kubernetes operator influxdb-operator 实践

1.代码,镜像, 二进制文件

  https://github.com/xishengcai/influxdata-operator.git   fork 之官方

2. build 包中已经编译好二进制文件

  cd influxdata-operator
  docker build ./build
  docker images
  docker tag {image_id} influxdb-operator:v1.0

  

3. 创建存储

  创建 pv,pvc, storageclass

kuectl create -f deploy/storage.yaml

  修改pv  nodeAffinity

  nodeAffinity:
    required:
      nodeSelectorTerms:
      - matchExpressions:
        - key: beta.kubernetes.io/arch
          operator: In
          values:
          - amd64 

  pvc 中声明storageclass 

  pv 中也声明storageclass

  pvc 最终绑定到符合storageclass条件的pv, 如果没有符合条件的pv则会pending

  storageclass 动态绑定,将pv归类,pvc 绑定符合声明的那一类pv

 

 

 

4. 创建目录

mkdir /var/lib/influxdb

  

5. 创建influxdb-operator

kubectl create -f bundle.yaml

  crd:Influxd

    规定了 镜像,副本数目

apiVersion: influxdata.com/v1alpha1
kind: Influxdb
metadata:
  name: influxdb
spec:
  # Add fields here
  size: 1
  baseImage: influxdb:1.6.3-alpine
  imagePullPolicy: IfNotPresent
  pod:
    resources:
      limits:
        cpu: 8
        memory: 16Gi
      requests:
        cpu: 0.1
        memory: 256Mi
    persistentVolumeClaim:
      metadata:
        name: "influxdb-data-pvc"
      spec:
        accessModes: [ "ReadWriteOnce" ]
        storageClassName: standard-resize
        resources:
          requests:
            storage: 8Gi

  operator: influxdata-operator

    启动operator deployment, 根据crd中的信息控制influxdb statefulset 副本状态

    operator源码中的controller-manager 监听 influxdb  statefulset 的状态,而 crd中存的就是期望的目标状态,如果不一致则进行update。

    listwatch下有add, delete, update,主要靠这三个方法实现期望状态。

    

apiVersion: apps/v1
kind: Deployment
metadata:
  name: influxdata-operator
spec:
  replicas: 1
  selector:
    matchLabels:
      name: influxdata-operator
  template:
    metadata:
      labels:
        name: influxdata-operator
    spec:
      serviceAccountName: influxdata-operator
      containers:
        - name: influxdata-operator
          # Replace this with the built image name
          image: influxdb-operator:v1.0
          ports:
          - containerPort: 60000
            name: metrics
          command:
          - influxdata-operator
          imagePullPolicy: IfNotPresent
          env:
            - name: WATCH_NAMESPACE
              valueFrom:
                fieldRef:
                  fieldPath: metadata.namespace
            - name: OPERATOR_NAME
              value: "influxdata-operator"

  

6. 检查pod状态

kubectl get pod

 

 

 

7. 获取service ip

kubectl get svc

  通过kubectl get all 可以看到default namespace创建的所有的资源对象

 

8.golang 代码测试

 代码地址:https://github.com/xishengcai/go_learn/blob/master/client/influxdb.go

   

 

9. 数据备份

  创建pv, pvc

  创建 crd: Backup

  kubectl create -f influxdata_v1alpha1_backup_cr_pv.yaml

apiVersion: influxdata.com/v1alpha1
kind: Backup
metadata:
  name: influxdb-backup
spec:
  podname: "influxdb-0"
  containername: "influxdb"
  # [ -database <db_name> ] Optional: If not specified, all databases are backed up.
  databases:
  # [ -shard <ID> ] Optional: If specified, then -retention <name> is required.
  shard:
  # [ -retention <rp_name> ] Optional: If not specified, the default is to use all retention policies. If specified, then -database is required.
  retention:
  # [ -start <timestamp> ] Optional: Not compatible with -since.
  start:
  # [ -end <timestamp> ] Optional:  Not compatible with -since. If used without -start, all data will be backed up starting from 1970-01-01.
  end:
  # [ -since <timestamp> ] Optional: Use -start instead, unless needed for legacy backup support.
  since:
  storage:
    provider: in2

   

  log:kubectl logs -f influxdata-operator-b5c7cdfdf-7sr85

    

    

 

10. 问题

  1.不同pod的pv没有实现数据同步

  2.没有读写分离

posted @ 2019-03-20 11:09  瓦肯船长  阅读(647)  评论(0编辑  收藏  举报