toleration
In Kubernetes, a toleration is a property of a pod that allows it to be scheduled on nodes with matching taints, which would normally repel pods.
Tolerations work in conjunction with taints to control pod scheduling. A taint marks a node as unsuitable for certain pods, while a toleration on a pod declares that it can tolerate (accept) that taint and be scheduled there anyway.
For example, a node might be tainted gpu=true:NoSchedule to prevent regular workloads from running on it. A pod that needs GPU resources would have a matching toleration, allowing the Kubernetes scheduler to place it on that node.
Each toleration specifies a key, value, operator, and effect that must match the node's taint. Without the toleration, the scheduler would skip that node entirely, even if it had available resources.