在Kubernetes中调度器的核心功能是自动分配Pod到合适节点但通过以下高级特性可精细化控制调度逻辑满足复杂场景需求一、节点亲和性与反亲和性Node Affinity/Anti-Affinity通过节点标签Labels和注解Annotations实现定向调度或规避调度支持硬性规则强制和软性规则建议。硬性规则**requiredDuringSchedulingIgnoredDur****ingExecution**强制Pod必须满足条件才能调度否则调度失败。示例禁止Pod调度到特定可用区affinity: nodeAffinity: requiredDuringSchedulingIgnoredDuringExecution: nodeSelectorTerms: - matchExpressions: - key: failure-domain.beta.kubernetes.io/zone operator: NotIn values: [asia-south1-a] # 排除该区域节点软性规则**preferredDuringSchedulingIgnoredDu****ringExecution**优先尝试匹配条件若不满足则降级调度。示例优先调度到指定可用区affinity: nodeAffinity: preferredDuringSchedulingIgnoredDuringExecution: - weight: 100 # 权重越高优先级越高 preference: matchExpressions: - key: failure-domain.beta.kubernetes.io/zone operator: In values: [asia-south1-a] # 优先选择此区域关键操作符In、NotIn、Exists等支持复杂逻辑组合。二、Pod亲和性与反亲和性Pod Affinity/Anti-Affinity基于已有Pod的分布控制新Pod的调度位置常用于高可用或性能优化场景。1、Pod反亲和性避免同节点部署示例确保Redis缓存Pod分散在不同节点affinity: podAntiAffinity: requiredDuringSchedulingIgnoredDuringExecution: - labelSelector: matchExpressions: - key: app operator: In values: [redis] topologyKey: kubernetes.io/hostname # 以节点为调度域2、Pod亲和性同节点部署示例将Web Pod调度到已有Redis Pod的节点affinity: podAffinity: requiredDuringSchedulingIgnoredDuringExecution: - labelSelector: matchExpressions: - key: app operator: In values: [redis] topologyKey: kubernetes.io/hostname # 与Redis同节点拓扑域**topologyKey**kubernetes.io/hostname节点级隔离failure-domain.beta.kubernetes.io/zone可用区级隔离三、组合策略实现复杂调度场景示例Web应用与缓存Pod的协同调度目标每个节点仅运行1个缓存Pod反亲和性。Web Pod需与缓存Pod同节点亲和性。配置方案# 缓存Deployment反亲和性 affinity: podAntiAffinity: requiredDuringSchedulingIgnoredDuringExecution: - labelSelector: matchExpressions: - key: app operator: In values: [redis] topologyKey: kubernetes.io/hostname # 禁止同节点多副本 # Web Deployment亲和性反亲和性 affinity: podAntiAffinity: requiredDuringSchedulingIgnoredDuringExecution: - labelSelector: matchExpressions: - key: app operator: In values: [web] topologyKey: kubernetes.io/hostname # Web自身也分散部署 podAffinity: requiredDuringSchedulingIgnoredDuringExecution: - labelSelector: matchExpressions: - key: app operator: In values: [redis] topologyKey: kubernetes.io/hostname # 必须与Redis同节点四、其他高级调度机制污点与容忍度Taints and Tolerations节点标记污点如NoSchedulePod需声明容忍度才能调度。自定义调度器完全接管调度逻辑如基于Python开发调度器绑定指定节点。调度框架插件扩展默认调度器例如修改打分策略如MostAllocated优化资源利用率。总结特性适用场景规则类型节点亲和性/反亲和性节点级定向调度如区域隔离硬性/软性Pod亲和性/反亲和性Pod协同部署如微服务同节点硬性/软性 拓扑域污点与容忍度节点状态感知如维护模式节点标记 Pod声明通过组合上述策略可精确控制Pod分布平衡资源利用率、容灾与性能需求。实际应用中需注意规则冲突如硬性规则叠加导致无节点可用建议优先使用软性规则保障调度弹性。