |
| openstackgerrit | Zack Cornelius proposed openstack/nova master: Implement discard for file backed memory https://review.openstack.org/576994 | 16:45 |
| *** gjayavelu has quit IRC | 16:56 | |
| *** gjayavelu has joined #openstack-placement | 17:09 | |
| cdent | nice digging on that aggregate stuff efried | 17:23 |
| efried | cdent: Ugh, thanks. Fixing it (everywhere) is going to be uuuuugly. | 17:23 |
| efried | cdent: But at least it validates the need to have put generations on aggregates. | 17:24 |
| *** e0ne has joined #openstack-placement | 17:30 | |
| openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Tighten up ReportClient use of generation https://review.openstack.org/556669 | 17:40 |
| *** tssurya has quit IRC | 17:54 | |
| openstackgerrit | Eric Fried proposed openstack/nova master: Tighten up ReportClient use of generation https://review.openstack.org/556669 | 17:58 |
| efried | cdent: In case you're following - that oughtta do it ^ | 17:58 |
| cdent | that's in my queue, but slogging through meeting at the moment. mleh | 17:59 |
| efried | further work is needed to resolve the actual race. Possibly with retries. | 17:59 |
| efried | Think I'm gonna take the afternoon off, in case anyone asks. | 18:17 |
| *** efried is now known as efried_pto | 18:20 | |
| cdent | good choice efried_pto . you work too much. | 18:20 |
| efried_pto | so says my wife | 18:20 |
| cdent | she's right | 18:20 |
| cdent | and you can tell her I said so | 18:21 |
| *** tssurya has joined #openstack-placement | 18:44 | |
| *** tssurya has quit IRC | 18:48 | |
| *** e0ne has quit IRC | 19:17 | |
| *** tssurya has joined #openstack-placement | 19:19 | |
| *** gjayavelu has quit IRC | 19:33 | |
| openstackgerrit | Lenny Verkhovsky proposed openstack/nova master: WIP workaround for RHEL7.5 qemu 2.0 https://review.openstack.org/577280 | 19:35 |
| *** gjayavelu has joined #openstack-placement | 20:33 | |
| *** takashin has joined #openstack-placement | 20:43 | |
| openstackgerrit | Matt Riedemann proposed openstack/nova master: Address nits from consumer generation https://review.openstack.org/577227 | 20:44 |
| openstackgerrit | Chris Dent proposed openstack/nova master: WIP: [placement] Add /reshaper handler for POST https://review.openstack.org/576927 | 20:50 |
| *** gjayavelu has quit IRC | 20:52 | |
| *** cdent has quit IRC | 20:53 | |
| openstackgerrit | Takashi NATSUME proposed openstack/nova-specs master: Create specs directory for Stein https://review.openstack.org/573602 | 21:15 |
| *** tssurya has quit IRC | 21:35 | |
| *** nicolasbock has quit IRC | 21:54 | |
| *** tssurya has joined #openstack-placement | 21:56 | |
| *** tssurya has quit IRC | 22:00 | |
| *** edmondsw has quit IRC | 22:24 | |
| *** edmondsw has joined #openstack-placement | 22:26 | |
| *** edmondsw has quit IRC | 22:35 | |
| *** edmondsw has joined #openstack-placement | 22:35 | |
| *** tssurya has joined #openstack-placement | 22:38 | |
| openstackgerrit | Merged openstack/nova master: Add check if neutron "binding-extended" extension is available https://review.openstack.org/523548 | 22:39 |
| *** edmondsw has quit IRC | 22:40 | |
| *** tssurya has quit IRC | 22:43 | |
| openstackgerrit | Zack Cornelius proposed openstack/nova master: Fix nits from change Ia7cf4414feb335b3c2e863b4c8b4ff559b275c34 https://review.openstack.org/577314 | 22:44 |
| openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in test_compute_api.py (4) https://review.openstack.org/568462 | 23:58 |
| openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in virt/test_block_device.py https://review.openstack.org/566153 | 23:59 |
| openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in nova/tests/unit/virt/xenapi/stubs.py https://review.openstack.org/568412 | 23:59 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!