| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2024-10-29 | |||
| 16:34:47 | sean-k-mooney | well frontend qos policies are use by the hypervior (libvirt in this case) | |
| 16:35:03 | sean-k-mooney | to enforce it where as backend qos polcies are done on the stroage backend | |
| 16:35:04 | tkajinam | but that qos is associated with flavor, right ? | |
| 16:35:09 | sean-k-mooney | no | |
| 16:35:16 | sean-k-mooney | its assocated witht eh volume | |
| 16:35:30 | sean-k-mooney | we also have the ablity to do it via the flavor for nova created block devices | |
| 16:35:38 | sean-k-mooney | using the quota:* extra specs | |
| 16:36:12 | sean-k-mooney | i belive the current state is you can configure hard limits today | |
| 16:36:15 | sean-k-mooney | but not burst | |
| 16:36:27 | sean-k-mooney | and the brust limist are what they want to enabel | |
| 16:36:34 | sean-k-mooney | but that was my question to the cinder folk | |
| 16:36:50 | tkajinam | ok | |
| 16:36:51 | sean-k-mooney | to have them confirm what is and is not expected to be requestable vai cinder today | |
| 16:36:57 | tkajinam | I thought the qos specs are not propagated via connection info but it's set in storage backend | |
| 16:37:08 | tkajinam | I'll recheck it and fix my comments in that spec if needed | |
| 16:37:28 | sean-k-mooney | you set them on the voluem type and im not sure how they are propegated to nova | |
| 16:37:37 | sean-k-mooney | so this is where we need input form the cider folk | |
| 16:38:00 | sean-k-mooney | i.e are tehy on teh attachmetn? conection info? do we need to look them up ectra | |
| 16:38:05 | bauzas | can we end the meeting ? | |
| 16:38:12 | sean-k-mooney | yep i think so | |
| 16:38:15 | tkajinam | yeah we can discuss details in the spec | |
| 16:38:17 | sean-k-mooney | this can be an async dicussion | |
| 16:38:41 | bauzas | cool | |
| 16:38:46 | bauzas | then thanks all | |
| 16:38:49 | opendevmeet | Log: https://meetings.opendev.org/meetings/nova/2024/nova.2024-10-29-16.00.log.html | |
| 16:38:49 | opendevmeet | Minutes (text): https://meetings.opendev.org/meetings/nova/2024/nova.2024-10-29-16.00.txt | |
| 16:38:49 | opendevmeet | Minutes: https://meetings.opendev.org/meetings/nova/2024/nova.2024-10-29-16.00.html | |
| 16:38:49 | opendevmeet | Meeting ended Tue Oct 29 16:38:49 2024 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | |
| 16:38:49 | bauzas | #endmeeting | |
| 16:42:09 | sean-k-mooney | tkajinam: for what its worht i have only ever used cinder qos with ceph using the backend mode so i have nerver personally tried to use the frontend mode | |
| 16:42:16 | sean-k-mooney | i think you can use it with cinder-lvm | |
| 16:42:25 | sean-k-mooney | but im fuzzy on the details | |
| 16:43:45 | sean-k-mooney | thats a long way to say i think this coudl be configured and enabeld in tempest or whitebox tests and validated end to end in 1st party ci. | |
| 16:44:52 | sean-k-mooney | so im not really conserned about the testablity ro the libvirt xml generaion just the api contract between nova/cinder | |
| 16:46:11 | sean-k-mooney | i woudl hope the exisating cidner tempest plugins or core tempest could verify frontend qos and the proposer could jsut extend that (in adddtion to unit/fucntional tests) | |
| 16:46:15 | tkajinam | https://github.com/openstack/cinder/blob/master/cinder/volume/driver.py#L2698-L2756 | |
| 16:47:15 | tkajinam | ahh, wait | |
| 16:47:16 | tkajinam | https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L1831-L1832 | |
| 16:47:21 | tkajinam | ok I understand what you mean | |
| 16:48:00 | sean-k-mooney | right cinder has an enum https://github.com/openstack/cinder/blob/master/cinder/volume/qos_specs.py#L33 for diffent type of qos | |
| 16:48:34 | sean-k-mooney | and as you foudn the generic manager code shoudl include the qos if its front-end or both | |
| 16:48:58 | sean-k-mooney | it looks like its part of the connection info | |
| 16:49:16 | sean-k-mooney | in a optional qos_specs section | |
| 16:49:23 | tkajinam | yeah | |
| 16:49:42 | sean-k-mooney | looking at https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L1836-L1839 | |
| 16:50:00 | sean-k-mooney | im not sure they the burst values they are propsoing are supproted by cinder | |
| 16:50:16 | sean-k-mooney | so i think there shoudl be a cinder spec first to extend the set of qos types | |
| 16:50:23 | sean-k-mooney | and then we woudl just enable them | |
| 16:50:58 | sean-k-mooney | the cinder team may be ok just using the nova spec for both but we would need them to say one way or another | |
| 16:54:22 | tkajinam | sean-k-mooney, quickly looking at cinder code it seems cinder has no mechanism to validate keys in qos specs | |
| 16:54:43 | tkajinam | the documentation even describes the "delay" which is not actually used IIUC https://docs.openstack.org/api-ref/block-storage/v3/index.html#id768 | |
| 16:54:59 | sean-k-mooney | it filters them when puting them in the connection info | |
| 16:55:19 | sean-k-mooney | so it may not filter for backend qos polciies btu it does for frontend ones | |
| 16:56:16 | tkajinam | I mean there is not filtering mechanism in api layer, so what we may need to update is the filtering logic in manager.py which you pointed | |
| 16:56:45 | sean-k-mooney | yes which cinder may or may not consider an api change | |
| 16:57:16 | sean-k-mooney | so they may be happy with jsut the nova spec and say a bug on there side or they may ask for a spec on both sides if they consider it an api change | |
| 16:57:29 | sean-k-mooney | i posted this on the spec more or less | |
| 16:57:49 | sean-k-mooney | although not in those exact words | |
| 16:58:35 | tkajinam | yeah I saw these | |
| 16:59:33 | sean-k-mooney | if you think any addtional context woudl help please reply so the author can condier that. are you overall infaovr or agasitn this feature? | |
| 17:05:02 | tkajinam | sean-k-mooney, I now think that's a good idea. My -1 came from the fact that I overlooked that frontend qos handling | |
| 17:05:57 | opendevreview | Merged openstack/nova master: [ovo]Add igb value to hw_vif_model image property https://review.opendev.org/c/openstack/nova/+/928456 | |
| 17:20:21 | opendevreview | Merged openstack/nova master: [libvirt]Support hw_vif_model = igb https://review.opendev.org/c/openstack/nova/+/928584 | |
| 17:20:31 | opendevreview | Merged openstack/nova master: [doc]Developer doc about PCI and SRIOV testing https://review.opendev.org/c/openstack/nova/+/928834 | |
| 17:20:39 | sean-k-mooney | gibi: :) ^ | |
| 17:23:06 | opendevreview | Takashi Kajinami proposed openstack/nova master: Detect AMD SEV-ES support https://review.opendev.org/c/openstack/nova/+/925685 | |
| 17:23:06 | opendevreview | Takashi Kajinami proposed openstack/nova master: Migrate MEM_ENCRYPTION_CONTEXT from root provider https://review.opendev.org/c/openstack/nova/+/921814 | |
| 17:23:07 | opendevreview | Takashi Kajinami proposed openstack/nova master: libvirt: Launch instances with SEV-ES memory encryption https://review.opendev.org/c/openstack/nova/+/926106 | |
| 17:23:07 | opendevreview | Takashi Kajinami proposed openstack/nova master: Add hw_mem_encryption_model image property https://review.opendev.org/c/openstack/nova/+/927706 | |
| 17:26:58 | gibi | sean-k-mooney: \o/ | |
| 17:27:08 | opendevreview | Takashi Kajinami proposed openstack/nova master: Add hw_mem_encryption_model image property https://review.opendev.org/c/openstack/nova/+/927706 | |
| 17:27:08 | gibi | a quick win for Epoxy | |
| 17:27:08 | opendevreview | Takashi Kajinami proposed openstack/nova master: Detect AMD SEV-ES support https://review.opendev.org/c/openstack/nova/+/925685 | |
| 17:27:09 | opendevreview | Takashi Kajinami proposed openstack/nova master: libvirt: Launch instances with SEV-ES memory encryption https://review.opendev.org/c/openstack/nova/+/926106 | |
| 20:43:20 | opendevreview | Artom Lifshitz proposed openstack/nova master: WIP: pwr mgmt: power down free PCPUS when updating compute node https://review.opendev.org/c/openstack/nova/+/932926 | |
| #openstack-nova - 2024-10-30 | |||
| 07:00:46 | opendevreview | Luca Miccini proposed openstack/nova master: TEST DNM check CI https://review.opendev.org/c/openstack/nova/+/933702 | |
| 07:08:26 | opendevreview | Luca Miccini proposed openstack/nova master: TEST DNM check CI https://review.opendev.org/c/openstack/nova/+/933702 | |
| 12:33:04 | opendevreview | Balazs Gibizer proposed openstack/nova master: Route shared storage RPC to evac dest at startup https://review.opendev.org/c/openstack/nova/+/933734 | |
| 15:29:52 | opendevreview | Balazs Gibizer proposed openstack/nova master: Reproduce bug 2085975 in functional https://review.opendev.org/c/openstack/nova/+/933753 | |
| 15:29:52 | opendevreview | Balazs Gibizer proposed openstack/nova master: Route shared storage RPC to evac dest at startup https://review.opendev.org/c/openstack/nova/+/933734 | |
| 15:30:12 | sean-k-mooney | that was fast | |
| 15:30:49 | gibi | I had proper focus time | |
| 15:31:11 | sean-k-mooney | :) | |
| 15:32:29 | sean-k-mooney | skiming the repoducer the logic makes sense | |
| 15:32:49 | sean-k-mooney | ill be interested to see if that is stable in ci as we have had issues with using condition variables like that before | |
| 15:33:08 | sean-k-mooney | its seams reasonable to me at a glance however | |
| 15:34:07 | sean-k-mooney | oh the rpc_exception car is set via wrapped_check_shared_storage | |
| 15:34:26 | sean-k-mooney | i think nonlocal is only aviable in newerversion fo python but perhaps that includes 3.9 | |
| 15:35:26 | sean-k-mooney | oh no looks like that is valid form old version of python 3 | |
| 15:36:09 | sean-k-mooney | https://peps.python.org/pep-3104/ | |
| 15:50:26 | gibi | yeah we will see I can run couple of rechecks to see if it is stable | |
| 15:51:01 | gibi | without the events I cannot create a stable reproduction as I need to stop evac int he middle. So we can only have unit tests then | |
| 17:26:43 | sean-k-mooney | gibi: so regarding https://github.com/openstack/nova/blob/110849f7f972d8cb536a9c96c4cb59f06e357eeb/nova/conf/compute.py#L1087-L1104 | |
| 17:26:50 | sean-k-mooney | the help text is actully wrong | |
| 17:27:03 | sean-k-mooney | nova will not update teh network info cache on its own | |
| 17:27:21 | sean-k-mooney | it will howver react to the network-changed external event form neutron if something changes | |
| 17:27:28 | sean-k-mooney | and update the relevent port | |
| 17:29:27 | gibi | sean-k-mooney: I haven't looked at the impl of the heal code yet. But I can do that fur sure. If we want to change the default then we definitly need to change that conf doc as it recommends against it. | |
| 17:33:48 | sean-k-mooney | right it was true at teh tiem becuase ther ewere many out of tree neutron drivers in production use | |
| 17:34:03 | sean-k-mooney | that is really not the case anymore | |