| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-09-02 | |||
| 09:19:36 | stephenfin | *Not | |
| 09:19:56 | alex_xu | oh, totally forget that option | |
| 09:20:17 | stephenfin | alex_xu: nah, if all they do is upgrade their compute nodes, nothing will change. They'll still be reporting and consuming VCPUs | |
| 09:20:55 | openstackgerrit | Gorka Eguileor proposed openstack/nova master: Use os-brick locking for volume attach and detach https://review.opendev.org/614190 | |
| 09:20:57 | stephenfin | Things only change when the operator sets cpu_dedicated_set or cpu_shared_set on the controller, and set the as-yet-unnamed option on the controller | |
| 09:21:41 | stephenfin | when 'cpu_dedicated_set' is set, the compute node will start reporting PCPUs and transform existing inventory of pinned instances (if any) from VCPU to PCPU | |
| 09:21:50 | alex_xu | right | |
| 09:21:52 | stephenfin | not before that though | |
| 09:22:07 | alex_xu | stephenfin: so we still have that workaround option, in case people need it | |
| 09:23:31 | stephenfin | alex_xu: Yes, see https://review.opendev.org/#/c/671801/32/nova/conf/workarounds.py | |
| 09:23:42 | stephenfin | I'm going to have to rename and move that though, of course | |
| 09:24:35 | bauzas | stephenfin: you got hit by a merge conflict FWIW | |
| 09:24:36 | bauzas | https://review.opendev.org/#/c/671793/16 | |
| 09:24:45 | stephenfin | yeah, I'll tackle that shortly | |
| 09:24:48 | stephenfin | expecting a lot of those | |
| 09:25:54 | alex_xu | stephenfin: ok, if the operator wants to upgrade the host os, then he can rolling upgrade the host os and nova-compute, without PCPU translating and cpu_dedicated_set. After all the nodes upgrade the new os and new nova-compute, then update the cpu_dedicated_set, to reshape the inventory, is that correct? | |
| 09:26:11 | stephenfin | yes | |
| 09:26:19 | alex_xu | so cool | |
| 09:26:44 | alex_xu | stephenfin: we really document those somewhere, like release note upgrade section? | |
| 09:27:28 | stephenfin | alex_xu: I've started here https://review.opendev.org/#/c/671793/16/releasenotes/notes/cpu-resources-d4e6a0c12681fa87.yaml | |
| 09:27:32 | stephenfin | But I'll add more as I go | |
| 09:27:38 | bauzas | alex_xu: yeah that's the plan | |
| 09:27:41 | alex_xu | stephenfin: cool | |
| 09:27:49 | bauzas | alex_xu: nothing changes unless the operator triggers it | |
| 09:28:07 | alex_xu | bauzas: yea, that is much simpler | |
| 09:28:08 | bauzas | but when they trigger, then it'll be a reshape | |
| 09:28:27 | bauzas | stephenfin: and that also implies we need to provide some nova-status check command | |
| 09:28:37 | bauzas | for upgrading to Unicorn | |
| 09:28:42 | alex_xu | stephenfin: also something fun https://review.opendev.org/#/c/674895/22/nova/virt/libvirt/driver.py, due to I updated a wrong option | |
| 09:29:18 | sean-k-mooney | alex_xu: dsid stephenfin answer your os upgrade question | |
| 09:29:31 | alex_xu | sean-k-mooney: yes, perfectly answered | |
| 09:30:59 | gibi | bauzas, stephenfin: hi! there are +2s from mriedem on the next couple patches in the support-move-ops-with-qos-ports series. starting here https://review.opendev.org/#/c/655110/ | |
| 09:31:23 | bauzas | gibi: yeah, I was planning to put my eyes on your series after stephenfin's one | |
| 09:31:32 | gibi | bauzas: cool, thanks | |
| 09:44:39 | yaawang | stephenfin: alex_xu Hi, I had update the patches of vCPU models selection, can you take a look if you have time :) https://review.opendev.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/cpu-model-selection | |
| 09:45:27 | alex_xu | yaawang: got it, will check it again | |
| 09:45:49 | yaawang | alex_xu: thk :) | |
| 09:45:57 | alex_xu | yaawang: np | |
| 10:06:16 | openstackgerrit | sean mooney proposed openstack/nova master: Libvirt: report storage bus traits https://review.opendev.org/666914 | |
| 10:06:17 | openstackgerrit | sean mooney proposed openstack/nova master: Add transform_image_metadata request filter https://review.opendev.org/665775 | |
| 10:06:17 | openstackgerrit | sean mooney proposed openstack/nova master: libvirt: use domain capabilities to get supported device models https://review.opendev.org/666915 | |
| 10:20:04 | openstackgerrit | Brin Zhang proposed openstack/nova-specs master: Support set port vnic type to create server https://review.opendev.org/672400 | |
| 10:25:04 | openstackgerrit | Brin Zhang proposed openstack/nova-specs master: Support set port vnic type to create server https://review.opendev.org/672400 | |
| 10:51:19 | openstackgerrit | Stephen Finucane proposed openstack/nova master: trivial: Rewrap definitions of 'NUMACell' https://review.opendev.org/674395 | |
| 10:51:20 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Start checking compute usage in functional tests https://review.opendev.org/678902 | |
| 10:51:20 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Simplify 'fakelibvirt.HostInfo' object https://review.opendev.org/678861 | |
| 10:51:21 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Fold in argument to '_update_provider_tree_for_vgpu' https://review.opendev.org/676729 | |
| 10:51:21 | openstackgerrit | Stephen Finucane proposed openstack/nova master: objects: Rename 'fields' import to 'obj_fields' https://review.opendev.org/674103 | |
| 10:51:22 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: '_get_(v|p)cpu_total' to '_get_(v|p)cpu_available' https://review.opendev.org/672693 | |
| 10:51:22 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Start reporting PCPU inventory to placement https://review.opendev.org/671793 | |
| 10:51:23 | openstackgerrit | Stephen Finucane proposed openstack/nova master: libvirt: Start reporting 'HW_CPU_HYPERTHREADING' trait https://review.opendev.org/675571 | |
| 10:51:23 | openstackgerrit | Stephen Finucane proposed openstack/nova master: hardware: Differentiate between shared and dedicated CPUs https://review.opendev.org/671800 | |
| 10:51:24 | openstackgerrit | Stephen Finucane proposed openstack/nova master: Add reshaper for PCPU https://review.opendev.org/674895 | |
| 10:51:24 | openstackgerrit | Stephen Finucane proposed openstack/nova master: Add support for translating CPU policy extra specs, image meta https://review.opendev.org/671801 | |
| 10:54:26 | openstackgerrit | Chris Dent proposed openstack/nova master: Use microversion in put allocations in test_report_client https://review.opendev.org/679622 | |
| 11:10:52 | stephenfin | alex_xu, bauzas: Still working on changing the scheduler config option and adding resize functional tests for cpu-resources, but I've moved all the trivial patches up to the front of the series, if you fancy taking a look (starts here https://review.opendev.org/#/c/678861/) | |
| 11:11:09 | stephenfin | gibi: Yeah, that's on my list for today (t) | |
| 11:11:11 | stephenfin | *(y) | |
| 11:11:37 | gibi | stephenfin: thanks | |
| 11:49:33 | openstackgerrit | Chris Dent proposed openstack/nova master: Tune up db.instance_get_all_uuids_by_hosts https://review.opendev.org/679627 | |
| 11:59:24 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Nice to have test coverage for If1f465112b8e9b0304b8b5b864b985f72168d839 https://review.opendev.org/679629 | |
| 12:07:35 | aspiers | stephenfin: thanks for the reviews. BTW https://review.opendev.org/#/c/636334/ is missing W+1 | |
| 12:08:14 | aspiers | stephenfin: I'm working on adding the checks you requested, for live migration / suspend etc. | |
| 12:21:03 | kashyap | aspiers: Nice, seems like the SEV work is almost merged. | |
| 12:21:14 | aspiers | kashyap: Yes, pretty close now :) | |
| 12:21:56 | aspiers | It's almost a year since I uploaded the Stein spec :-o | |
| 12:22:13 | aspiers | So this has been a year-long project or more already | |
| 12:23:19 | kashyap | It required a lot of 'supporting infra' work, which is also useful for other features (machine types rework; domCapabilities, etc) | |
| 12:23:28 | aspiers | Right | |
| 12:23:42 | aspiers | I'm glad it turned out to have other uses. Didn't anticipate that originally. | |
| 12:24:17 | kashyap | aspiers: Unrelated - SEV _requires_ 'q35', isn't it? | |
| 12:25:28 | aspiers | kashyap: http://eavesdrop.openstack.org/irclogs/%23openstack-nova/%23openstack-nova.2019-07-30.log.html#t2019-07-30T09:21:35 | |
| 12:25:30 | aspiers | ;-) | |
| 12:26:03 | kashyap | aspiers: Haha, my memory is like the bumblebee's -- can't recall anything more than past 7 seconds | |
| 12:26:07 | aspiers | XD | |
| 12:26:20 | aspiers | Similar thing happened to me this morning | |
| 12:27:20 | kashyap | aspiers: I asked because, the next question I was looking for was: is there a check in the code where: "if SEV && !q35 --> throw an error" | |
| 12:27:52 | sean-k-mooney | kashyap: you could only do that check on the compute node currenlty | |
| 12:28:27 | sean-k-mooney | i guess you could push it to placemetn by not reporting the sev resouces if the machine type is set to pc | |
| 12:28:27 | kashyap | sean-k-mooney: Right, it matters only on compute nodes anyway. | |
| 12:28:45 | sean-k-mooney | kashyap: my point is it woudl cause a late reschdule | |
| 12:28:46 | kashyap | sean-k-mooney: What I was wondering is, _if_ such a thing exists already | |
| 12:28:58 | aspiers | Guys, no need to speculate ;-) | |
| 12:29:06 | aspiers | https://review.opendev.org/#/c/644565/49/nova/virt/libvirt/driver.py@5104 | |
| 12:29:42 | sean-k-mooney | im not sure overriding the machine type is a good idea | |
| 12:30:07 | aspiers | We could bail there instead | |
| 12:30:37 | kashyap | Oh, at _what_ point does the override happens? | |
| 12:30:52 | aspiers | kashyap: At *that* point ^^^ | |
| 12:31:03 | sean-k-mooney | its in the driver when its generating the xml | |
| 12:31:07 | kashyap | Right. Override can cause a potential surprise for the operator, doesn't it? | |
| 12:31:25 | sean-k-mooney | kashyap: yes and maybe live migration issues | |
| 12:31:32 | aspiers | Live migration is unsupported | |
| 12:31:48 | aspiers | See above message I wrote to stephenfin ^^^ | |
| 12:32:06 | kashyap | aspiers: Ah, nod, noticed. | |
| 12:32:08 | aspiers | But it will probably be supported eventually | |
| 12:32:14 | sean-k-mooney | also you can request teh machine type in the image https://github.com/openstack/glance/blob/master/etc/metadefs/compute-libvirt-image.json#L34-L38 | |
| 12:32:43 | aspiers | sean-k-mooney: Yes, that's why image_meta is a parameter to get_machine_type_no_default() in the code I just linked | |
| 12:33:08 | kashyap | aspiers: Thinking out loud: do we need an additional function for handling SEV-specific quirks for machine types, instead of squeezing it into the generic guest_machine_type()? | |
| 12:33:10 | aspiers | https://review.opendev.org/#/c/644565/49/nova/virt/libvirt/utils.py@539 | |
| 12:33:25 | sean-k-mooney | yes but overrideing an explcit machine_type in the image is not correct | |