Kubernetes 使用哪种算法将 pods 分配给节点？
Which algorithim does Kubernetes use to assign pods to nodes?
这更多的是成本估算问题，而不是如何使用节点亲和力等功能。
所以基本上 m pods 有一些限制，例如：
- 特定
Deployments / StatefulSets 的每个 pod 应该在不同的 kubernetes 节点上
特定 Deployments / StatefulSets 的 - pods 应在 3 个可用性区域上平衡
现在，我想知道我需要多少个节点（所有相同类型）来托管给定的一组 Deployments / StatefulSets。
我首先认为这更像是要使用匈牙利算法解决的分配问题，但这在多维约束等术语中似乎要复杂得多。
Kubernetes 根据
等许多约束分配 pods
- 资源需求
- 资源存在（节点容量）
- 节点选择器（如果有）或关联规则
- 相似性规则的权重
这篇文章同样适用于：https://kubernetes.io/docs/concepts/scheduling-eviction/kube-scheduler/
另外：https://kubernetes.io/docs/concepts/workloads/pods/pod-topology-spread-constraints/
我建议阅读：https://kubernetes.io/docs/concepts/scheduling-eviction/
据我所知，kube-scheduler 默认使用的算法在 github here.
中有描述
它解释了它是如何工作的。它首先过滤不符合pods要求的节点，例如资源请求 > 节点上的可用资源、亲和力等
然后使用排序算法来确定最合适的节点。如需更深入地了解
参考网友Harsh Manvar, I will add a few more information from myself. This topic was covered in the documentation, as described by my predecessor. Besides her, you can find very good materials here很好的回答：
What Happens When You Create a Pod On a Kubernetes Cluster?
In a matter of seconds, the Pod is up and running on one of the cluster nodes. However, a lot has happened within those seconds. Let’s see:
- While scanning the API server (which it is continuously doing), the Kubernetes Scheduler detects that there is a new Pod without a nodeName parameter. The nodeName is what shows which node should be owning this Pod.
- The Scheduler selects a suitable node for this Pod and updates the Pod definition with the node name (though the nodeName parameter).
- The kubelet on the chosen node is notified that there is a pod that is pending execution.
- The kubelet executes the Pod, and the latter starts running on the node.
您还可以找到有关调度进程和调度程序算法的 tutorial。
这更多的是成本估算问题，而不是如何使用节点亲和力等功能。
所以基本上 m pods 有一些限制，例如：
- 特定
Deployments/StatefulSets的每个 pod 应该在不同的 kubernetes 节点上
特定 - pods 应在 3 个可用性区域上平衡
Deployments / StatefulSets 的 现在，我想知道我需要多少个节点（所有相同类型）来托管给定的一组 Deployments / StatefulSets。
我首先认为这更像是要使用匈牙利算法解决的分配问题，但这在多维约束等术语中似乎要复杂得多。
Kubernetes 根据
等许多约束分配 pods- 资源需求
- 资源存在（节点容量）
- 节点选择器（如果有）或关联规则
- 相似性规则的权重
这篇文章同样适用于：https://kubernetes.io/docs/concepts/scheduling-eviction/kube-scheduler/
另外：https://kubernetes.io/docs/concepts/workloads/pods/pod-topology-spread-constraints/
我建议阅读：https://kubernetes.io/docs/concepts/scheduling-eviction/
据我所知，kube-scheduler 默认使用的算法在 github here.
中有描述它解释了它是如何工作的。它首先过滤不符合pods要求的节点，例如资源请求 > 节点上的可用资源、亲和力等
然后使用排序算法来确定最合适的节点。如需更深入地了解
参考网友Harsh Manvar, I will add a few more information from myself. This topic was covered in the documentation, as described by my predecessor. Besides her, you can find very good materials here很好的回答：
What Happens When You Create a Pod On a Kubernetes Cluster? In a matter of seconds, the Pod is up and running on one of the cluster nodes. However, a lot has happened within those seconds. Let’s see:
- While scanning the API server (which it is continuously doing), the Kubernetes Scheduler detects that there is a new Pod without a nodeName parameter. The nodeName is what shows which node should be owning this Pod.
- The Scheduler selects a suitable node for this Pod and updates the Pod definition with the node name (though the nodeName parameter).
- The kubelet on the chosen node is notified that there is a pod that is pending execution.
- The kubelet executes the Pod, and the latter starts running on the node.
您还可以找到有关调度进程和调度程序算法的 tutorial。