| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2023-02-01 | |||
| 16:55:31 | opendevreview | Artom Lifshitz proposed openstack/python-novaclient master: Bump microversion to 2.95 https://review.opendev.org/c/openstack/python-novaclient/+/872418 | |
| 17:08:30 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Remove legacy migrations https://review.opendev.org/c/openstack/nova/+/872428 | |
| 17:08:31 | opendevreview | Stephen Finucane proposed openstack/nova master: db: Remove the legacy 'migration_version' table https://review.opendev.org/c/openstack/nova/+/872429 | |
| 17:08:42 | stephenfin | sean-k-mooney: gibi: that should fix SQLA 2.0 compat ^ | |
| 17:19:41 | stephenfin | I think it's okay to drop them completely. We've supported automatic migration to alembic since Wallaby. Antelope will be 5 releases later which spans even the biggest fast-forward upgrade interval. Also, even with FFU we expect folks to run DB upgrades on each version so | |
| 17:23:44 | opendevreview | Dan Smith proposed openstack/nova master: Check our nodes for hypervisor_hostname changes https://review.opendev.org/c/openstack/nova/+/872220 | |
| 17:23:45 | opendevreview | Dan Smith proposed openstack/nova master: Protect against a deleted node id file https://review.opendev.org/c/openstack/nova/+/872204 | |
| 17:23:45 | opendevreview | Dan Smith proposed openstack/nova master: Move comment about _destroy_evacuated_instances() https://review.opendev.org/c/openstack/nova/+/872348 | |
| 17:23:46 | opendevreview | Dan Smith proposed openstack/nova master: Abort startup if nodename conflict is detected https://review.opendev.org/c/openstack/nova/+/872432 | |
| 17:23:52 | dansmith | gdi | |
| 18:11:05 | sean-k-mooney | stephenfin: ill take a look shortly | |
| 18:37:01 | opendevreview | Sylvain Bauza proposed openstack/nova master: cpu: interfaces for managing state and governor https://review.opendev.org/c/openstack/nova/+/868236 | |
| 18:37:02 | opendevreview | Sylvain Bauza proposed openstack/nova master: libvirt: let CPUs be power managed https://review.opendev.org/c/openstack/nova/+/821228 | |
| 18:37:02 | opendevreview | Sylvain Bauza proposed openstack/nova master: WIP: enable cpus when an instance is spawning https://review.opendev.org/c/openstack/nova/+/868237 | |
| 18:37:27 | bauzas | sean-k-mooney: looks like there is a discrepancy between the guest pcpu and the numa topology blob : | |
| 18:37:33 | bauzas | https://paste.opendev.org/show/bcAxuCSeroU2VHjfqpUx/ | |
| 18:40:31 | sean-k-mooney | thats the bug i reported last week | |
| 18:40:53 | sean-k-mooney | oh the power seriese | |
| 18:41:31 | sean-k-mooney | bauzas: let me check if you are usign the right data set | |
| 18:41:32 | bauzas | nevermind, I found the rootcase | |
| 18:41:37 | sean-k-mooney | ok | |
| 18:41:50 | sean-k-mooney | by the way just so you are aware | |
| 18:42:02 | sean-k-mooney | core 0 cant generally be turnned off | |
| 18:42:07 | sean-k-mooney | its special in the kernel | |
| 18:42:26 | sean-k-mooney | more genrelaly the first core in a socket can be special in the same way | |
| 18:42:57 | bauzas | sean-k-mooney: found the discrepancy reason : https://paste.opendev.org/show/bAQTjXRRqaxXPO11HPq6/ | |
| 18:43:11 | bauzas | tl;dr: cpu_pinning property is wrong | |
| 18:43:30 | bauzas | I'll change my functest to use the pcpuset | |
| 18:43:39 | bauzas | s/use/verify | |
| 18:43:45 | sean-k-mooney | or your reading it wtong | |
| 18:44:13 | sean-k-mooney | it looks correct to me | |
| 18:44:38 | sean-k-mooney | cpu_pinning_raw={0=0,1=1,2=2,3=3,4=6} that is a dict to logical guest core to host core | |
| 18:45:16 | sean-k-mooney | so topology.cpu_pinning returned {0, 1, 2, 3, 6} | |
| 18:45:20 | bauzas | I'm confused | |
| 18:45:29 | sean-k-mooney | which are the host cores its the vm cores are pinned too | |
| 18:45:29 | bauzas | are those numbers the vcpu ones ? | |
| 18:45:41 | sean-k-mooney | the key 0-4 | |
| 18:45:54 | sean-k-mooney | are logical guest cpu cores 0-4 | |
| 18:46:12 | sean-k-mooney | the values are the ids of the host cores those vcpus are pinned too | |
| 18:46:15 | opendevreview | Sylvain Bauza proposed openstack/nova master: WIP: enable cpus when an instance is spawning https://review.opendev.org/c/openstack/nova/+/868237 | |
| 18:46:26 | bauzas | then, the guest.vcpu.cpu set is wrong | |
| 18:46:36 | bauzas | from libvirt | |
| 18:46:42 | sean-k-mooney | can you show me the libvirt xml | |
| 18:47:01 | bauzas | gimme me a sec | |
| 18:47:33 | bauzas | calling the domain I guess | |
| 18:47:58 | sean-k-mooney | you sould use the instance numa toplogy blob | |
| 18:48:26 | sean-k-mooney | that is the singel source or truth | |
| 18:51:19 | bauzas | Laptop freezed, had to reboot | |
| 18:52:37 | opendevreview | Dan Smith proposed openstack/nova master: Stable compute uuid functional tests https://review.opendev.org/c/openstack/nova/+/872441 | |
| 19:12:47 | opendevreview | Dan Smith proposed openstack/nova master: Stable compute uuid functional tests https://review.opendev.org/c/openstack/nova/+/872441 | |
| 19:35:07 | opendevreview | Sylvain Bauza proposed openstack/nova master: WIP: enable cpus when an instance is spawning https://review.opendev.org/c/openstack/nova/+/868237 | |
| 19:35:22 | bauzas | sean-k-mooney: updated ^ | |
| 19:38:22 | sean-k-mooney | bauzas: ack ill take a look tomorrwo | |
| 19:38:39 | sean-k-mooney | dansmith: ah i was going to ask but i see that there was an issue with ironic on your series | |
| 19:39:18 | dansmith | sean-k-mooney: only because I aded that missing node file check before the ironic exclusion | |
| 19:39:27 | dansmith | before I added that, it worked fine on ironic | |
| 19:39:39 | dansmith | but it's running another job with the two reversed now, should be fine, but we should wait to be sure | |
| 19:40:05 | sean-k-mooney | ack i can take a look again tomrrow | |
| 19:40:25 | sean-k-mooney | i also see you stared adding func test in a follow up to codify some of the manual tests | |
| 19:41:16 | sean-k-mooney | oh and you adressed the compute node create traceback | |
| 19:41:35 | sean-k-mooney | cool let us know when its ready to re review | |
| 19:43:16 | sean-k-mooney | i think you have adressed everything i found in my manual testing at this point i can quickly run true the list again tomorrow | |
| 19:48:29 | dansmith | sean-k-mooney: no, I can't really address the traceback (on startup) without a change to oslo.service AFAIK | |
| 19:48:39 | dansmith | but it will no longer be a trace if it happens during periodic | |
| 19:49:07 | sean-k-mooney | i was refering to https://review.opendev.org/c/openstack/nova/+/872432/1/nova/compute/manager.py | |
| 19:49:15 | sean-k-mooney | sorry not that | |
| 19:49:22 | sean-k-mooney | https://review.opendev.org/c/openstack/nova/+/872432/1/nova/compute/resource_tracker.py | |
| 19:49:39 | dansmith | sean-k-mooney: if you could re-+W this before you go, that can still merge https://review.opendev.org/c/openstack/nova/+/872220/3 | |
| 19:49:44 | dansmith | and then we'll have less for tomorrow | |
| 19:50:01 | sean-k-mooney | sure | |
| 19:50:06 | dansmith | sean-k-mooney: yeah, but in order to get service startup to abort, we still have to raise and you'll get a trace in the logs | |
| 19:50:14 | dansmith | sean-k-mooney: that one was just hit by a rebase accidentally | |
| 19:50:15 | sean-k-mooney | i was just skiming the later two patches by the way | |
| 19:50:58 | sean-k-mooney | dansmith: ok and the agent will abort on start in that case? | |
| 19:51:11 | dansmith | yes, | |
| 19:51:14 | sean-k-mooney | ill test it to tomrrow either way just wondering what to expect | |
| 19:51:16 | sean-k-mooney | cool | |
| 19:51:36 | dansmith | the reason it wasn't before is we swallow and ignore Exception, except for specific ones, so now this makes InvalidConfiguration abort, if startup=True | |
| 19:51:56 | sean-k-mooney | presumably in a decorator | |
| 19:52:03 | dansmith | so that patch is mostly just to make sure we catch the duplicate error specifically, turn it into InvalidConfiguration, and then allow InvalidConfiguration on startup to abort us | |
| 19:52:06 | simondodsley | Question from a customer using Queens (I know EOL and unsupported), is there a way to migrate a boot volume to a new backend wit ha shutdown instance? | |
| 19:52:07 | dansmith | no | |
| 19:52:24 | dansmith | just in our own wrapper around update_available_resource() | |
| 19:52:32 | sean-k-mooney | ah ok | |
| 19:52:33 | dansmith | you'll see when you look closer.. it was fairly obvious, there are just lots of layers | |
| 19:53:01 | sean-k-mooney | ack | |
| 19:53:24 | sean-k-mooney | simondodsley: so ha shutdown isntance you mean isntancce ha is in use but the instance is stopped | |
| 19:53:48 | simondodsley | i beleive that is what they are asking | |
| 19:53:51 | sean-k-mooney | simondodsley: if its a boot form volume guest i would still expec a cinder volume retry or volume migration could be used | |
| 19:54:12 | sean-k-mooney | nova does not have any apis for this so you would have to ask the cinder folks i think | |
| 19:54:37 | simondodsley | they tried the cinder retype and it dodn't work. Nova got confused and attached the new volume with the wrong vd device | |
| 19:54:49 | simondodsley | or is that definatley a cinder/os-brick thing? | |
| 19:55:21 | sean-k-mooney | vd device? | |
| 19:55:35 | sean-k-mooney | as in /dev/vda /dev/vdb in the guest | |
| 19:56:17 | sean-k-mooney | if so the device path in the guest is not actully used when you are using libvirt | |
| 19:56:24 | sean-k-mooney | we cant actully contol that | |
| 19:56:54 | sean-k-mooney | but if you mean on the host that sound like a bug but im not sure if its a nova one of os-brick/cinder one | |
| 19:57:37 | sean-k-mooney | most of the host block device managemnt is done os-brick | |
| 20:01:23 | sean-k-mooney | dansmith: oh ya that was right in front of me https://review.opendev.org/c/openstack/nova/+/872432/1/nova/compute/manager.py#10491 | |
| 20:02:19 | dansmith | yep, that's it, and you can see earlier we raise on startup for reshape things | |