| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2022-08-19 | |||
| 18:37:30 | gibi | anyhow my brain is toasted and my wife just arrived so I have to log off. don't think too much about this issue it is late friday anyhow | |
| 18:37:44 | gibi | see you all on Monday | |
| 18:37:49 | sean-k-mooney | ok | |
| 18:37:54 | sean-k-mooney | im going to call it a day too | |
| 18:38:06 | gibi | have a nice weekend | |
| 18:38:36 | sean-k-mooney | you too ill try and look at this on monday after i rebase the vdpa patches | |
| 18:58:22 | opendevreview | Dan Smith proposed openstack/nova-specs master: WIP: Robustify Compute Node Hostnames https://review.opendev.org/c/openstack/nova-specs/+/853837 | |
| 18:58:36 | dansmith | sean-k-mooney: artom: ^ | |
| 18:59:08 | dansmith | that's a big chunk of work, which we may never do, but I thought it was probably good to document some of the things we could/should do to make this better | |
| 18:59:23 | dansmith | either to point to and say "see, too big, never going to happen" or the opposite | |
| 18:59:49 | dansmith | I've been thinking about the first work item for a long time and I think we should probably do that for safety even if we don't do any of the rest of it | |
| 19:02:33 | artom | That's kind of in the same vein as https://bugzilla.redhat.com/show_bug.cgi?id=1965419, which came up before when another customer renamed their hosts | |
| 19:02:44 | artom | Now that I think about it, it may have been that exact same KCS | |
| 19:02:51 | artom | Because it was a 10 -> 13 FFU | |
| 19:02:58 | artom | Sorry, leaking downstream here | |
| 19:04:21 | sean-k-mooney | i mean we were broken in 16.1 requirenign neutron ot add a new config option | |
| 19:04:23 | sean-k-mooney | https://bugzilla.redhat.com/show_bug.cgi?id=1900500 | |
| 19:04:51 | sean-k-mooney | resource_provider_default_hypervisor | |
| 19:06:27 | sean-k-mooney | https://github.com/openstack/neutron/commit/577217c52d677ba35ca78b87c06302d506f66ff9 and https://github.com/openstack/neutron/commit/ddf0fef28b7095724c8ba27f3275d0dad2252251 | |
| 19:06:45 | sean-k-mooney | were added to neutorn to work aorund changes in ooo | |
| 19:30:15 | opendevreview | Dan Smith proposed openstack/nova-specs master: WIP: Robustify Compute Node Hostnames https://review.opendev.org/c/openstack/nova-specs/+/853837 | |
| #openstack-nova - 2022-08-20 | |||
| 06:18:38 | opendevreview | Takashi Natsume proposed openstack/nova master: Add missing descriptions in HACKING.rst https://review.opendev.org/c/openstack/nova/+/853054 | |
| 06:18:59 | opendevreview | Takashi Natsume proposed openstack/nova master: doc: Update a PTL guide https://review.opendev.org/c/openstack/nova/+/852275 | |
| 06:35:52 | opendevreview | Merged openstack/nova stable/train: Don't unset Instance.old_flavor, new_flavor until necessary https://review.opendev.org/c/openstack/nova/+/751367 | |
| 09:06:23 | opendevreview | Merged openstack/nova master: doc: Update a PTL guide https://review.opendev.org/c/openstack/nova/+/852275 | |
| 15:37:55 | opendevreview | Merged openstack/nova master: enable blocked VDPA move operations https://review.opendev.org/c/openstack/nova/+/832330 | |
| 21:21:04 | opendevreview | Rico Lin proposed openstack/nova master: Add locked_memory extra spec and image property https://review.opendev.org/c/openstack/nova/+/778347 | |
| 21:21:05 | opendevreview | Rico Lin proposed openstack/nova master: libvirt: Add vIOMMU device to guest https://review.opendev.org/c/openstack/nova/+/830646 | |
| 21:21:05 | opendevreview | Rico Lin proposed openstack/nova master: Add traits for viommu model https://review.opendev.org/c/openstack/nova/+/844507 | |
| #openstack-nova - 2022-08-21 | |||
| 06:06:07 | opendevreview | Takashi Natsume proposed openstack/nova master: Fix a deprecation warning about threading.Thread https://review.opendev.org/c/openstack/nova/+/853869 | |
| #openstack-nova - 2022-08-22 | |||
| 03:10:32 | opendevreview | zhangqing proposed openstack/nova stable/train: remove redundant statement in SecurityGroupDefaultRulesController https://review.opendev.org/c/openstack/nova/+/853901 | |
| 03:11:24 | opendevreview | zhangqing proposed openstack/nova stable/train: remove redundant statement https://review.opendev.org/c/openstack/nova/+/853901 | |
| 07:29:58 | Uggla | Hello o/ | |
| 07:37:57 | gibi | Uggla: welcome back | |
| 07:38:16 | Uggla | Hi gibi ! | |
| 07:51:50 | gibi | fyi folks, we have "Non-client library freeze: August 25th, 2022 (R-6 week)" which is this week. So if you have anything depending on os-traits, os-resource-classes, os-vif, etc then those dependencies needs to land this week | |
| 08:06:14 | opendevreview | Rajesh Tailor proposed openstack/nova master: Fix rescue volume-based instance https://review.opendev.org/c/openstack/nova/+/852737 | |
| 08:08:33 | opendevreview | Jan Hartkopf proposed openstack/nova master: add support for updating server's user_data https://review.opendev.org/c/openstack/nova/+/816157 | |
| 08:52:19 | gibi | sean-k-mooney[m]: this is the code that causes the missing PCI device failure https://github.com/openstack/nova/blob/99dd3f75cd23a4ff419c20826f5abfcfed417889/nova/pci/manager.py#L485-L503 in https://review.opendev.org/c/openstack/nova/+/853835 (I needed fresh brains for it to find) | |
| 09:10:12 | sean-k-mooney | ah right just getting started but ill pull your patch and see if i can reporduce locally and take a look | |
| 09:11:40 | gibi | I need to refactor that piece of code and move it to the Instance ovo | |
| 09:13:21 | sean-k-mooney | i was thinking about this since we last spoke. is there any reason not to have the consturctor generate a uuid automitically when we constuct the pci request objects | |
| 09:13:41 | sean-k-mooney | since we will now be creating these on both the neuton and non nueutron path | |
| 09:14:03 | gibi | sean-k-mooney: yes, that is a good point too | |
| 09:14:07 | gibi | sean-k-mooney: I will do that | |
| 09:14:33 | sean-k-mooney | do you recall what test failed? | |
| 09:14:48 | sean-k-mooney | i guess it will be in the zuul logs | |
| 09:15:17 | sean-k-mooney | test_cold_migrate_server_with_pci | |
| 09:17:38 | gibi | yes that one | |
| 09:18:29 | gibi | and it fails as when the libvirt driver tries to get the PciDevice objects of the instance to generate the xml it gets [] as the above linked piece of code assumes request_id = None means flavor based PCI request | |
| 09:18:33 | gibi | and I break that assumption | |
| 09:18:54 | sean-k-mooney | we have a function to figure that out | |
| 09:19:03 | sean-k-mooney | which should be used instead | |
| 09:19:20 | sean-k-mooney | three is a source atribute | |
| 09:19:39 | sean-k-mooney | https://github.com/openstack/nova/blob/master/nova/objects/instance_pci_requests.py#L48-L54 | |
| 09:20:00 | gibi | exactly | |
| 09:22:52 | Uggla | gibi, sean-k-mooney any objection to rename the options as proposed by stephen here : https://review.opendev.org/c/openstack/python-openstackclient/+/831902/comments/9b4913bd_cee0ed57 | |
| 09:23:33 | sean-k-mooney | i havent looked at it but ill check quickly | |
| 09:23:54 | sean-k-mooney | oh he wants to use the ciso no prefix | |
| 09:23:59 | sean-k-mooney | i personaly hate that | |
| 09:26:44 | gibi | we have examples with --no-* already in the client and the flag's doc clearly states that this means unpin so I'm OK | |
| 09:27:03 | sean-k-mooney | we do but we also have set and unset i belive | |
| 09:27:05 | sean-k-mooney | just checkign that now | |
| 09:27:40 | sean-k-mooney | yes opnestack flavor set and unset | |
| 09:27:59 | gibi | but that is not a flag but a subcommand | |
| 09:29:10 | sean-k-mooney | yes but i think openstack server unshleve --unset-az | |
| 09:29:16 | sean-k-mooney | would make sense | |
| 09:30:15 | gibi | I have nothing against that either | |
| 09:30:27 | gibi | stephenfin: are you around? | |
| 09:40:08 | sean-k-mooney | gibi: so printing the xml there are not hostdev elements which is why its getting None for elem | |
| 09:40:15 | gibi | yes | |
| 09:40:21 | gibi | I figured it out this morning | |
| 09:40:36 | sean-k-mooney | i find that very odd that adding the request id woudl have resulted in that | |
| 09:41:04 | gibi | nova uses PciDevice.request_id == None to signal flavor based PCI devs | |
| 09:41:09 | sean-k-mooney | so yes modifying get_instance_pci_devs | |
| 09:41:19 | sean-k-mooney | is likely the way to go | |
| 09:41:21 | gibi | yes | |
| 09:41:28 | sean-k-mooney | well we do in some placees but not all | |
| 09:41:53 | gibi | yes | |
| 09:47:30 | sean-k-mooney | oh its becasue we have the PciDevice object not the pci request objects here | |
| 09:48:06 | sean-k-mooney | i was going to just add or device.source == objects.InstancePCIRequest.FLAVOR_ALIAS | |
| 09:48:22 | sean-k-mooney | but device is not an InstancePciREquest object | |
| 09:49:04 | sean-k-mooney | we have the pci request too | |
| 09:50:43 | gibi | yes | |
| 09:51:11 | gibi | there is somewhere a generic code that does PciDevice.request_id = InstancePCIRequest.request_id | |
| 09:51:18 | gibi | which is I think correct | |
| 09:52:16 | sean-k-mooney | that basically waht im trying locally | |
| 09:52:50 | sean-k-mooney | im doing a set comprehention to get the flavor request ids and then checking if the current device is in that when request_id is none | |
| 09:54:08 | sean-k-mooney | https://paste.opendev.org/show/bqJUPAfRGibXn8mslgdU/ | |
| 09:54:09 | gibi | I added https://paste.opendev.org/show/b1nvksM7Fw7F4w5vnDKw/ to Instance ovo and replaced the get_instance_pci_devs calls with it and it seems to work | |
| 09:54:12 | sean-k-mooney | that seams to work | |
| 09:55:30 | sean-k-mooney | you could do that but you can do it in the existing fucntion without changing the signiture or moving it | |
| 09:56:05 | gibi | yeah but 1) I have the implicit not-providing-request-id meaning give me the flavor based CPI devs | |
| 09:56:10 | gibi | s/have/hate/ | |
| 09:56:34 | gibi | I want to make the query explicit by the caller | |
| 09:56:49 | sean-k-mooney | yep i get that | |
| 09:56:56 | gibi | 2) also request_id == 'all' is /o\ | |
| 09:57:15 | sean-k-mooney | the duality because of the fact that this code predated neutron sriov i think | |
| 09:57:26 | sean-k-mooney | and source | |
| 09:57:31 | gibi | yes, it is old code, it served wll but I retired it now :D | |