polaris.run.pool.ResourcePool

class polaris.run.pool.ResourcePool(nodes: List[NodeResources], local_node: str | None = None, names_gpus: bool = True)

The allocation’s free resources, and what may be started against them.

Variables:

local_node (str) – The node the scheduler itself is running on, which is the only one it can confine an unlaunched step to.

__init__(nodes: List[NodeResources], local_node: str | None = None, names_gpus: bool = True)

Parameters:

• nodes (list of polaris.run.allocation.NodeResources) – What the allocation holds, as its nodes reported it

• local_node (str, optional) – The node the scheduler is on. Defaults to the first, which is where a batch script runs.

• names_gpus (bool, optional) – Whether a placement names the GPUs it gives. Not where the batch system reserves GPUs by count, since it picks which itself

Methods

__init__(nodes[, local_node, names_gpus])
release(reservation)	Hand back everything a step was given.
reserve(step)	Give a step what it needs, if what it needs is free right now.
resident(node_name)	The steps on one node right now.
why_impossible(step)	Say why a step could never run here, or None if it could.