hadoop 部署在k8s

1. 编写好hadoop.yaml

apiVersion: v1
kind: ConfigMap
metadata:
  name: kube-hadoop-conf
  namespace: default
data:
  HDFS_MASTER_SERVICE: hadoop-hdfs-master
  HDOOP_YARN_MASTER: hadoop-yarn-master
---
apiVersion: v1
kind: Service
metadata:
  name: hadoop-hdfs-master
spec:
  type: NodePort
  selector:
    app: hdfs-master
  ports:
    - name: rpc
      port: 9000
      targetPort: 9000
    - name: http
      port: 50070
      targetPort: 50070
      nodePort: 32007
---
apiVersion: v1
kind: Pod
metadata:
  name: hdfs-master
  labels:
    app: hdfs-master
spec:
  containers:
    - name: hdfs-master
      image: k8s.gcr.io/ccmls/hadoop:latest
      imagePullPolicy: IfNotPresent
      ports:
        - containerPort: 9000
        - containerPort: 50070
      env:
        - name: HADOOP_NODE_TYPE
          value: namenode
        - name: HDFS_MASTER_SERVICE
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDFS_MASTER_SERVICE
        - name: HDOOP_YARN_MASTER
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDOOP_YARN_MASTER
  restartPolicy: Always
---
apiVersion: v1
kind: Pod
metadata:
    name: hadoop-datanode-1
    labels:
      app: hadoop-datanode-1
spec:
  containers:
    - name: hadoop-datanode-1
      image: k8s.gcr.io/ccmls/hadoop:latest
      imagePullPolicy: IfNotPresent
      ports:
        - containerPort: 9000
        - containerPort: 50070
      env:
        - name: HADOOP_NODE_TYPE
          value: datanode
        - name: HDFS_MASTER_SERVICE
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDFS_MASTER_SERVICE
        - name: HDOOP_YARN_MASTER
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDOOP_YARN_MASTER
  restartPolicy: Always
---
apiVersion: v1
kind: Pod
metadata:
    name: hadoop-datanode-2
    labels:
      app: hadoop-datanode-2
spec:
  containers:
    - name: hadoop-datanode-2
      image: k8s.gcr.io/ccmls/hadoop:latest
      imagePullPolicy: IfNotPresent
      ports:
        - containerPort: 9000
        - containerPort: 50070
      env:
        - name: HADOOP_NODE_TYPE
          value: datanode
        - name: HDFS_MASTER_SERVICE
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDFS_MASTER_SERVICE
        - name: HDOOP_YARN_MASTER
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDOOP_YARN_MASTER
  restartPolicy: Always
---
apiVersion: v1
kind: Pod
metadata:
    name: hadoop-datanode-3
    labels:
      app: hadoop-datanode-3
spec:
  containers:
    - name: hadoop-datanode-3
      image: k8s.gcr.io/ccmls/hadoop:latest
      imagePullPolicy: IfNotPresent
      ports:
        - containerPort: 9000
        - containerPort: 50070
      env:
        - name: HADOOP_NODE_TYPE
          value: datanode
        - name: HDFS_MASTER_SERVICE
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDFS_MASTER_SERVICE
        - name: HDOOP_YARN_MASTER
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDOOP_YARN_MASTER
  restartPolicy: Always
---
apiVersion: v1
kind: Service
metadata:
  name: hadoop-yarn-master
spec:
  type: NodePort
  selector:
    app: yarn-master
  ports:
     - name: "8030"
       port: 8030
     - name: "8031"
       port: 8031
     - name: "8032"
       port: 8032
     - name: http
       port: 8088
       targetPort: 8088
       nodePort: 32088
---
apiVersion: v1
kind: Pod
metadata:
  name: yarn-master
  labels:
    app: yarn-master
spec:
  containers:
    - name: yarn-master
      image: k8s.gcr.io/ccmls/hadoop:latest
      imagePullPolicy: IfNotPresent
      ports:
        - containerPort: 9000
        - containerPort: 50070
      env:
        - name: HADOOP_NODE_TYPE
          value: resourceman
        - name: HDFS_MASTER_SERVICE
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDFS_MASTER_SERVICE
        - name: HDOOP_YARN_MASTER
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDOOP_YARN_MASTER
  restartPolicy: Always
---
apiVersion: v1
kind: Service
metadata:
  name: yarn-node-1
spec:
  clusterIP: None
  selector:
    app: yarn-node-1
  ports:
     - port: 8040
---
apiVersion: v1
kind: Service
metadata:
  name: yarn-node-2
spec:
  clusterIP: None
  selector:
    app: yarn-node-2
  ports:
     - port: 8040
---
apiVersion: v1
kind: Service
metadata:
  name: yarn-node-3
spec:
  clusterIP: None
  selector:
    app: yarn-node-3
  ports:
     - port: 8040
---
apiVersion: v1
kind: Pod
metadata:
  name: yarn-node-1
  labels:
    app: yarn-node-1
spec:
  containers:
    - name: yarn-node-1
      image: k8s.gcr.io/ccmls/hadoop:latest
      imagePullPolicy: IfNotPresent
      ports:
        - containerPort: 8040
        - containerPort: 8041
        - containerPort: 8042
      env:
        - name: HADOOP_NODE_TYPE
          value: yarnnode
        - name: HDFS_MASTER_SERVICE
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDFS_MASTER_SERVICE
        - name: HDOOP_YARN_MASTER
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDOOP_YARN_MASTER
  restartPolicy: Always
---
apiVersion: v1
kind: Pod
metadata:
  name: yarn-node-2
  labels:
    app: yarn-node-2
spec:
  containers:
    - name: yarn-node-2
      image: k8s.gcr.io/ccmls/hadoop:latest
      imagePullPolicy: IfNotPresent
      ports:
        - containerPort: 8040
        - containerPort: 8041
        - containerPort: 8042
      env:
        - name: HADOOP_NODE_TYPE
          value: yarnnode
        - name: HDFS_MASTER_SERVICE
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDFS_MASTER_SERVICE
        - name: HDOOP_YARN_MASTER
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDOOP_YARN_MASTER
  restartPolicy: Always
---
apiVersion: v1
kind: Pod
metadata:
  name: yarn-node-3
  labels:
    app: yarn-node-3
spec:
  containers:
    - name: yarn-node-3
      image: k8s.gcr.io/ccmls/hadoop:latest
      imagePullPolicy: IfNotPresent
      ports:
        - containerPort: 8040
        - containerPort: 8041
        - containerPort: 8042
      env:
        - name: HADOOP_NODE_TYPE
          value: yarnnode
        - name: HDFS_MASTER_SERVICE
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDFS_MASTER_SERVICE
        - name: HDOOP_YARN_MASTER
          valueFrom:
            configMapKeyRef:
              name: kube-hadoop-conf
              key: HDOOP_YARN_MASTER
  restartPolicy: Always

这个yaml文件包含一个ConfigMap，5个Service，8个pod，这里需要注意的是ConfigMap中HDFS_MASTER_SERVICE和HDOOP_YARN_MASTER不要使用IP，使用HDFS service的名称，否则datanode将会连接不上namenode，出现错误【ipc.Client: Retrying connect to server: xxx:9000.】

1. 执行创建命令

kubectl create -f hadoop.yaml

1. 检查是否创建成功

• 查看config map

kubectl get configmap -o wide 

• 查看service

kubectl get svc -o wide

• 查看pod

kubectl get po -o wide

• 通过浏览器访问HDFS管理界面【http://kubernet-master-ip:32007】

全部正常，搭建成功！
但是上面的搭建方式是以单个POD声明的，这种方式不稳定，如果不小心删除后就没有了，其实我们可以使用Replication Controller方式进行搭建，这样的话始终可以确保保留相应数量的POD，具体的yaml文件如下：
apiVersion: v1
kind: ConfigMap
metadata:
  name: kube-hadoop-conf
  namespace: default
data:
  HDFS_MASTER_SERVICE: hadoop-hdfs-master
  HDOOP_YARN_MASTER: hadoop-yarn-master
---
apiVersion: v1
kind: Service
metadata:
  name: hadoop-hdfs-master
spec:
  type: NodePort
  selector:
    name: hdfs-master
  ports:
    - name: rpc
      port: 9000
      targetPort: 9000
    - name: http
      port: 50070
      targetPort: 50070
      nodePort: 32007
---
apiVersion: v1
kind: ReplicationController
metadata:
  name: hdfs-master
  labels:
    name: hdfs-master
spec:
  replicas: 1
  selector:
    name: hdfs-master
  template:
    metadata:
      labels:
        name: hdfs-master
    spec:
      containers:
        - name: hdfs-master
          image: 10.3.13.184/library/kubernetes-hadoop:latest
          imagePullPolicy: IfNotPresent
          ports:
            - containerPort: 9000
            - containerPort: 50070    
          env:
            - name: HADOOP_NODE_TYPE
              value: namenode
            - name: HDFS_MASTER_SERVICE
              valueFrom:
                configMapKeyRef:
                  name: kube-hadoop-conf
                  key: HDFS_MASTER_SERVICE
            - name: HDOOP_YARN_MASTER
              valueFrom:
                configMapKeyRef:
                  name: kube-hadoop-conf
                  key: HDOOP_YARN_MASTER
      restartPolicy: Always
---
apiVersion: v1
kind: ReplicationController
metadata:
  name: hadoop-datanode
  labels:
    app: hadoop-datanode
spec:
  replicas: 3
  selector:
    name: hadoop-datanode
  template:
    metadata:
      labels:
        name: hadoop-datanode
    spec:
      containers:
        - name: hadoop-datanode
          image: 10.3.13.184/library/kubernetes-hadoop:latest
          imagePullPolicy: IfNotPresent
          ports:
            - containerPort: 9000
            - containerPort: 50070    
          env:
            - name: HADOOP_NODE_TYPE
              value: datanode
            - name: HDFS_MASTER_SERVICE
              valueFrom:
                configMapKeyRef:
                  name: kube-hadoop-conf
                  key: HDFS_MASTER_SERVICE
            - name: HDOOP_YARN_MASTER
              valueFrom:
                configMapKeyRef:
                  name: kube-hadoop-conf
                  key: HDOOP_YARN_MASTER
      restartPolicy: Always
---
apiVersion: v1
kind: Service
metadata:
  name: hadoop-yarn-master
spec:
  type: NodePort
  selector:
    name: yarn-master
  ports:
     - name: "8030"       
       port: 8030
     - name: "8031"     
       port: 8031
     - name: "8032"
       port: 8032     
     - name: http
       port: 8088
       targetPort: 8088
       nodePort: 32088
---
apiVersion: v1
kind: ReplicationController
metadata:
  name: yarn-master
  labels:
    name: yarn-master
spec:
  replicas: 1
  selector:
    name: yarn-master
  template:
    metadata:
      labels:
        name: yarn-master
    spec:
      containers:
        - name: yarn-master
          image: 10.3.13.184/library/kubernetes-hadoop:latest
          imagePullPolicy: IfNotPresent
          ports:
            - containerPort: 9000
            - containerPort: 50070    
          env:
            - name: HADOOP_NODE_TYPE
              value: resourceman
            - name: HDFS_MASTER_SERVICE
              valueFrom:
                configMapKeyRef:
                  name: kube-hadoop-conf
                  key: HDFS_MASTER_SERVICE
            - name: HDOOP_YARN_MASTER
              valueFrom:
                configMapKeyRef:
                  name: kube-hadoop-conf
                  key: HDOOP_YARN_MASTER          
      restartPolicy: Always
---
apiVersion: v1
kind: Service
metadata:
  name: yarn-node
spec:
  clusterIP: None
  selector:
    name: yarn-node
  ports:
     - port: 8040
---
apiVersion: v1
kind: ReplicationController
metadata:
  name: yarn-node
  labels:
    name: yarn-node
spec:
  replicas: 3
  selector:
    name: yarn-node
  template:
    metadata:
      labels:
        name: yarn-node
    spec:
      containers:
        - name: yarn-node
          image: 10.3.13.184/library/kubernetes-hadoop:latest
          imagePullPolicy: IfNotPresent
          ports:
            - containerPort: 8040
            - containerPort: 8041   
            - containerPort: 8042        
          env:
            - name: HADOOP_NODE_TYPE
              value: yarnnode
            - name: HDFS_MASTER_SERVICE
              valueFrom:
                configMapKeyRef:
                  name: kube-hadoop-conf
                  key: HDFS_MASTER_SERVICE
            - name: HDOOP_YARN_MASTER
              valueFrom:
                configMapKeyRef:
                  name: kube-hadoop-conf
                  key: HDOOP_YARN_MASTER          
      restartPolicy: Always