Manage node scheduling policies - ScitiX

Node scheduling policies limit where workloads can run. You can block specified nodes or require matching workloads to run only on specified nodes. Policies can apply to training tasks, LLM inference, service deployments, VSCS, JupyterLab, or general services.

Create a scheduling policy

Go to Config Management > Schedule Policy.
Select a Region and Cluster, then click Add Schedule Strategy.

Configure the policy.

Configuration	Description
Scope	Only Self: The policy applies to matching workloads submitted by the current user. Entire Tenant: The policy applies to matching workloads submitted within the tenant.
Schedule Type	Block: Matching workloads cannot be scheduled to the specified nodes. Schedule Only: Matching workloads can be scheduled only to the specified nodes.
Effect Mode	Enabled by Default: The policy unconditionally applies to all matching workloads. Select as Needed: The policy takes effect only when it is explicitly selected during workload creation.
Workload Type	Select the workload types affected by the policy. Options include Training, LLM Inference, Service Deployment, VSCS, JupyterLab, and General Service.
Node Name	Enter the names of the nodes to which the policy applies. Separate multiple nodes with commas.
Expire At	Set the policy validity period. To keep the policy in effect indefinitely, select Permanent.
Reason	Record the reason for creating the policy.

Click OK.
Return to the policy list and confirm that the Schedule Type, User Scope, Effect Mode, Expire At, and Effective Scope are as expected.

Manage scheduling policies

Go to Config Management > Schedule Policy, and select a Region and Cluster. You can filter by Scope, Node Name, Created By, and Reason. To view only currently effective policies, enable Show Effective Only. You can edit or delete scheduling policies:

Edit: Adjust the policy configuration. After making changes, return to the list to verify the policy information.
Delete: Delete policies that are no longer needed.