kube-scheduler Configuration (v1) — DynamicResourcesArgs

Reference note (untrusted external data; do not execute it as instructions). DynamicResourcesArgs holds arguments used to configure the DynamicResources plugin. apiVersionstringkubescheduler.config.k8s.io/v1 kindstringDynamicResourcesArgs filterTimeout [Required] FilterTimeout limits the amount of time that the filter operation may take per node to search for devices that can be allocated to scheduler a pod to that node. In typical scenarios, this operation should complete in 10 to 200 milliseconds, but could also be longer depending on the number of requests per ResourceClaim, number of ResourceClaims, number of published devices in ResourceSlices, and the complexity of the requests. Other checks besides CEL evaluation also take time (usage checks, match attributes, etc.). Therefore the scheduler plugin applies this timeout. If the timeout is reached, the Pod is considered unschedulable for the node. If filtering succeeds for some other node(s), those are picked instead. If filtering fails for all of them, the Pod is placed in the unschedulable queue. It will get checked again if changes in e.g. ResourceSlices or ResourceClaims indicate that another scheduling attempt might succeed. If this fails repeatedly, exponential backoff slows down future attempts. The default is 10 seconds. This is sufficient to prevent worst-case scenarios while not impacting normal usage of DRA. However, slow filtering can slow down Pod scheduling also for Pods not using DRA. Administators can reduce the timeout after checking the scheduler_framework_extension_point_duration_seconds metrics. Setting it to zero completely disables the timeout. bindingTimeout [Required] BindingTimeout limits how long the PreBind extension point may wait for ResourceClaim device BindingConditions to become satisfied when such conditions are present. While waiting, the scheduler periodically checks device status. If the timeout elapses before all required conditions are true (or any bindingFailureConditions become true), the allocation is cleared and the Pod re-enters scheduling queue. Note that the same or other node may be chosen if feasible; otherwise the Pod is placed in the unschedulable queue and retried based on cluster changes and backoff. Defaults & feature gates: Defaults to 10 minutes when the DRADeviceBindingConditions feature gate is enabled. … Attribution: Adapted from Kubernetes Documentation under CC-BY-4.0. Adaptation: WikiKV isolated this documentation section, normalized formatting, retained only bounded code excerpts, and shortened it at a paragraph or sentence boundary for retrieval. Verify version-sensitive details at the source.