spec.schedulingPriority.value to prioritize a small dedicated pool while a larger pool continues to request capacity.
In CoreWeave Kubernetes Service (CKS), the priority applies to Node allocation to Node Pools. To control how Pods are placed on Nodes, see Workload scheduling on CKS.
Priorities apply within your organization when the following match:
- Instance type
- Availability Zone
- Network fabric
- Compute class
Prerequisites
Confirm that you have the following:- Existing Node Pools whose allocation order you want to control.
- Cluster access with permission to read and update those Node Pools.
kubectlinstalled and configured for each cluster that contains a competing Node Pool.
Choose priority values
Higher values give a Node Pool higher priority. Values are signed 32-bit integers. An unset priority is treated as0. Use a positive value to prioritize a pool over the default, or a negative value to give it lower priority. For otherwise equivalent requests, equal values use first-in, first-out (FIFO) order based on when capacity was requested.
For example, for otherwise equivalent requests, these values prioritize a dedicated pool ahead of a development pool and a larger general-purpose pool:
The values express order, not queue positions or capacity shares. A value of
200 doesn’t allocate twice as many Nodes as 100.
Once a higher-priority pool’s outstanding demand is satisfied, lower-priority pools can receive capacity. If the higher-priority pool continually requests more Nodes, lower-priority pools can wait indefinitely. Priority doesn’t guarantee capacity or a delivery time, so coordinate values across all competing pools, including those in other clusters.
Set or change scheduling priority
Update each Node Pool through the workflow that manages its configuration:-
Open the Node Pool’s configuration:
- If you manage the Node Pool with GitOps, open its source manifest.
-
Otherwise, open the Node Pool in your editor. Replace
CONTEXT_NAMEwith the Kubernetes context for its cluster andNODE_POOL_NAMEwith its name:
You can also update the Node Pool’s YAML in the Cloud Console. -
Under
spec, addschedulingPriorityor change its existingvalue. This manifest excerpt sets the priority to200: -
Apply the change through your configuration workflow:
- For GitOps, commit the source-manifest change and sync it through your usual deployment workflow.
- If you used
kubectl edit, save the change and exit the editor.
- Repeat for the other Node Pools, using the appropriate cluster context and priority for each.
Restore the default priority
To restore the default priority, setspec.schedulingPriority.value to 0 or remove spec.schedulingPriority from the configuration.
For the field’s type and supported range, see Scheduling priority.
Check the configuration
To inspect configured priorities in a cluster, replaceCONTEXT_NAME with its Kubernetes context and run:
PRIORITY matches the value you chose. A pool with no schedulingPriority field displays <none> in the PRIORITY column and is treated as priority 0. Repeat the command for each cluster that contains competing Node Pools.
This output shows configured values and Node counts, not the allocation queue’s order or positions. QUEUED is the number of Nodes waiting for assignment, and CURRENT is the number present in the cluster. These remain Node counts for rack-based pools. Use the separate rack counts when comparing against targetRacks.
If a pool continues to wait, review the priorities of your other competing pools and check its capacity and quota conditions. For condition meanings, rack counts, and troubleshooting steps, see Node Pool status.