pod 没有将状态从未就绪更改为就绪

Question

我在我的 k8s pod 中使用就绪探测功能。为此，探测检查文件 /tmp/healthy 是否存在。如果不是，探测标记 pod UN-READY。我观察到如果我再次创建文件 /tmp/healthy，pod 不会返回到 READY 状态。这是否意味着在 pod 生命周期中只有 READY--> UN-READY 而不是相反的路径？

/home/ravi/for_others/ric/stub>kubectl describe pod -n myns  deployment-eterm
Name:         deployment-eterm
Namespace:    myns
Priority:     0
Status:       Running
IP:           192.168.252.87
IPs:
  IP:           192.168.252.87
Controlled By:  ReplicaSet/deployment-myns-eterm-micro-6b896556c8
Containers:
    Liveness:       exec [/bin/sh -c /tmp/liveliness-status-collector.sh] delay=60s timeout=1s period=2s #success=1 #failure=5
    Readiness:      exec [/bin/sh -c cat /tmp/healthy] delay=60s timeout=1s period=10s #success=1 #failure=1
    Environment Variables from:
      configmap-myns-eterm-env-micro  ConfigMap  Optional: false
    Environment:                       <none>
    Mounts:
      /var/run/secrets/kubernetes.io/serviceaccount from default-token-57jpz (ro)
Conditions:
  Type              Status
  Initialized       True
  Ready             False
  ContainersReady   False
  PodScheduled      True
Tolerations:     node.kubernetes.io/not-ready:NoExecute for 300s
                 node.kubernetes.io/unreachable:NoExecute for 300s
Events:
  Type     Reason     Age               From                    Message
  ----     ------     ----              ----                    -------
  Normal   Scheduled  9m11s             default-scheduler       Successfully assigned myns/deployment-eterm to mnode
  Normal   Pulled     9m10s             kubelet, mnode  Container image "porics.microlab.com:5000/eterm:latest" already present on machine
  Normal   Created    9m10s             kubelet, mnode  Created container container-myns-eterm
  Normal   Started    9m10s             kubelet, mnode  Started container container-myns-eterm
  Normal   Pulled     9m10s             kubelet, mnode  Container image "porics.microlab.com:5000/config-proxy:latest" already present on machine
  Normal   Created    9m10s             kubelet, mnode  Created container container-myns-configproxy
  Normal   Started    9m10s             kubelet, mnode  Started container container-myns-configproxy
  Warning  Unhealthy  3s (x3 over 23s)  kubelet, mnode  Readiness probe failed: cat: /tmp/healthy: No such file or directory
/home/ravi/for_others/ric/stub>

Answer 1

以下说法不正确：

Does it mean, in pod lifecycle, there is the only one path of READY--> UN-READY but not vice-versa?

您可以运行下面的实验来查看 pod 转换：

使用以下清单文件创建一个 pod：

apiVersion: v1
kind: Pod
metadata:
  labels:
    test: readiness
  name: readiness-test
spec:
  nodeName: kube-master
  containers:
  - name: liveness
    image: bash
    command: ['bash','-c', 'while true;do echo "$(date): creating /tmp/healthy"; touch /tmp/healthy; sleep 10; echo "$(date): deleting /tmp/healthy";rm /tmp/healthy ;sleep 10;done']
    readinessProbe:
      exec:
        command:
        - cat
        - /tmp/healthy
      initialDelaySeconds: 5
      periodSeconds: 2

以上清单将导致 pod 每 10 秒转换为 ready/not-ready。

kubectl get pod -w
NAME             READY   STATUS    RESTARTS   AGE
readiness-test   0/1     Running   0          79s
readiness-test   1/1     Running   0          83s
readiness-test   0/1     Running   0          97s
readiness-test   1/1     Running   0          103s
readiness-test   0/1     Running   0          117s
readiness-test   1/1     Running   0          2m3s
readiness-test   0/1     Running   0          2m17s
readiness-test   1/1     Running   0          2m23s
readiness-test   0/1     Running   0          2m37s
readiness-test   1/1     Running   0          2m43s
readiness-test   0/1     Running   0          2m57s
readiness-test   1/1     Running   0          3m3s
readiness-test   0/1     Running   0          3m17s
readiness-test   1/1     Running   0          3m23s
readiness-test   0/1     Running   0          3m37s
readiness-test   1/1     Running   0          3m43s
readiness-test   0/1     Running   0          3m57s
readiness-test   1/1     Running   0          4m3s

您可以看到准备状态正在发生变化：

while true; do k get pod readiness-test -o jsonpath='{.status.containerStatuses[*].ready}{"\n"}';sleep 2 ;done
true
true
true
true
true
true
false
false
true
true
true
true
true
true
false
false
true
true
true
true
true
true
false
false
true
true
true
true
true
false
false
false

pod 没有将状态从未就绪更改为就绪

pod is not changing state from unready to ready

kubernetes

kubernetes-pod