def _get_agent_metadata(self, node_ip: str) -> AgentMetadata:
node = self._nodes_by_ip.get(node_ip)
if not node:
return AgentMetadata(state=AgentState.ORPHANED)
return AgentMetadata(
agent_id=node.metadata.name,
allocated_resources=allocated_node_resources(
self._pods_by_ip[node_ip]),
batch_task_count=self._count_batch_tasks(node_ip),
state=(AgentState.RUNNING
if self._pods_by_ip[node_ip] else AgentState.IDLE),
task_count=len(self._pods_by_ip[node_ip]),
total_resources=total_node_resources(node),
)
def _get_pod_unschedulable_reason(
self, pod: KubernetesPod) -> PodUnschedulableReason:
pod_resource_request = total_pod_resources(pod)
for node_ip, pods_on_node in self._pods_by_ip.items():
node = self._nodes_by_ip.get(node_ip)
if node:
available_node_resources = (
total_node_resources(
node, self._excluded_pods_by_ip.get(node_ip, [])) -
allocated_node_resources(pods_on_node))
if pod_resource_request < available_node_resources:
return PodUnschedulableReason.Unknown
return PodUnschedulableReason.InsufficientResources
def get_resource_allocation(self, resource_name: str) -> float:
return sum(
getattr(allocated_node_resources(self._pods_by_ip[node_ip]),
resource_name)
for node_ip, node in self._nodes_by_ip.items())
def get_resource_allocation(self, resource_name: str) -> float:
return sum(
getattr(allocated_node_resources(pod), resource_name)
for pod in self._pods_by_ip.values())
def get_resource_pending(self, resource_name: str) -> float:
return getattr(
allocated_node_resources(
[p for p, __ in self.get_unschedulable_pods()]), resource_name)
def get_resource_excluded(self, resource_name: str) -> float:
return sum(
getattr(
allocated_node_resources(
self._excluded_pods_by_ip.get(node_ip, [])), resource_name)
for node_ip in self._nodes_by_ip.keys())
