Pod Descheduler问题：处于CrashLoopBackOff状态的Pod无法被驱逐

核心配置问题 #

你的RemoveFailedPods策略存在两个关键错误，导致CrashLoopBackOff状态的Pod无法被驱逐：

1. 无效字段failed_pods_with_reason：Kubernetes 1.25版本的Descheduler（对应release-1.25分支）中，RemoveFailedPods策略的参数里没有该字段，你添加的CrashLoopBackOff不会被识别，需将其移入reasons数组。
2. 原因拼写错误：OutOfcpu应改为OutOfCPU（CPU大写），否则该原因的Pod也无法被匹配。

修正后的ConfigMap配置 #

---
apiVersion: v1
kind: ConfigMap
metadata:
  name: descheduler-policy-configmap
  namespace: kube-system
data:
  policy.yaml: |
    apiVersion: "descheduler/v1alpha1"
    kind: "DeschedulerPolicy"
    evictFailedBarePods: false
    strategies:
      "RemoveDuplicates":
        enabled: true
      "RemovePodsViolatingInterPodAntiAffinity":
        enabled: true
      "LowNodeUtilization":
        enabled: true
        params:
          nodeResourceUtilizationThresholds:
            thresholds:
              "cpu" : 20
              "memory": 20
              "pods": 20
            targetThresholds:
              "cpu" : 50
              "memory": 50
              "pods": 50
      "RemoveFailedPods":
        enabled: true
        params:
          failedPods:
            reasons:
            - OutOfCPU  # 修正拼写错误
            - CreateContainerConfigError
            - Evicted
            - Unhealthy
            - Error
            - Completed
            - BackOff
            - CrashLoopBackOff  # 移至正确的配置数组
            includingInitContainers: true
            # excludeOwnerKinds:
            # - "Job"
            minPodLifetimeSeconds: 120 # 注意：实际为2分钟，注释标注的1小时有误
      "PodLifeTime":
        enabled: true
        params:
          podLifeTime:
            maxPodLifeTimeSeconds: 60 # 注意：实际为1分钟，注释标注的7天有误
            states:
            - "Pending"
            - "PodInitializing"
            - "ContainerCreating"
      "RemovePodsHavingTooManyRestarts":
        enabled: true
        params:
          podsHavingTooManyRestarts:
            podRestartThreshold: 5
            includingInitContainers: true
            runningPods:
              enabled: false
            failedPods:
              enabled: true

后续验证步骤 #

1. 更新ConfigMap：

   kubectl apply -f <你的ConfigMap文件路径>
   

2. 重启Descheduler Pod（或等待下一次CronJob触发，确保新配置加载）：

   kubectl delete pods -n kube-system -l <你的Descheduler Pod标签>
   

3. 查看Descheduler日志，确认配置无报错：

   kubectl logs -n kube-system <descheduler-pod-name>
   

4. 检查CrashLoopBackOff状态的Pod是否被正常驱逐。

另外注意：配置中minPodLifetimeSeconds和maxPodLifeTimeSeconds的实际值与注释描述不符，若需对应注释的时长，请调整数值。

内容的提问来源于stack exchange，提问作者0ptimus