Pod Scheduling 失败
错误信息
0/3 节点是 available: 3 节点(s) 有 taints 那个 pod didn't tolerate
快速摘要
scheduler cannot 位置 Pod on 任何节点 due 到 taints, resource 限制, affinity 规则, or 节点 selectors.
为什么发生
节点 taints prevent Pod 从是 scheduled on available 节点
节点 selector or 节点 affinity 规则做 不 match 任何节点
Pod anti-affinity 规则 prevent co-位置使用 existing Pods
Insufficient resources on 所有 matching 节点
最小示例
spec:
nodeSelector:
disktype: ssd # no nodes have this labelspec:
nodeSelector:
disktype: ssd
tolerations:
- key: dedicated
operator: Equal
value: gpu
effect: NoSchedulePod cannot 是 scheduled 因为否 节点 match selector or tolerate taints. fix 添加 tolerations.
如何诊断
描述 Pod:
kubectl 描述 pod <pod>— 检查事件小节检查节点 taints:
kubectl 获取节点 -o 自定义-列=NAME:.metadata.name,TAINTS:.spec.taints验证 nodeSelector, affinity, and tolerations in Pod spec
检查 available resources on 节点:
kubectl 描述节点
如何修复
添加 tolerations 到 Pod spec for 节点 taints
Adjust nodeSelector or affinity 规则到 match available 节点
添加更多节点 or 移除 taints 从 existing 节点
Reduce resource 请求所以 Pod fits on available 节点
如何预防
Document 节点 taints and 必需 tolerations
测试 Pod specs 针对 cluster constraints 之前 deploying
Monitor scheduling failures and alert on 它们