| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-03-10 | |||
| 17:47:26 | sean-k-mooney | and then we backlevel before sending | |
| 17:47:27 | bauzas | it's not just the compute | |
| 17:47:34 | dansmith | sean-k-mooney: the rules are similar to the rpc interface, but you have to be able to support both version of the object across the boundary.. compute has to support the 1.x object in case another compute sends it (although likely not the case for compute_node), | |
| 17:47:38 | bauzas | or the object being passed over the wire | |
| 17:47:39 | dansmith | and of course conductor has to be able to support both | |
| 17:48:04 | bauzas | and other services also have to understand the new minimum version | |
| 17:48:08 | sean-k-mooney | yep that is more or less what i was assuming regarding supproting both | |
| 17:48:14 | bauzas | if they get the object | |
| 17:48:23 | bauzas | I mean, if they rehydrate it | |
| 17:48:56 | dansmith | yup | |
| 17:48:59 | bauzas | sean-k-mooney: which ovo object are you considering to bump ? | |
| 17:49:11 | sean-k-mooney | its somewhat bounded now that we have the check for min compute service version now | |
| 17:49:14 | bauzas | computenode is also used by a lot of services, hence my concern | |
| 17:49:28 | sean-k-mooney | e.g. that it now enforce n+1 max delta | |
| 17:49:49 | sean-k-mooney | bauzas: well im not nessisaly suggesting we do that one next cycle | |
| 17:49:58 | sean-k-mooney | just trying to figure out when we are allowed too | |
| 17:50:03 | bauzas | yup, gotcha | |
| 17:50:22 | bauzas | should we now discuss about raising the API minimums, just for the fun ? | |
| 17:50:33 | bauzas | REST* APIs | |
| 17:50:43 | bauzas | :) | |
| 17:50:58 | sean-k-mooney | well we have a todo to stop declaring the numa_toplogy object as a stign adn declar it as an ovo filed | |
| 17:50:59 | dansmith | bauzas: it is, but not passed around between them as much right? | |
| 17:51:01 | sean-k-mooney | form like juno | |
| 17:51:26 | sean-k-mooney | bauzas: so remove in 2.0 came up when i review artoms patchs | |
| 17:51:38 | sean-k-mooney | which is why it was on my mind since we addded anouther one of those comments | |
| 17:52:05 | bauzas | dansmith: trying to think about it | |
| 17:52:18 | bauzas | dansmith: as a nested object, can't promise it isn't used | |
| 17:52:37 | dansmith | what other object includes computenode in it? | |
| 17:52:37 | sean-k-mooney | give that todo is 5 years old https://github.com/openstack/nova/blob/ab07507e5cfce6232fef373d07ff92ea704541da/nova/objects/compute_node.py#L84-L86 | |
| 17:52:48 | bauzas | good question | |
| 17:53:03 | dansmith | bauzas: you said "as a nested object" ... | |
| 17:53:45 | bauzas | dansmith: the Service object at least, I remember writing the nesting | |
| 17:54:03 | dansmith | okay but I don't think that gets passed around either | |
| 17:54:10 | bauzas | hah, and the RequestSpec one | |
| 17:54:21 | dansmith | reqspec has computenode in it? | |
| 17:54:25 | bauzas | the SchedulerRetries one | |
| 17:54:47 | bauzas | it has a list of cn objects in it | |
| 17:55:01 | sean-k-mooney | do we use that any more | |
| 17:55:07 | bauzas | which itself is nested in the spec object | |
| 17:55:09 | sean-k-mooney | i tought we dont do retreis since placment | |
| 17:55:17 | sean-k-mooney | we use alternate hosts | |
| 17:55:33 | dansmith | that seems weird because those are from different DBs, but maybe we just jam them in there to pass around for the retries? | |
| 17:56:09 | sean-k-mooney | dansmith: we should check if its still needed i think that might have been for when we used the retryfilter | |
| 17:56:12 | bauzas | sean-k-mooney: good point, it's no longer used | |
| 17:56:19 | dansmith | yeah | |
| 17:56:34 | dansmith | well, I certainly hadn't remembered that we do that, but changing it would mean more bumps, so.. | |
| 17:56:41 | dansmith | however, | |
| 17:56:51 | dansmith | I'm not sure it's really likely to be a problem | |
| 17:57:04 | dansmith | it's sourced from the control services, so if they put 1.x in there until everyone is upgraded, | |
| 17:57:06 | dansmith | it should be fine | |
| 17:57:56 | sean-k-mooney | ya again just raised this because i was trying to under stand when we can adress the Do X in next major version comments that are in some objects | |
| 17:58:07 | sean-k-mooney | many of which have been there for years at this point | |
| 18:01:40 | sean-k-mooney | https://github.com/openstack/nova/commit/1337890ace918fa2555046c01c8624be014ce2d8 was the instance one so i guess reviewing that is a good start. | |
| 18:02:46 | sean-k-mooney | oh ya i rembere the obj_relationships stuff | |
| 18:05:05 | sean-k-mooney | oh i see we had a _BaseInstance and InstanceV1 and InstanceV2 class for a cycle kind of like the proxy for rpc | |
| 18:06:18 | sean-k-mooney | then that patch actully drop V1 after we were shoudl everytin supported v2 | |
| 18:07:47 | sean-k-mooney | i vaguely remeber this happening https://github.com/openstack/nova/commit/713d8cb0777afb9fe4f665b9a40cac894b04aacb added 2 objects. | |
| 19:18:04 | openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Add 'VDPA' to 'PciDeviceType' https://review.opendev.org/c/openstack/nova/+/777481 | |
| 19:18:04 | openstackgerrit | Stephen Finucane proposed openstack/nova master: add constants for vnic type vdpa https://review.opendev.org/c/openstack/nova/+/770474 | |
| 19:18:05 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Add guest generation for vDPA https://review.opendev.org/c/openstack/nova/+/770532 | |
| 19:18:05 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Add vDPA nodedev parsing https://review.opendev.org/c/openstack/nova/+/770533 | |
| 19:18:06 | openstackgerrit | Stephen Finucane proposed openstack/nova master: pci: Add vDPA vnic to PCI request mapping and filtering https://review.opendev.org/c/openstack/nova/+/778350 | |
| 19:18:07 | openstackgerrit | Stephen Finucane proposed openstack/nova master: add hw:mlock extra spec https://review.opendev.org/c/openstack/nova/+/778347 | |
| 19:18:08 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Move PCI host device parsing to 'host' https://review.opendev.org/c/openstack/nova/+/779851 | |
| 19:18:11 | stephenfin | sean-k-mooney: there you go | |
| 19:22:49 | sean-k-mooney | thanks. ill take a look and test it out shortly. | |
| 19:31:18 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Move PCI host device parsing to 'host' https://review.opendev.org/c/openstack/nova/+/779851 | |
| 19:31:18 | openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Add 'VDPA' to 'PciDeviceType' https://review.opendev.org/c/openstack/nova/+/777481 | |
| 19:31:19 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Add vDPA nodedev parsing https://review.opendev.org/c/openstack/nova/+/770533 | |
| 19:31:20 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Add guest generation for vDPA https://review.opendev.org/c/openstack/nova/+/770532 | |
| 19:31:21 | openstackgerrit | Stephen Finucane proposed openstack/nova master: add hw:mlock extra spec https://review.opendev.org/c/openstack/nova/+/778347 | |
| 19:31:21 | openstackgerrit | Stephen Finucane proposed openstack/nova master: pci: Add vDPA vnic to PCI request mapping and filtering https://review.opendev.org/c/openstack/nova/+/778350 | |
| 20:12:54 | openstackgerrit | Merged openstack/nova master: libvirt: Stop passing around virt_type, caps https://review.opendev.org/c/openstack/nova/+/775689 | |
| 20:13:31 | openstackgerrit | Merged openstack/nova master: libvirt: Add missing type hints https://review.opendev.org/c/openstack/nova/+/775688 | |
| 20:23:14 | openstackgerrit | sean mooney proposed openstack/nova master: Support per port numa policies with SR-IOV https://review.opendev.org/c/openstack/nova/+/773792 | |
| 20:24:45 | sean-k-mooney | artom: ^ im having troble getting the socket test i addded here working https://review.opendev.org/c/openstack/nova/+/773792/10/nova/tests/functional/libvirt/test_pci_sriov_servers.py#1254 | |
| 20:31:28 | artom | sean-k-mooney, I wonder if it's because I set the 'socket' attribute on the pools in my filter method | |
| 20:31:45 | artom | And the pools that get passed to remove_device() don't have it | |
| 20:31:54 | artom | So it fails the equality test | |
| 20:32:23 | sean-k-mooney | its there in the host state object {"physnets": [], "tunneled": false}, "nova_object.changes": ["physnets", "tunneled"]}, "socket": 1}, | |
| 20:32:58 | sean-k-mooney | actully ill copy this to pastbin that not waht i wanted | |
| 20:34:04 | sean-k-mooney | http://paste.openstack.org/show/803443/ | |
| 20:34:09 | artom | del pool_keys['socket'] | |
| 20:34:09 | artom | if 'socket' in pool_keys: | |
| 20:34:09 | artom | del pool_keys['devices'] | |
| 20:34:09 | artom | Hrmm, yeah, if I add: | |
| 20:34:22 | artom | In _find_pool() | |
| 20:34:25 | artom | It makes the tests pass | |
| 20:34:28 | artom | OK, I have to run | |
| 20:34:37 | artom | Hopefully that helped a little bit, talk in ~1 hour | |
| 20:35:19 | sean-k-mooney | ok ya that might be the issue | |
| 20:36:17 | sean-k-mooney | but if it is your current patch might break neutron sriov ports? ill do some testing locally and try and see if its just a functest issue or a real one | |
| 20:37:34 | sean-k-mooney | the other polices do work however so i dont know it might just be an edgecase i need to fix to make both work correctly | |
| 20:38:07 | artom | I suspect it's a logic error in my patches | |
| 20:38:17 | artom | Though I do wonder how come my own functional tests passed...? | |
| 21:15:44 | sean-k-mooney | artom: i get the error on real hardware too. | |
| 21:31:08 | artom | sean-k-mooney, I wonder if if has something to do with VFs and PFs? | |
| 21:34:01 | sean-k-mooney | im not sure | |
| 21:34:12 | sean-k-mooney | did you test your code on real hardware | |
| 21:35:47 | sean-k-mooney | hum if i do some hacking in /sys to put all my nics on numa 0 i can boot again | |