Skip to main content
Version: v1.5

Workload Scheduling

Nova determines where Kubernetes workloads run across a fleet of workload clusters.

At a high level, scheduling involves:

  1. Identifying the workload to place
  2. Determining which workload clusters are eligible
  3. Selecting the most appropriate workload cluster
  4. Applying the workload to that workload cluster

Nova supports two primary ways to control placement:

  • Annotation-based scheduling – directly select a workload cluster
  • Policy-based scheduling – define placement rules and let Nova choose

Policy-based scheduling is the recommended approach for most workloads because it separates placement intent from the workload manifests themselves.

Scheduling Flow​

  1. A workload is submitted to the Nova control plane
  2. Nova evaluates placement requirements
  3. Eligible workload clusters are identified
  4. Nova makes a placement decision
  5. The workload is applied to the selected workload cluster

Scheduling Methods​

Annotation-based Scheduling​

Annotation-based scheduling is useful when you know exactly which workload cluster should run a workload.

For details, see Annotation-based Scheduling.

Policy-based Scheduling​

Policy-based scheduling uses a SchedulePolicy to select workloads, identify eligible workload clusters, and apply placement behavior.

For details, see Policy-based Scheduling.