Created attachment 1836509 [details] vcpus number in flavor page Description of problem: It is possible to use more vcpus than the limit shows. Version-Release number of selected component (if applicable): core_puddle=RHOS-16.2-RHEL-8-20211013.n.1 How reproducible: always Steps to Reproduce: 1. Add new flavor with 1 vcpu --> Add another flavor with 3 vcpus --> Add a third flavor with 2 vcpus 2. Add a new instance --> Attach the first flavor to the instance 3. Open the instance from step 2 --> Attach the second flavor from step 1 --> Save 4. Open the instance from step 2 --> Attach the third flavor from step 1 --> Save 5. Go to Admin tab --> Compute --> Hypervisors --> Check the vcpus pie chart Actual results: The vcpus pie chart shows 'Used 6 of 4' vcpus(!) Expected results: It will not be possible to add more vcpus than the limit (in this case 4) Additional info:
Horizon is not responsible for enforcing limits. Moving to Nova.
I suspect this is coming from Horizon using the hypervisors API instead of querying the Placement resource providers? In that sense, this is similar to BZ 1846881, where I've said that Nova has made a point of removing usage information from the `hypervisors` API [1] as it was "frequently misleading or outright wrong [and] can be better queried from placement" - and we do not intend to re-add anything back to it. I'm moving this back to Horizon, feel free to close as a duplicate of 1846881. [1] https://docs.openstack.org/nova/latest/reference/api-microversion-history.html#maximum-in-wallaby
*** This bug has been marked as a duplicate of bug 1846881 ***