| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-06-07 | |||
| 23:34:29 | alex_xu | jaypipes: Can I use trait to pass down baremetal hardware configuration to the ironic? it's like there a, b ,c BIOS confiugrations, then I have three traits mapping to this three configurations. And attach this three traits to the node, and placement can ensure anyone of this three traits schedule to the right node. Then ironic configure the BM based on this trait. | |
| #openstack-nova - 2018-06-08 | |||
| 00:48:22 | openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: placement: Use INNER JOIN for requied traits https://review.openstack.org/573113 | |
| 01:05:20 | openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: Test alloc_cands with indirectly sharing RPs https://review.openstack.org/519601 | |
| 01:20:48 | openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: Fix nits in nested provider allocation candidates(2) https://review.openstack.org/573475 | |
| 01:23:38 | mriedem_afk | alex_xu: that sounds like a nightmare to manage | |
| 01:24:03 | mriedem | alex_xu: there the discussion in dublin about using glare to store bm config and then nova could pass that along to ironic | |
| 01:40:05 | alex_xu | mriedem_afk: not sure how much people deploy glare... | |
| 01:40:29 | mriedem_afk | alex_xu: this could give them a reason | |
| 01:40:38 | mriedem_afk | but yes agree with you | |
| 01:41:11 | alex_xu | mriedem_afk: just hope a simple design | |
| 01:42:15 | alex_xu | mriedem_afk: what is the point about manage nightmare? | |
| 01:43:30 | mriedem_afk | managing bios config traits per node for thousands of nodes | |
| 01:43:54 | mriedem_afk | if it's all automated by some other external service, then maybe not a big deal | |
| 01:44:10 | mriedem_afk | but traits have to be mapped to flavors too | |
| 01:44:43 | mriedem_afk | if you have something like tacker that is single tenant and just creates flavors on demand, then maybe not a big deal (again) | |
| 01:48:40 | alex_xu | mriedem_afk: so the key problem is we should decouple the configuration and the flavor? | |
| 01:54:11 | mriedem_afk | yeah i think so | |
| 01:54:49 | mriedem | the idea in dublin was to post your config in glare, get an id handle, and then pass that artifact_id with nova server create and nova would get the artifact from glare and pass it on to the virt driver | |
| 01:56:27 | alex_xu | mriedem: yea, i see, why we say no for the instance metadata? | |
| 01:58:07 | mriedem | you'd need to read https://review.openstack.org/#/c/408151/ | |
| 01:58:24 | mriedem | also, unrelated, but it looks like i'm able to filter servers with --ip6 even when using compute api version 2.1 | |
| 01:58:32 | mriedem | but --ip6 filter support was added in 2.5 | |
| 01:58:40 | alex_xu | mriedem: ah, thanks, I'm looking for this spec in the ironic, then I didn't found it... | |
| 01:59:05 | mriedem | oh i know why | |
| 01:59:08 | mriedem | i'm running as admin | |
| 02:00:07 | alex_xu | mriedem: yea, in memory, we expose filter to the non-admin with microversion | |
| 02:01:30 | mriedem | ah yes | |
| 02:01:37 | mriedem | ip6 is ignored for non-admin before 2.5 | |
| 02:03:38 | pvc | hi | |
| 02:03:43 | pvc | can anyone help me with gpu passthorught | |
| 02:04:42 | mriedem | alex_xu: re: https://etherpad.openstack.org/p/compute-api-microversion-gap-in-osc | |
| 02:04:57 | mriedem | pvc: naichuans might be able to | |
| 02:05:28 | pvc | I've successfully passthrough the video card of my compute node to my instance | |
| 02:05:43 | naichuans | :) | |
| 02:05:49 | pvc | but when im installing nvidia driver it seems that my compute node is rebooting or cant access and cant ping | |
| 02:06:16 | alex_xu | mriedem: thanks god we needn't add support for every microversion in osc | |
| 02:06:25 | pvc | Then im rebooting my compute node so i can ping it again but when i reboot my VM i cannot access again my compute node | |
| 02:06:29 | pvc | any idea naichuans? | |
| 02:06:48 | naichuans | pvc: Which hypervisor arew you using? | |
| 02:06:58 | pvc | im using Centos Queens | |
| 02:07:23 | pvc | i think i misconfigured on compute node? | |
| 02:07:35 | pvc | but i can see the videocard on my instance | |
| 02:07:47 | pvc | but fail when running nvidia-smi | |
| 02:08:07 | pvc | NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running. ubuntu 16.04 | |
| 02:09:49 | naichuans | strange, install driver looks no effect on compute node. Do you change anything in nova.conf | |
| 02:10:00 | pvc | yes wait | |
| 02:12:01 | pvc | naichuans http://paste.openstack.org/show/722937/ | |
| 02:12:15 | naichuans | pvc: will check it | |
| 02:12:17 | pvc | you want to see grub.cgf too on my compute? | |
| 02:12:32 | naichuans | yes, plase | |
| 02:15:51 | pvc | naichuans http://paste.openstack.org/show/722938/ | |
| 02:16:04 | pvc | im planning to test tensorflow | |
| 02:16:10 | naichuans | got it~ | |
| 02:17:22 | pvc | this is the flavor and image i used | |
| 02:17:40 | pvc | naichuans http://paste.openstack.org/show/722939/ | |
| 02:19:36 | pvc | do i need to add something? | |
| 02:20:52 | openstackgerrit | Shuo Liu proposed openstack/nova-specs master: fix spelling mistake https://review.openstack.org/573479 | |
| 02:21:57 | naichuans | pvc: haven't find anything wrong yet. I will check if I could found something. By the way, I saw you are using KVM, maybe bauzas: could help because he is an expert on vgpu in KVM. | |
| 02:22:21 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in unit/network/test_neutronv2.py (2) https://review.openstack.org/573481 | |
| 02:23:07 | pvc | hi sir bauzas | |
| 02:25:03 | pvc | do you have docs on how to install driver for ubuntu naichuans | |
| 02:26:54 | naichuans | pvc: No, but for VM side, I think it is no different with physical machines, you can get them from google | |
| 02:27:39 | pvc | its weird that when i spawn up an instance i cant reach my compute node | |
| 02:27:57 | pvc | do i need to use UEFI? | |
| 02:29:30 | naichuans | I have no idea yet. But that's would not effect compute node. | |
| 02:35:24 | pvc | is this right naichuans http://paste.openstack.org/show/722940/ | |
| 02:36:39 | pvc | is my step okay now queens release? | |
| 02:36:47 | pvc | i mean the way i configure my compute | |
| 02:40:53 | naichuans | pvc: looks OK for me, will confirm it with hypervisor guys | |
| 02:41:59 | pvc | thank you so much naichuans | |
| 02:42:14 | pvc | i need to enable it :( for tensorflow migration | |
| 02:42:48 | naichuans | pvc: Np~ | |
| 02:49:44 | openstackgerrit | Shuo Liu proposed openstack/nova-specs master: fix spelling mistake in rst file https://review.openstack.org/573487 | |
| 03:04:16 | naichuans | pvc: I wonder you are using the wrong pci vendor/prodcut id. Did you find the id in /sys/bus/pci? | |
| 03:26:16 | openstackgerrit | Steve Kowalik proposed openstack/nova master: Switch to oslo_messaging.ConfFixture.transport_url https://review.openstack.org/573046 | |
| 03:36:33 | pvc | naichuans where on sys/bus/pci | |
| 03:37:58 | pvc | naichuans http://paste.openstack.org/show/722945/ | |
| 03:40:30 | naichuans | pvc: yes, you could find some information like `00:02.0 Display controller [0380]: Intel Corporation Iris Pro Graphics P6300 ` using lspci, then find ids in `/sys/bus/pci/devices/0000:00:02.0`, do `cat device`/`cat vendor` | |
| 03:41:16 | naichuans | mostly vendor should be `0x8086`, your configuration looks different. | |
| 03:57:32 | pvc | wait naichuans | |
| 03:58:56 | pvc | here naichuans http://paste.openstack.org/show/722946/ | |
| 03:59:25 | pvc | http://paste.openstack.org/show/722947/ | |
| 05:09:17 | naichuans | pvc: yes, it is different from your configure in nova.conf http://paste.openstack.org/show/722937/ | |
| 05:10:20 | pvc | may i know what should i input there? naichuans? can you do a paste again? | |
| 05:14:20 | pvc | naichuans thank you | |
| 05:14:29 | naichuans | pvc: could you try http://paste.openstack.org/show/722948/? | |
| 05:14:51 | naichuans | pvc: Not sure it can fix your issue or not, just have a try :) | |
| 05:15:00 | pvc | okay thank you so much :) | |
| 05:15:08 | naichuans | Np | |
| 05:22:19 | pvc | on the /etc/modprobe.d should i edit something or my config is okay now? i use vfio-pci | |
| 05:22:57 | pvc | naichuans http://paste.openstack.org/show/722949/ | |
| 05:24:43 | naichuans | pvc: I'm not clear about that configure, sorry | |
| 05:30:17 | pvc | i need to hide nouveau right? | |
| 05:39:21 | naichuans | I didn't try, maybe native driver is better | |
| 06:19:10 | pvc | naichuans how can i enable UEFI support? | |
| 06:25:47 | openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: Return all resources in provider_summaries https://review.openstack.org/558045 | |
| 06:25:48 | openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: Return all providers - multiple providers https://review.openstack.org/559480 | |
| 06:25:49 | openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: Return all providers - single provider https://review.openstack.org/573545 | |
| 06:37:35 | openstackgerrit | Tetsuro Nakamura proposed openstack/nova master: Fix nits in nested provider allocation candidates(2) https://review.openstack.org/573475 | |
| 06:45:12 | openstackgerrit | Naichuan Sun proposed openstack/nova master: xenapi(N-R-P): Add API to support vgpu resource provider create https://review.openstack.org/520313 | |