Auto-discover additional fields for cortex filtering #40
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This was referenced Dec 29, 2025
PhilippMatthes
force-pushed
the
cortex-filtering
branch
5 times, most recently
from
ce406f3 to
e344b1c
Compare
This comment was marked as resolved.
This comment was marked as resolved.
PhilippMatthes
added a commit
to cobaltcore-dev/cortex
that referenced
this pull request
Jan 5, 2026
## Background For virtual machines spawned on the kvm hypervisor, we want to no longer use nova and placement as source of truth. Instead, filters should use the hypervisor crd exposed by the [hypervisor operator](github.com/cobaltcore-dev/openstack-hypervisor-operator) and populated by the [node agent](https://github.com/cobaltcore-dev/kvm-node-agent). This contribution replaces the implementation of all filters that were originally ported from nova accordingly. Afterward, we can disable filters in nova one-by-one, moving the compute placement logic over to cortex. > [!TIP] > You can use the newly added [mirror tool](93fdcc0) to mirror hypervisor resources from our compute cluster over to the local cluster. ## Completion - [x] ~internal/scheduling/decisions/nova/plugins/filters/filter_compute_capabilities.go~ (REMOVED) - [x] internal/scheduling/decisions/nova/plugins/filters/filter_capabilities.go (NEW) - [x] internal/scheduling/decisions/nova/plugins/filters/filter_correct_az.go - [x] internal/scheduling/decisions/nova/plugins/filters/filter_external_customer.go - [x] internal/scheduling/decisions/nova/plugins/filters/filter_has_accelerators.go - [x] internal/scheduling/decisions/nova/plugins/filters/filter_has_enough_capacity.go - [x] internal/scheduling/decisions/nova/plugins/filters/filter_has_requested_traits.go - [x] internal/scheduling/decisions/nova/plugins/filters/filter_host_instructions.go - [x] internal/scheduling/decisions/nova/plugins/filters/filter_maintenance.go (NEW) - [x] internal/scheduling/decisions/nova/plugins/filters/filter_packed_virtqueue.go - [x] ~internal/scheduling/decisions/nova/plugins/filters/filter_project_aggregates.go~ (REMOVED) - [x] internal/scheduling/decisions/nova/plugins/filters/filter_allowed_projects.go (NEW) - [x] ~internal/scheduling/decisions/nova/plugins/filters/filter_disabled.go~ (REMOVED) - [x] internal/scheduling/decisions/nova/plugins/filters/filter_status_conditions.go (NEW) ## Dependencies > [!NOTE] > The scope of this PR is to establish a minimum viable scheduling pipeline with the current state. Extensive refactorings, for example of the filter for requested traits, are out of scope. Hypervisor operator PR: cobaltcore-dev/openstack-hypervisor-operator#217 KVM node agent PR: cobaltcore-dev/kvm-node-agent#40
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as resolved.
This comment was marked as resolved.
notandy
requested changes
Jan 5, 2026
Member
Author
|
Refactored and tested: apiVersion: kvm.cloud.sap/v1
kind: Hypervisor
status:
allocation:
cpu: "3"
memory: 4064Mi
capabilities:
cpuArch: x86_64
# ...
capacity:
cpu: "128"
memory: 1056566456Ki
cells:
- allocation:
cpu: "3"
memory: 4064Mi
capacity:
cpu: "64"
memory: 528110060Ki
cellID: 0
- allocation:
cpu: "0"
memory: "0"
capacity:
cpu: "64"
memory: 528456396Ki
cellID: 1
# ...
domainCapabilities:
arch: x86_64
hypervisorType: ch
supportedCpuModes:
- mode/host-passthrough
supportedDevices:
- video
- video/none
supportedFeatures: []
libVirtVersion: 48.0.0 # logic was simplified
numInstances: 2 # logic was simplified
# ... |
notandy
approved these changes
Jan 6, 2026
PhilippMatthes
added a commit
to cobaltcore-dev/openstack-hypervisor-operator
that referenced
this pull request
Jan 7, 2026
## Background In [this pull request](cobaltcore-dev/cortex#441) we implemented a cortex filtering pipeline for KVM. This pipeline uses the hypervisor CRD as single source of truth to find out on which hypervisors a vm can be scheduled. To complete this implementation, we need to extend the hypervisor CRD. ## Tasks Support filtering based on hypervisor type and other capabilities: - [x] The capabilities struct should be extended to support the hypervisor type. Later, we will probably need to extend this struct further. - [x] Add fields for supported devices (e.g. video device), cpu modes, and features (for migration filtering) Capacity filtering: - [x] We need a spec + status not only for the size of the host, but also for the currently used capacity. This will be used by cortex to filter out hosts without the required capacity. This scheduling logic can be made more intelligent in the future, by mapping out individual numa cells and including reserved capacity. Pinned projects: - [x] Provide a spec to declare pinned projects on this hypervisor. (Bonus) - [x] Add numa cell capacity & allocation information so we can implement numa sensitive initial placement When finished: - [x] Ensure backwards compatibility so we can roll this out without any issues ## Dependencies > [!NOTE] > The scope of this PR is to establish a minimum viable scheduling pipeline in cortex, with the least amount of changes possible. Refactorings of the hypervisor crd spec can follow if needed. KVM node agent PR: cobaltcore-dev/kvm-node-agent#40
PhilippMatthes
force-pushed
the
cortex-filtering
branch
from
350a605 to
6bda51b
Compare
Member
Author
|
^ Rebased on current main |
Merging this branch changes the coverage (2 decrease, 3 increase)
Coverage by fileChanged files (no unit tests)
Please note that the "Total", "Covered", and "Missed" counts above refer to code statements instead of lines of code. The value in brackets refers to the test coverage of that file in the old version of the code. Changed unit test files
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In this pull request we implemented a cortex filtering pipeline for KVM. This pipeline uses the hypervisor CRD as single source of truth to find out on which hypervisors a vm can be scheduled. To complete this implementation, we extended the hypervisor crd in this pull request. The hypervisor crd pull request added additional fields and removed outdated ones, which need to be autodiscovered in the kvm node agent. The following fields are now populated:
Support filtering based on hypervisor type and other capabilities:
Capacity filtering:
(Bonus)
When done:
Note
The scope of this PR is to establish a minimum viable scheduling pipeline in cortex, with the least amount of changes possible. Refactorings of the hypervisor crd spec can follow if needed.