SchedulingOptimization CRD reference

SchedulingOptimization is a cluster-scoped request asking Stork to re-evaluate whether running workloads are still placed where the scheduler would put them today and to converge them back. The storage driver creates one CR per request (for example after a pool failover moves volume replicas). Stork only updates status; the creator owns deletion..

SchedulingOptimization

Field	Description	Type
apiVersion	APIVersion defines the versioned schema of this representation of an object. Servers should convert recognized schemas to the latest internal value, and may reject unrecognized values. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#resources	string
kind	Kind is a string value representing the REST resource this object represents. Servers may infer this from the endpoint the client submits requests to. Cannot be updated. In CamelCase. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds	string
spec	SchedulingOptimizationSpec holds one named field per optimization kind, each with its own config block; future kinds are additive named fields. At least one kind must be enabled for the request to do anything.	object
status	SchedulingOptimizationStatus is the status of a scheduling optimization request	object

spec fields

Field	Description	Type
spec.vmHyperconvergence	VMHyperconvergence requests remediation of KubeVirt VMs that are no longer running on a node holding a replica of their volumes.	object
spec.vmHyperconvergence.coolOffPeriod	CoolOffPeriod overrides the cluster-wide default for the minimum time between Stork-triggered live migrations of the same VM.	string
spec.vmHyperconvergence.enabled	Enabled is the request marker: true means run VM hyperconvergence remediation. The typical request sets only this field.	boolean
spec.vmHyperconvergence.maxParallelMigrations	MaxParallelMigrations overrides the cluster-wide default for the engine's per-run worker configuration. In practice this sets the taskmanager's distributor buffer size — a backpressure hint, not a hard semaphore — plus the workers value the engine reports on its logs and clamp events. The hard cap on concurrently-running live migrations comes from KubeVirt's own MigrationConfiguration.parallelMigrationsPerCluster (default 5), which queues excess VMIMs on its side.	integer

status fields

Field	Description	Type
status.finishTimestamp	FinishTimestamp records when the run reached the final stage	string
status.lastUpdateTimestamp	LastUpdateTimestamp is bumped on every status write — a liveness signal for the creator while counters sit still during a long migration	string
status.reason	Reason carries a short explanation of the terminal Status whenever there is something worth stating. Populated on non-clean outcomes (candidates outstanding, driver failure) and on clean no-op passes where the run completed successfully without doing any work — an opt-out CR (no optimization kind enabled, or the kind's Enabled=false) finalizes as Successful with a reason string; an already-converged fleet finalizes as Successful with an empty reason.	string
status.stage	SchedulingOptimizationStageType is where in the workflow the request is	string
status.startTimestamp	StartTimestamp records when the request was promoted and the run started	string
status.status	SchedulingOptimizationStatusType is how the request is going	string
status.vmHyperconvergence	VMHyperconvergence holds the result of VM hyperconvergence remediation, mirroring the spec field of the same name	object
status.vmHyperconvergence.details	Details lists per-VM outcomes for evaluator-side skips and every candidate the migration engine ran. Bounded by maxNonHyperconvergedPerRun in the evaluator. Successful/Skipped/Failed counters are always exact regardless of Details content.	array
status.vmHyperconvergence.details.name	Name of the VirtualMachineInstance	string
status.vmHyperconvergence.details.namespace	Namespace of the VirtualMachineInstance	string
status.vmHyperconvergence.details.reason	Reason is the skip reason or failure text, set when Status is not Successful. The format is "<Enum>: <detail>" — the stork enum (e.g. MigrationRunFailed, MigrationTimedOut, EvaluationError), a colon separator, and a detail string. For failed migrations the detail begins with "vmim=<ns>/<name>" and may be followed by KubeVirt's own failureReason (e.g. "MigrationRunFailed: vmim=ns/mig-vmi-a: unable to attach volume"). For an EvaluationError skip the detail is the raw evaluator error (e.g. "EvaluationError: persistentvolumeclaims "foo" not found"). Clients matching on the enum should split on the first ": " rather than comparing the full string. The detail is capped to keep CR size bounded.	string
status.vmHyperconvergence.details.status	Status is the per-VM outcome	string
status.vmHyperconvergence.failed	Failed counts migrations attempted but reported failed by KubeVirt	integer
status.vmHyperconvergence.skipped	Skipped counts non-hyperconverged candidates that a filter blocked	integer
status.vmHyperconvergence.successful	Successful counts live migrations that completed successfully	integer