/api/configs/pool

Pool config is used to configure compute pools for workload distribution and resource management.

Top-Level Configuration

Top-level configuration is used to configure compute pools.

Field	Type	Description	Default Values
pools	Dict[String, Pool]	Dictionary of pool name to compute pool configurations.	{}

Pool

Field	Type	Description	Default Values
name	String	Unique identifier for the pool.	Required field
description	String	Human-readable description of the pool and its purpose.	None
enable_maintenance	Boolean	Whether maintenance mode is enabled for the pool.	False
backend	String	Name of the backend associated with this pool.	None
default_platform	String	Default platform identifier to use for tasks in this pool.	None
default_exec_timeout	String	Default execution timeout for tasks. Must be in the format of <integer><unit> (for example, 10m, 1h, 1d).	Inherited from default_exec_timeout workflow config
default_queue_timeout	String	Default queue timeout for tasks. Must be in the format of <integer><unit> (for example, 10m, 1h, 1d).	Inherited from default_queue_timeout workflow config
max_exec_timeout	String	Maximum allowed execution timeout for tasks. Must be in the format of <integer><unit> (for example, 10m, 1h, 1d).	Inherited from max_exec_timeout workflow config
max_queue_timeout	String	Maximum allowed queue timeout for tasks. Must be in the format of <integer><unit> (for example, 10m, 1h, 1d).	Inherited from max_queue_timeout workflow config
default_exit_actions	Dict[String, Integer]	Default actions to perform when tasks exit with a specific exit code. The available actions are: COMPLETE, FAIL, RESCHEDULE.	{}
resources	Dict[String, Resource Constraint]	Resource allocation configuration for the pool.	{}
common_default_variables	Dict[String, String]	Default values for variables used in pod templates.	{}
common_resource_validations	Array[String]	List of resource validation names applied to all platforms in the pool. Read more about resource validation in Resource Validation.	[]
common_pod_template	Array[String]	List of pod template names applied to all platforms in the pool. Read more about pod templates in Pod Templates.	[]
common_group_templates	Array[String]	List of group template names applied to all task groups in the pool. These Kubernetes resources are created before the group’s pods. Read more about group templates in Group Templates.	[]
platforms	Dict[String, Platform]	Dictionary of platform configurations available in this pool.	{}
topology_keys	Array[TopologyKey]	Ordered list of topology levels available in this pool, from coarsest to finest granularity. Enables topology-aware scheduling when non-empty. Only supported on pools backed by KAI Scheduler v0.12 or later. See Topology-Aware Scheduling for details.	[]

Resource Constraint

For more explanations and examples, see Scheduler Configuration.

Field	Type	Description	Default Values
guarantee	Integer	Guaranteed minimum number of resources allocated to the pool for non-preemptible workflows.	-1
maximum	Integer	Maximum number of resources that can be allocated to the pool.	-1
weight	Integer	Scheduling weight for resource allocation priority relative to other pools.	1

Warning

To set up priority and preemption for pools sharing the same resource nodes, admins need to configure ALL pools with the guarantee, weight, and maximum settings. Otherwise, preemption will not be enabled across pools (e.g. Pool A cannot preempt a workflow from Pool B).

TopologyKey

Field	Type	Description	Default Values
key	String	User-friendly name for this topology level. This is the name users reference in their workflow spec topology requirements (e.g., zone, rack, gpu-clique).	Required field
label	String	The Kubernetes node label used to identify nodes at this topology level (e.g., topology.kubernetes.io/zone, nvidia.com/gpu-clique).	Required field

Platform

Field	Type	Description	Default Values
description	String	Human-readable description of the platform.	None
host_network_allowed	Boolean	Whether tasks can use host networking.	False
privileged_allowed	Boolean	Whether privileged containers are allowed.	False
allowed_mounts	Array[String]	List of mount paths that are allowed for tasks.	[]
default_mounts	Array[String]	Default mount configurations applied to tasks.	[]
default_variables	Dict[String, String]	Default values for variables used in pod templates. If the variable is defined in the pool setting, this will override the pool setting.	{}
resource_validations	Array[String]	Platform-specific resource validation rules. Read more about resource validation in Resource Validation.	[]
override_pod_template	Array[String]	Pod template overrides specific to this platform. Read more about pod templates in Pod Templates.	[]