| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-03-27 | |||
| 08:37:28 | jianghuaw_ | I think we still have to restrict each pGPU only support one vGPU type. | |
| 08:37:33 | openstackgerrit | Silvan Kaiser proposed openstack/nova master: Exec systemd-run with privileges in Quobyte driver https://review.openstack.org/554195 | |
| 08:37:44 | gibi | bauzas: OK, then good for you | |
| 08:37:46 | jianghuaw_ | so root RP(compute) -> child RP (pGPU)<vgpu inventory> | |
| 08:38:14 | bauzas | jianghuaw_: if so, we should be asking operators to say which type for each PCI device, right? | |
| 08:38:28 | bauzas | if we want to support multiple types | |
| 08:38:51 | jianghuaw_ | I guess yes. | |
| 08:38:56 | bauzas | and also, I'm not sure that Intel has the same problem than nVidia where only one type is possible for each pGPU | |
| 08:39:11 | jianghuaw_ | For xen, it's root RP(compute) -> child RP (pGPU group)<vgpu inventory> | |
| 08:39:40 | bauzas | but each group is only having one type, right? | |
| 08:39:52 | bauzas | if so, how do you say which type for each group ? | |
| 08:41:08 | jianghuaw_ | all of the same type PGPUs will be put into the same group. | |
| 08:41:27 | bauzas | not sure I understand | |
| 08:41:50 | bauzas | each physical device is having multiple possible types | |
| 08:41:54 | jianghuaw_ | so still depending the configure option to restrict each grpu only has one type enabled. | |
| 08:42:32 | jianghuaw_ | But if there are multiple types of PGPUs exists in single compute. it can support different types. | |
| 08:43:05 | bauzas | say I'm doing enabled_vgpu_types=nvidia-1,nvidia-2 | |
| 08:43:12 | bauzas | (or the name you use in Xen) | |
| 08:43:31 | bauzas | if I have 2 pGPUs, each one supporting both types, what Xen will have ? | |
| 08:43:39 | bauzas | two pGPU groups ? | |
| 08:43:42 | jianghuaw_ | Also we can ask operators to customize the pGPU groups. Then we can put some pgpus into group1 and then other others be group2. | |
| 08:43:52 | jianghuaw_ | and each can enable different vGPU types. | |
| 08:45:45 | jianghuaw_ | switching to a meeting. | |
| 08:47:35 | jichen | sahid: sorry for delay, I replied in the patch and saw your question here | |
| 08:47:48 | bauzas | jianghuaw_: mmmm, not sure I like that | |
| 08:55:26 | kashyap | melwitt: When you're back: the ListOpt() trick works -- https://review.openstack.org/#/c/534384/ | |
| 09:04:11 | openstackgerrit | Theodoros Tsioutsias proposed openstack/nova-specs master: Add PENDING vm state https://review.openstack.org/554212 | |
| 09:10:30 | openstackgerrit | Sylvain Bauza proposed openstack/nova-specs master: Proposes NUMA topology with RPs https://review.openstack.org/552924 | |
| 09:18:56 | openstackgerrit | Konstantinos Samaras-Tsakiris proposed openstack/nova master: Add `hide_hypervisor_id` flavor extra_spec https://review.openstack.org/555861 | |
| 09:24:01 | ktibi | Hi, how can I disable the compatibility check for live migration ? because I have two CPU model : Nehalem & Nehalem-IBRS and migration fail :/ | |
| 09:36:49 | openstackgerrit | Tetiana Lashchova proposed openstack/nova-specs master: Allow modification of user-data via the server update https://review.openstack.org/547964 | |
| 09:41:06 | jianghuaw_ | bauzas, came back from a call meeting. I understood your concern. The problem is that we can't handle the dynamically changing vGPU capacities. | |
| 09:41:22 | openstackgerrit | Bhagyashri Shewale proposed openstack/nova-specs master: Disallow rotation parameter 0 for 'createBackup' API https://review.openstack.org/511825 | |
| 09:41:42 | jianghuaw_ | I mean the available vGPUs for one type will be impacted by other types belongs to the same pGPU. | |
| 09:42:15 | jianghuaw_ | that's why we have to keep the restriction that each PGPU (or pgpu group) can only expose one vGPU type. | |
| 09:42:24 | kaisers1 | mikal: I'm not sure which direction to go in https://review.openstack.org/#/c/554195/ , could you pls review regarding Stephens comments? | |
| 09:43:27 | bauzas | jianghuaw_: sure but I don't think it's a problem | |
| 09:44:17 | bauzas | jianghuaw_: at least for libvirt, what I know is that if I'm providing multiple inventories, then when I'll create the first mdev, it'll automatically update the inventories of the other types to be total=0 | |
| 09:45:34 | jianghuaw_ | bauzas, will it cause conflict? | |
| 09:46:36 | jianghuaw_ | in the case multiple vGPUs have been allocated from multiple inventories | |
| 09:47:53 | jianghuaw_ | Before create the first mdev, all inventories will have >0 vGPUs available. Right? | |
| 09:48:39 | jianghuaw_ | If yes, it's possible to allocate vGPUs from different inventories. | |
| 09:49:31 | jianghuaw_ | Then the first mdev creation will result into the other inventories' total be 0. the allocatoin for the other vGPUs will fail. right? | |
| 09:50:43 | openstackgerrit | Matthew Booth proposed openstack/nova-specs master: Add serial numbers for local disks https://review.openstack.org/556565 | |
| 09:52:37 | openstackgerrit | Silvan Kaiser proposed openstack/nova master: Exec systemd-run with privileges in Quobyte driver https://review.openstack.org/554195 | |
| 09:55:28 | stephenfin | kashyap: This of any interest? Abandoning if not https://review.openstack.org/#/c/348394/ | |
| 09:56:52 | kashyap | stephenfin: That's another piecemeal way of fixing the current mess of handling different OVMF binaries | |
| 09:57:15 | kashyap | stephenfin: Can be abandoned, IMHO. And we should handle it globally for all distributions | |
| 09:57:50 | kashyap | https://review.openstack.org/#/c/506720/ | |
| 09:58:16 | kashyap | stephenfin: Which is in turn a bit predicated on this RFC I started for libvirt and QEMU: [RFC] Defining firmware (OVMF, et al) metadata format & file | |
| 09:58:19 | kashyap | https://lists.nongnu.org/archive/html/qemu-devel/2018-03/msg01978.html | |
| 09:58:20 | stephenfin | kashyap: Cool, done | |
| 10:01:01 | stephenfin | sahid: Any chance you could abandon these, now that bauzas' vGPU series has merged? https://review.openstack.org/#/q/topic:pci-mdev-support-compaq+(status:open+OR+status:merged) | |
| 10:15:40 | sahid | stephenfin: i'm expecting gerrit to abandon them automatically at some point | |
| 10:16:22 | stephenfin | Ah, it doesn't do that. Someone (typically sdague in the past) had to run a script or something | |
| 10:27:51 | kashyap | stephenfin: Question for you here: https://review.openstack.org/#/c/530924/ | |
| 10:28:49 | kashyap | Also, sigh (existing fault), naming nuisance: 'disk_cachemodes' (should be 'disk_cache_modes') | |
| 10:29:30 | kashyap | Changing such things at this point would be futile (for OCD's sake), as it might break scripts, & other tools, etc :-( | |
| 10:56:52 | openstackgerrit | Chris Dent proposed openstack/nova master: [placement] Filter resource providers by forbidden traits in db https://review.openstack.org/556472 | |
| 10:56:52 | openstackgerrit | Chris Dent proposed openstack/nova master: [placement] Filter allocation candidates by forbidden traits in db https://review.openstack.org/556660 | |
| 10:56:53 | openstackgerrit | Chris Dent proposed openstack/nova master: [placement] Parse forbidden traits in query strings https://review.openstack.org/556819 | |
| 10:56:53 | openstackgerrit | Chris Dent proposed openstack/nova master: [placement] Support forbidden traits in API https://review.openstack.org/556820 | |
| 10:57:25 | cdent | Okay, with that done, I can brew up to start the spec dive | |
| 11:03:17 | openstackgerrit | Surya Seetharaman proposed openstack/nova master: Scheduling Optimization: Remove cell0 from the list of candidates https://review.openstack.org/556821 | |
| 11:05:45 | sean-k-mooney | melwitt: not that my vote counts in the actul ballot but i had assumed efried had a +/-1.5 for a while so i'd also be +1 on the proposal to add you to the core team. | |
| 11:11:25 | kashyap | sean-k-mooney: What do you mean 1.5? | |
| 11:12:23 | kashyap | (Also: IMHO, even if your vote doesn't "count" by some measure, active contributors such as you expressing their views is equally important, in my books.) | |
| 11:12:28 | sean-k-mooney | kashyap: that while efried does not have +2 right currently there are some area of the code where his +/- 1 are treated as +/- 2 due to the experience he has demonstrated in that area | |
| 11:12:57 | kashyap | I see. I guessed as much, just wanted to double-confirm. | |
| 11:13:42 | kashyap | sean-k-mooney: Unrelated bait (apologies): I think this is ready: https://review.openstack.org/#/c/534384/ | |
| 11:13:46 | sean-k-mooney | kashyap: it does not "count" because the vote is between members of the core team rather then an open vote. but yes i know its still valuble for active memeber to express there support/decent | |
| 11:14:01 | sean-k-mooney | kashyap: oh yes i skimmed it yesterday | |
| 11:14:32 | openstackgerrit | Yikun Jiang (Kero) proposed openstack/nova master: Record the host info in EventReporter https://review.openstack.org/556746 | |
| 11:14:43 | kashyap | sean-k-mooney: Thanks! Tests pass, backport concerns addressed, rel note in place (removed needless verbosity per dansmith's review) | |
| 11:15:01 | sean-k-mooney | i see you got the release nots job passing | |
| 11:15:12 | kashyap | Yep! I got the rST dark magic right | |
| 11:15:43 | sean-k-mooney | ill quickly review again but last time i looked it seamed fine to me | |
| 11:16:09 | kashyap | sean-k-mooney: Thanks, the core change is quite small: I limit the 'choices' keyword arg to one value (PCID) | |
| 11:16:12 | sean-k-mooney | i saw dans comment and kind of aggreed but not strongly enough to ask for a respin | |
| 11:16:28 | kashyap | sean-k-mooney: Yeah; I reworded it, and moved the CPU models info into the config file | |
| 11:16:37 | kashyap | As Operators can assume that it can be applied to all models. | |
| 11:16:59 | kashyap | As we know, some virtual CPU models include it (like the Haswell variants), some don't (Nehalem, etc). | |
| 11:19:51 | sean-k-mooney | item_type=types.String( | |
| 11:19:53 | sean-k-mooney | choices=['pcid'] | |
| 11:19:55 | sean-k-mooney | ) | |
| 11:19:57 | sean-k-mooney | thats new to me | |
| 11:20:23 | sean-k-mooney | kashyap: is that syntax documented in oslo somewhere? | |
| 11:20:45 | kashyap | sean-k-mooney: Not quote, I found it from stephenfin's use here: https://github.com/openstack/nova/blob/cd15c3d/nova/conf/vnc.py#L226,L232 | |
| 11:20:50 | kashyap | s/quote/quite/ | |
| 11:20:59 | kashyap | sean-k-mooney: I only skimmed the oslo_config/cfg.py, though | |
| 11:21:21 | kashyap | https://github.com/openstack/oslo.config/blob/master/oslo_config/cfg.py | |
| 11:21:22 | sean-k-mooney | kashyap: well i have used choice elements for string fields before just never lists | |
| 11:21:44 | kashyap | sean-k-mooney: It's actually documented here: https://github.com/openstack/oslo.config/blob/master/oslo_config/cfg.py#L52 | |
| 11:24:09 | sean-k-mooney | ya so its not that you are using a listOpt that has a choice filed, you are instead create a list of sting enums which have the chioce element | |
| 11:24:14 | sean-k-mooney | but it works | |
| 11:24:55 | kashyap | Exactly :P It's still using the 'String' type | |
| 11:26:32 | openstackgerrit | Elod Illes proposed openstack/nova stable/ocata: WIP: Functional test for regression bug #1713783 https://review.openstack.org/505160 | |
| 11:26:33 | openstack | bug 1713783 in OpenStack Compute (nova) ocata "After failed evacuation the recovered source compute tries to delete the instance" [High,In progress] https://launchpad.net/bugs/1713783 - Assigned to Balazs Gibizer (balazs-gibizer) | |
| 11:32:14 | kashyap | sean-k-mooney: Thanks for the review. Much appreciated | |
| 11:33:10 | sean-k-mooney | kashyap: no worries. i dont get enough time as i like to review but always feel free to ping me with a patch if you have one | |