| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-02-04 | |||
| 11:40:05 | stephenfin | As in the XML generated by nova or attributes of the instance? | |
| 11:40:30 | lyarwood | stephenfin: well that or the GuestConfig (?) objects we generate that in turn create the XML | |
| 11:40:49 | lyarwood | stephenfin: I don't really want to assert things in the XML if we have objects that work just as well tbh | |
| 11:41:20 | lyarwood | LibvirtConfigObjects is what I mean sorry | |
| 11:41:35 | lyarwood | np if there's no prior art, I'll hack something up now | |
| 11:41:50 | stephenfin | Hmm, I don't think so. The only example I can think of is the NUMA live migration tests, but that intercepts the live migration API call to validate the XML | |
| 11:42:22 | lyarwood | cool yeah I think that's the one I saw yesterday, I'll hack something up now | |
| 11:43:00 | lyarwood | _get_guest_config should work tbh | |
| 12:14:50 | sean-k-mooney | lyarwood: there isnt really one | |
| 12:15:01 | sean-k-mooney | unless you have a reference to the driver | |
| 12:15:17 | sean-k-mooney | normally the funct test start the services | |
| 12:15:53 | sean-k-mooney | and there is no rpc call that you can make to teh compute manager to get it so you need to use teh service instance ot get teh driver an then call functions on it. | |
| 12:16:01 | sean-k-mooney | but what you can do is use the logs | |
| 12:16:18 | sean-k-mooney | and you can technically get teh domain xmls form that if you had too | |
| 12:19:10 | lyarwood | yeah I've got it working | |
| 12:19:27 | lyarwood | we keep a reference to the driver so it's a normal passthrough mock | |
| 12:20:08 | sean-k-mooney | cool | |
| 12:46:22 | lyarwood | sean-k-mooney: so https://review.opendev.org/c/openstack/nova/+/774044/1/nova/utils.py would still let hw_machine_type through right | |
| 12:47:04 | sean-k-mooney | yes | |
| 12:47:30 | sean-k-mooney | but i know how to fix that im debating if i want to go with a simpler fix before this one | |
| 12:48:01 | sean-k-mooney | basically i notice we never included ramdisk_id and kernel_id when we converted image metadata to ovos | |
| 12:48:24 | sean-k-mooney | so this became much more involed of a change involving object changes | |
| 12:48:42 | lyarwood | kk | |
| 12:48:45 | sean-k-mooney | but it was quite late/early this moring when i figured that out and didnt want to start agin | |
| 12:49:40 | sean-k-mooney | it will include hw_machine_type because you are using the same key as the image proerty - a sufix | |
| 12:49:50 | sean-k-mooney | lyarwood: if you add a prefix it would not be included | |
| 12:50:22 | lyarwood | kk, well I think I might leave it, the actual behaviour means I don't need to make any other changes to the util method fetching the machine type | |
| 12:50:26 | sean-k-mooney | what im going to do in the simpler version is remove key = key[len(SM_IMAGE_PROP_PREFIX):] | |
| 12:50:50 | lyarwood | kk | |
| 12:50:54 | sean-k-mooney | and build the proerties differnetly so it only contianes the one with image_ | |
| 12:51:22 | sean-k-mooney | lyarwood:hehe | |
| 12:51:46 | lyarwood | so that would cause me to also add a system_metadata lookup in that util method | |
| 12:51:56 | lyarwood | that isn't the end of the world, it's what I had originally | |
| 12:52:05 | lyarwood | in a seperate change | |
| 12:52:30 | sean-k-mooney | well i wanted you to use image_ orginally as a prefix to avoid the extra lookup | |
| 12:52:35 | sean-k-mooney | we can do it either way | |
| 12:52:49 | sean-k-mooney | i was going to look at wriing the other patch to have someting backportable | |
| 12:52:54 | sean-k-mooney | assumeing we wanted too | |
| 12:53:07 | sean-k-mooney | if we dont then i can just finish fixing this patch | |
| 12:53:31 | lyarwood | that's also something I could do, always storing it as SM_IMAGE_PROP_PREFIX_hw_machine_type | |
| 12:53:41 | lyarwood | and that wouldn't bork me | |
| 12:53:45 | lyarwood | with this fix in place | |
| 12:53:57 | sean-k-mooney | ya | |
| 12:54:40 | sean-k-mooney | do you have an opipion on what way i should go with the fix. should i also do the version without the ovo change | |
| 12:59:18 | lyarwood | I don't at the moment tbh | |
| 12:59:33 | lyarwood | Let me post this and grab something to eat and then I'll think it through once I'm back | |
| 12:59:56 | sean-k-mooney | cool its more dod we need a backportable fix or not | |
| 13:00:21 | sean-k-mooney | my current fix uses the ovo field but we forgot to add at least 2 when we created the object | |
| 13:00:40 | sean-k-mooney | so it cant be backported but i can fix it without relying on the fields too | |
| 13:00:43 | sean-k-mooney | which would be | |
| 13:01:28 | sean-k-mooney | i can do both too so its not either or | |
| 13:02:11 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP libvirt: Record the machine_type of instances in system_metadata https://review.opendev.org/c/openstack/nova/+/767533 | |
| 13:02:13 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP nova-status: Add hw_machine_type check for libvirt instances https://review.opendev.org/c/openstack/nova/+/770643 | |
| 13:02:13 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP nova-manage: Add commands for managing instance machine type https://review.opendev.org/c/openstack/nova/+/769548 | |
| 13:04:24 | lyarwood | sean-k-mooney: https://review.opendev.org/c/openstack/nova/+/767533/5/nova/tests/functional/libvirt/test_machine_type.py - comments on the tests here would be appreciated btw | |
| 13:05:25 | sean-k-mooney | sure. i think i owe bauzas a review of his routed stuff first but i now have your open in front of me :) | |
| 13:08:51 | lyarwood | thanks | |
| 13:09:25 | bauzas | sean-k-mooney: I'll upload a new revision https://review.opendev.org/c/openstack/nova/+/773976 today later | |
| 13:10:37 | sean-k-mooney | bauzas: ok ill review lyarwood patch then now and ill look at yours when you push it | |
| 13:11:20 | bauzas | thanks | |
| 13:26:48 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Remove __unicode__() from nova unit test Exception https://review.opendev.org/c/openstack/nova/+/769894 | |
| 14:05:51 | openstackgerrit | Lee Yarwood proposed openstack/nova master: Add regression test for bug #1908075 https://review.opendev.org/c/openstack/nova/+/766976 | |
| 14:05:54 | openstackgerrit | Lee Yarwood proposed openstack/nova master: fup: Merge duplicate volume attachment checks https://review.opendev.org/c/openstack/nova/+/773380 | |
| 14:05:54 | openstackgerrit | Lee Yarwood proposed openstack/nova master: api: Reject volume attach requests when an active bdm exists https://review.opendev.org/c/openstack/nova/+/768472 | |
| 14:05:54 | openstack | bug 1908075 in OpenStack Compute (nova) "Nova allows a non-multiattach volume to be attached to multiple instances *if* its volume state is reset by an admin" [Undecided,In progress] https://launchpad.net/bugs/1908075 - Assigned to Lee Yarwood (lyarwood) | |
| 14:05:56 | lyarwood | stephenfin: ^ updated, would you mind hitting the changes below that fup as well? | |
| 14:37:03 | openstackgerrit | Ghanshyam proposed openstack/placement master: DNM: testing direct l-c https://review.opendev.org/c/openstack/placement/+/773813 | |
| 14:44:40 | lyarwood | sean-k-mooney: ^ sorry forgot to update the unit tests in that fist change | |
| 14:44:54 | lyarwood | argh git-review is still slow | |
| 14:45:15 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP nova-manage: Add commands for managing instance machine type https://review.opendev.org/c/openstack/nova/+/769548 | |
| 14:45:15 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP libvirt: Record the machine_type of instances in system_metadata https://review.opendev.org/c/openstack/nova/+/767533 | |
| 14:45:16 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP nova-status: Add hw_machine_type check for libvirt instances https://review.opendev.org/c/openstack/nova/+/770643 | |
| 14:55:32 | gibi | lyarwood, stephenfin: do we need both? https://review.opendev.org/c/openstack/nova/+/773727 https://review.opendev.org/c/openstack/nova/+/769920 | |
| 14:56:17 | stephenfin | Yes, I think so. The fixtures proves the stubbing isn't complete and would be useful even when it is to prevent regressions | |
| 14:56:40 | lyarwood | yeah what stephenfin said, already has shown a few things we missed AFAICT | |
| 15:04:35 | gibi | thanks | |
| 15:23:55 | gibi | I'm +2 on the fairly simple libvirt metadata feature https://review.opendev.org/c/openstack/nova/+/750552 | |
| 15:24:35 | gibi | so if some core has time then it is an easy win | |
| 15:26:35 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Add network and utils methods for getting routed networks and segments https://review.opendev.org/c/openstack/nova/+/773976 | |
| 15:26:36 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: WIP: Add a routed networks scheduler pre-filter https://review.opendev.org/c/openstack/nova/+/749068 | |
| 15:27:36 | bauzas | gibi: lemme look | |
| 15:28:03 | bauzas | gibi: btw. thanks for continuing to review the routed networks series | |
| 15:28:16 | gibi | bauzas: thanks | |
| 15:28:27 | bauzas | fwiw, I'm pretty done, just the last change needs to be having UTs and docs | |
| 15:33:09 | gibi | bauzas: ack, I will continue looking at it, actaully the self -1 made me stop so it is good that you stated now that it is basically ready | |
| 15:33:29 | bauzas | gibi: yeah I needed to add UTs | |
| 15:33:32 | bauzas | now it's done | |
| 15:33:44 | bauzas | those are easy peasy | |
| 15:37:00 | gibi | :) | |
| 15:44:31 | bauzas | gibi: concerns with reliability of the guest metadata information in https://review.opendev.org/c/openstack/nova/+/750552 | |
| 15:45:48 | sean-k-mooney | bauzas: reliablity? | |
| 15:46:03 | sean-k-mooney | this is an internal debug info | |
| 15:46:24 | sean-k-mooney | so if its a little out of sync i think its ok | |
| 15:47:24 | bauzas | sean-k-mooney: well, if so, we don't need it | |
| 15:47:45 | sean-k-mooney | we dont need it but it does make debuging from logs simpler | |
| 15:47:45 | bauzas | operators could get their infos by other means, right? | |
| 15:48:01 | sean-k-mooney | they could but this would be useful for us reading sosreports | |
| 15:48:06 | bauzas | sean-k-mooney: right, but then we need it to be reliable | |
| 15:48:07 | sean-k-mooney | where we cant | |
| 15:48:17 | sean-k-mooney | ya | |