| opendevreview | Jorge San Emeterio proposed openstack/nova master: Improving logging at '_allocate_mdevs'. https://review.opendev.org/c/openstack/nova/+/861146 | 07:55 |
|---|---|---|
| opendevreview | Amit Uniyal proposed openstack/nova master: WIP: Functional test for 1980720 https://review.opendev.org/c/openstack/nova/+/861076 | 10:30 |
| opendevreview | Pierre Libeau proposed openstack/nova master: Use admin_client to allocate port for instance https://review.opendev.org/c/openstack/nova/+/861172 | 12:04 |
| obutenko | Hello, All | 12:58 |
| obutenko | I currently working with new functionality based on the "PCI Device Tracking In Placement" spec. We would like to have the next behavior, placement should choose the device based on the flavor trait property and we want to use a new device_spec field, traits. According to the next configuration, we would like to have the next behavior. When we create an instance with m1.large flavor, we use PCI 0000:65:00.06, when we create an instance with m1.med | 12:58 |
| obutenko | To achieve this behavior we make the next configuration: | 12:59 |
| obutenko | openstack flavor set --property trait:CUSTOM_PCI_2=required m1.medium | 12:59 |
| obutenko | openstack flavor set --property trait:CUSTOM_PCI_3=required m1.large | 12:59 |
| sean-k-mooney | that is not how we wanted to enabel that | 12:59 |
| gibi | obutenko: the scheduler support is not merged yet for PCI Device Tracking in placement | 12:59 |
| sean-k-mooney | so currently the way we wanted to do the trati based selection was via the pci alais | 13:00 |
| *** dasm|off is now known as dasm | 13:00 | |
| sean-k-mooney | so the alias would have the trait request and the flavor woudl jsut request the alias as it does today | 13:00 |
| gibi | obutenko: see the open patches here https://review.opendev.org/q/topic:bp/pci-device-tracking-in-placement+status:open | 13:00 |
| obutenko | ok, thanks | 13:01 |
| sean-k-mooney | gibi do you have the traits bit doen already by the way i cant re | 13:02 |
| sean-k-mooney | found it https://review.opendev.org/c/openstack/nova/+/853316 | 13:02 |
| sean-k-mooney | obutenko: ^ that is what will enable triats based selection via the alias | 13:02 |
| sean-k-mooney | obutenko: we wanted to keep the layer of indirection provided by the alias beauce it simplifes the code and the configuration significantly | 13:03 |
| sean-k-mooney | in the future i would kind of like to see a way to model device-profiles or resouce templates or somethign similar like a flaovr at the api level that could perhaps unify how we do device managment in nova | 13:05 |
| sean-k-mooney | but i dont see us doing that for a year or more | 13:05 |
| sean-k-mooney | we have a number of diffent ways to consume reosuce form the host currently for vgpu/mdevs, pci/sriov, pmem, cpupinning, hugepages and cyborg | 13:06 |
| sean-k-mooney | the answer might be just use resouces: in the falvor for all of them however that is problematic as then its very hard to know how to mange that resouce type in nova so right now we use the alias to know its sriov/pci and a cyborg device-profile for there resouces | 13:08 |
| *** ministry is now known as __ministry | 13:11 | |
| *** knikolla[m] is now known as knikolla | 13:26 | |
| sean-k-mooney | gibi: i proably need to set out some time to help review and land the rest fo that by the way | 13:28 |
| opendevreview | Dan Smith proposed openstack/nova master: Test ceph-multistore with a real image https://review.opendev.org/c/openstack/nova/+/860864 | 13:28 |
| sean-k-mooney | we should try and cordiate some time after the ptg to try and compelte that before milestone 1 | 13:28 |
| *** blarnath is now known as d34dh0r53 | 13:32 | |
| gibi | sean-k-mooney: yes, and I have to set aside a time to fix the smallish things open there | 13:41 |
| gibi | yepp, lets coordinate on that after the PTG | 13:41 |
| opendevreview | Dan Smith proposed openstack/nova master: Test ceph-multistore with a real image https://review.opendev.org/c/openstack/nova/+/860864 | 14:09 |
| atmark | is there way to tell if a VM was created with --avaibility-zone? Just trying to determine if a VM is restricted to a AZ. | 14:14 |
| atmark | if a VM is restricted to AZ, a resize shouldn't bypass right? | 14:17 |
| gibi | atmark: https://docs.openstack.org/api-ref/compute/?expanded=show-server-details-detail#show-server-details has a field OS-EXT-AZ:availability_zone | 14:20 |
| gibi | atmark: and yes resize will keep the instance in the same AZ | 14:20 |
| atmark | gibi: I created two VMs, one with AZ specified and second one without, openstack server show both has OS-EXT-AZ:availability_zone | 14:25 |
| gibi | the one that you created without AZ ended up assigned to the default AZ | 14:28 |
| gibi | see the second bulletpoint of https://docs.openstack.org/nova/latest/admin/availability-zones.html | 14:29 |
| opendevreview | Dan Smith proposed openstack/nova master: Test ceph-multistore with a real image https://review.opendev.org/c/openstack/nova/+/860864 | 14:31 |
| sean-k-mooney | atmark: without lookign at the db you cant tell the differnce | 14:32 |
| sean-k-mooney | if you look at he request spec and instance table they will be differnt based on if it was booted with a az request or not | 14:33 |
| sean-k-mooney | btu its not visable form teh api | 14:33 |
| atmark | sean-k-mooney: ah, so just in the db. i'll look into those tables and do a diff. | 14:41 |
| atmark | thx | 14:42 |
| atmark | also, if it was booted with az request, resize shouldn't bypass right? | 14:43 |
| bauzas | atmark: if you created a instance with a specific AZ, then the resize scheduling will check for the hosts which ones are in the AZ | 15:06 |
| atmark | got it | 15:21 |
| *** tkajinam is now known as Guest2971 | 15:43 | |
| bauzas | melwitt: fancy me asking whoami-rajat for a nova-cinder session due to your topic https://etherpad.opendev.org/p/nova-antelope-ptg L192 ? | 15:55 |
| bauzas | first time I'm seeing an item in the etherpad that reflects some disussion for cinder | 15:55 |
| bauzas | ralonsoh: I just moved two topics from our Nova PTG etherpad and promoted them for our cross-project meeting https://etherpad.opendev.org/p/nova-antelope-ptg L59 | 16:00 |
| bauzas | ralonsoh: I guess you also have topics in mind for our meeting ? | 16:00 |
| melwitt | bauzas: oh, I didn't know there wasn't already going to be a cross-project session. I'm not sure it's worth creating a new one only because of one topic | 16:12 |
| bauzas | melwitt: well, there was no cinder-nova session planned until I saw your topic | 16:13 |
| bauzas | hence my question | 16:13 |
| bauzas | should we ask for one ? | 16:13 |
| melwitt | bauzas: I wouldn't worry about it. I had assumed there would be one, wrong assumption on my part | 16:14 |
| bauzas | melwitt: we can arrange that with whoami-rajat | 16:14 |
| bauzas | at least promoting this topic to the cinder folks so they can join | 16:14 |
| melwitt | bauzas: I would prefer not to burden with a last minute thing. sure, it doesn't hurt to let them know so they can see if they think it's interesting | 16:17 |
| bauzas | cool, will see then | 16:17 |
| melwitt | I think maybe I have the issue solved with the patch I proposed but I am not sure. if I've got it wrong, we will likely need to talk to them further | 16:18 |
| melwitt | ok, thanks | 16:18 |
| ralonsoh | bauzas, https://etherpad.opendev.org/p/neutron-antelope-ptg | 16:20 |
| opendevreview | Dan Smith proposed openstack/nova master: Test ceph-multistore with a real image https://review.opendev.org/c/openstack/nova/+/860864 | 16:20 |
| ralonsoh | what are the topics? | 16:20 |
| bauzas | ralonsoh: see https://etherpad.opendev.org/p/nova-antelope-ptg L59 | 16:21 |
| ralonsoh | I scheduled the two topics I had for Thursday from 13 to 15 UTC | 16:21 |
| bauzas | we have two topics for you folks | 16:22 |
| ralonsoh | bauzas, I think we can use the Friday slots, there is no agenda for the neutron operator hout | 16:23 |
| ralonsoh | hour* | 16:23 |
| ralonsoh | so I'll propose those topics to be discussed on Friday, ok? | 16:23 |
| bauzas | ralonsoh: hah, you think we need more than 2 hours for all topics ?. | 16:25 |
| bauzas | ralonsoh: if so, ack | 16:25 |
| ralonsoh | bauzas, we can start on thrusday | 16:26 |
| bauzas | yeah | 16:26 |
| ralonsoh | and if we need more time (more that 2 hours) | 16:26 |
| bauzas | let's be agile (c) :) | 16:26 |
| ralonsoh | we can use Friday too | 16:26 |
| bauzas | ++ | 16:26 |
| * bauzas needs to leave | 16:26 | |
| ralonsoh | I'll send the mail (tomorrow) | 16:26 |
| whoami-rajat | bauzas, melwitt we're up for a cross project section if we've topics | 16:37 |
| melwitt | whoami-rajat: I'm not totally sure we need to discuss the topic I had added on our etherpad. it's about the volume format vs snapshot format that came up on https://review.opendev.org/c/openstack/cinder/+/857528 I think maybe I have fixed my patch but I need a sanity check from one of yall. and if I've got it wrong on my patch, that's when we need further discussion iiuc | 16:40 |
| whoami-rajat | melwitt, ack, sorry I haven't been able to get time to take a look, will try to take out some time for it but we've light on agenda so a cross project shouldn't be a problem as well | 17:00 |
| melwitt | whoami-rajat: np, I only got it working recently. ok, thank you | 17:02 |
| opendevreview | Dan Smith proposed openstack/nova master: Test ceph-multistore with a real image https://review.opendev.org/c/openstack/nova/+/860864 | 18:18 |
| *** dasm is now known as dasm|off | 23:52 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!