# 创建故意崩溃的 Pod
cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: Pod
metadata:
  name: crash-demo
spec:
  containers:
  - name: bad-app
    image: busybox:1.36
    command: ["sh", "-c", "echo 'Starting...'; sleep 5; exit 1"]
EOF

# 排查流程：
# Step 1: 确认状态
kubectl get pod crash-demo
# STATUS: CrashLoopBackOff

# Step 2: 查看事件
kubectl describe pod crash-demo | grep -A15 Events
# Back-off restarting failed container

# Step 3: 查看日志（包括上一次）
kubectl logs crash-demo
kubectl logs crash-demo --previous
# Starting...

# Step 4: 结论 → 应用启动后 exit 1，需要检查应用配置
kubectl delete pod crash-demo

cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: Pod
metadata:
  name: oom-demo
spec:
  containers:
  - name: mem-hog
    image: busybox:1.36
    command:
    - sh
    - -c
    - dd if=/dev/zero of=/dev/shm/big bs=50M count=10
    resources:
      limits:
        memory: "100Mi"
EOF

# 排查
kubectl get pod oom-demo -w
# STATUS: OOMKilled

kubectl describe pod oom-demo | grep -A5 "State"
# Reason: OOMKilled

kubectl delete pod oom-demo

kubectl run bad-image --image=notexist/image:v99 --restart=Never

kubectl get pod bad-image
# STATUS: ImagePullBackOff / ErrImagePull

kubectl describe pod bad-image | grep -A5 Events
# Failed to pull image: notexist/image:v99

kubectl delete pod bad-image

# 模拟 Service 不通的场景
# 1. 创建后端但故意 label 不匹配
kubectl create deploy web-backend --image=nginx:alpine --replicas=2

# 2. 创建 Service 选错误的 label
cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: Service
metadata:
  name: bad-svc
spec:
  selector:
    app: non-existent     # 故意不匹配！
  ports:
  - port: 80
EOF

# 3. 排查
kubectl get endpoints bad-svc
# ENDPOINTS: <none>  ← 问题！

kubectl get pod -l app=web-backend --show-labels
# 实际 label: app=web-backend

kubectl get svc bad-svc -o jsonpath='{.spec.selector}'
# {"app":"non-existent"}

# 4. 修复
kubectl patch svc bad-svc -p '{"spec":{"selector":{"app":"web-backend"}}}'
kubectl get endpoints bad-svc
# 现在有 ENDPOINTS 了

# 5. 清理
kubectl delete deploy web-backend
kubectl delete svc bad-svc

# 模拟节点问题
# 1. 对节点加污点使其不可调度
kubectl taint node k8s-node1 test=block:NoSchedule

# 2. 创建 Pod 观察现象
kubectl run taint-test --image=nginx:alpine
kubectl get pod taint-test -o wide
# 被调度到其他节点

# 3. 如果所有节点都有污点？
kubectl describe pod taint-test | grep -A10 Events
# Warning  FailedScheduling  ...

# 4. 排查节点状况
kubectl describe node k8s-node1 | grep -A10 Taints
kubectl describe node k8s-node1 | grep -A5 Conditions
# MemoryPressure / DiskPressure / PIDPressure

# 5. 恢复
kubectl taint node k8s-node1 test-
kubectl delete pod taint-test

# 模拟 PVC 无法绑定
cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: impossible-pvc
spec:
  accessModes: ["ReadWriteMany"]
  resources:
    requests:
      storage: 100Ti    # 不可能有的容量
EOF

# 排查
kubectl get pvc impossible-pvc
# STATUS: Pending

kubectl describe pvc impossible-pvc | grep -A5 Events
# "no persistent volumes available..."

kubectl get pv
# 没有 100Ti 的 PV

kubectl delete pvc impossible-pvc