创建第一个 Volcano Job

apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: my-first-job
spec:
  minAvailable: 1
  schedulerName: volcano
  queue: default
  tasks:
    - replicas: 1
      name: worker
      template:
        spec:
          containers:
            - name: busybox
              image: busybox:latest
              command: ["echo", "Hello from Volcano!"]
          restartPolicy: OnFailure

apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: distributed-job
spec:
  # --- 调度配置 ---
  schedulerName: volcano            # 使用 Volcano 调度器
  queue: default                    # 提交到 default 队列
  minAvailable: 3                   # Gang Scheduling: 至少 3 个 Pod 才开始
  maxRetry: 5                       # 最大重试 5 次
  priorityClassName: high-priority  # 优先级类

  # --- 生命周期策略（Job 级，适用于所有 Task）---
  policies:
    - event: PodEvicted
      action: RestartJob            # Pod 被驱逐时重启整个 Job

  # --- 插件配置 ---
  plugins:
    ssh: []                         # 启用 SSH 免密（生成密钥对）
    env: []                         # 注入环境变量（VK_TASK_INDEX 等）
    svc: []                         # 创建 Headless Service（Pod 间网络发现）

  # --- Task 定义 ---
  tasks:
    # Task 1: Master 节点
    - replicas: 1
      name: master
      policies:
        - event: TaskCompleted
          action: CompleteJob       # Master 完成时标记 Job 完成
      template:
        spec:
          containers:
            - name: master
              image: nginx:latest
              resources:
                requests:
                  cpu: "1"
                  memory: "1Gi"
          restartPolicy: OnFailure

    # Task 2: Worker 节点
    - replicas: 6
      name: worker
      template:
        spec:
          containers:
            - name: worker
              image: nginx:latest
              resources:
                requests:
                  cpu: "500m"
                  memory: "512Mi"
          restartPolicy: OnFailure

tasks:
  - name: data-loader
    replicas: 1
    template: ...

  - name: trainer
    replicas: 4
    dependsOn:
      name: ["data-loader"]       # 等待 data-loader 完成后启动
      iteration: "all"            # "all" = 所有依赖完成; "any" = 任一完成
    template: ...

policies:
  # Pod 被驱逐时重启整个 Job
  - event: PodEvicted
    action: RestartJob

  # 使用超时：Job 未知状态超过 5 分钟后 Abort
  - event: JobUnknown
    action: AbortJob
    timeout: 5m

  # 基于退出码的策略：退出码 137（OOMKilled）时重启 Job
  - exitCode: 137
    action: RestartJob

plugins:
  env: []

plugins:
  ssh: []

# 自定义 SSH 密钥路径
plugins:
  ssh: ["--ssh-key-file-path=/home/user/.ssh"]

plugins:
  svc: []

# 自定义选项
plugins:
  svc: ["--publish-not-ready-addresses=true", "--disable-network-policy=false"]

apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: pi-calculation
spec:
  minAvailable: 3
  schedulerName: volcano
  queue: default
  tasks:
    - replicas: 3
      name: compute
      template:
        spec:
          containers:
            - name: pi
              image: perl:5.34
              command: ["perl", "-Mbignum=bpi", "-wle", "print bpi(2000)"]
              resources:
                requests:
                  cpu: "1"
          restartPolicy: Never

apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: mpi-job
spec:
  minAvailable: 3
  schedulerName: volcano
  plugins:
    ssh: []
    svc: []
  tasks:
    # MPI Master
    - replicas: 1
      name: mpimaster
      policies:
        - event: TaskCompleted
          action: CompleteJob
      template:
        spec:
          containers:
            - name: mpimaster
              image: volcanosh/example-mpi:0.0.3
              command:
                - /bin/sh
                - -c
                - |
                  MPI_HOST=`cat /etc/volcano/mpiworker.host | tr "\n" ","`;
                  mkdir -p /var/run/sshd; /usr/sbin/sshd;
                  mpiexec --allow-run-as-root --host ${MPI_HOST} -np 2 mpi_hello_world;
              ports:
                - containerPort: 22
                  name: mpijob-port
          restartPolicy: OnFailure

    # MPI Worker
    - replicas: 2
      name: mpiworker
      template:
        spec:
          containers:
            - name: mpiworker
              image: volcanosh/example-mpi:0.0.3
              command:
                - /bin/sh
                - -c
                - |
                  mkdir -p /var/run/sshd; /usr/sbin/sshd -D;
              ports:
                - containerPort: 22
                  name: mpijob-port
          restartPolicy: OnFailure

apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: pipeline-job
spec:
  schedulerName: volcano
  queue: default
  minAvailable: 1
  tasks:
    # 阶段 1：数据准备
    - replicas: 1
      name: data-prep
      template:
        spec:
          containers:
            - name: prep
              image: busybox:latest
              command: ["sh", "-c", "echo 'Preparing data...' && sleep 10"]
          restartPolicy: OnFailure

    # 阶段 2：训练（依赖阶段 1）
    - replicas: 5
      name: trainer
      dependsOn:
        name: ["data-prep"]
      template:
        spec:
          containers:
            - name: train
              image: busybox:latest
              command: ["sh", "-c", "echo 'Training...' && sleep 30"]
          restartPolicy: OnFailure

    # 阶段 3：评估（依赖阶段 2）
    - replicas: 1
      name: evaluator
      dependsOn:
        name: ["trainer"]
      policies:
        - event: TaskCompleted
          action: CompleteJob
      template:
        spec:
          containers:
            - name: eval
              image: busybox:latest
              command: ["sh", "-c", "echo 'Evaluating...' && sleep 5"]
          restartPolicy: OnFailure