环境准备

一个部署完整的k8s集群，版本1.18.1
系统版本：CentOS7.2
docker版本：1.13.1

172.22.21.77 dev-learn-77 master
172.22.21.78 dev-learn-78 slave
172.22.21.79 dev-learn-79 slave

[root@dev-learn-77 ~]# kubectl get node -o wide
NAME           STATUS   ROLES    AGE    VERSION   INTERNAL-IP    EXTERNAL-IP   OS-IMAGE                KERNEL-VERSION                CONTAINER-RUNTIME
dev-learn-77   Ready    master   5d5h   v1.18.1   172.22.21.77   <none>        CentOS Linux 7 (Core)   3.10.0-1062.18.1.el7.x86_64   docker://1.13.1
dev-learn-78   Ready    <none>   5d     v1.18.1   172.22.21.78   <none>        CentOS Linux 7 (Core)   3.10.0-1062.18.1.el7.x86_64   docker://1.13.1
dev-learn-79   Ready    <none>   5d1h   v1.18.1   172.22.21.79   <none>        CentOS Linux 7 (Core)   3.10.0-1062.18.1.el7.x86_64   docker://1.13.1
[root@dev-learn-77 ~]# 

使用的是主机Host网络
存储使用本地文件系统

准备yaml文件

• minio-distributed-daemonset.yaml
  注意环境变量中的MINIO_ACCESS_KEY和MINIO_SECRET_KEY，将会是登录时的用户名和密码，可以任意更改。

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: minio
  labels:
    app: minio
spec:
  selector:
    matchLabels:
      app: minio
  template:
    metadata:
      labels:
        app: minio
    spec:
      # We only deploy minio to the specified nodes. select your nodes by using `kubectl label node hostname1 -l minio-server=true`
      nodeSelector:
        minio-server: "true"
      # This is to maximize network performance, the headless service can be used to connect to a random host.
      hostNetwork: true
      # We're just using a hostpath. This path must be the same on all servers, and should be the largest, fastest block device you can fit.
      volumes:
      - name: storage
        hostPath:
          path: /mounts/minio1
      containers:
      - name: minio
        env:
        - name: MINIO_ACCESS_KEY
          value: "minio"
        - name: MINIO_SECRET_KEY
          value: "minio123"
        image: minio/minio:RELEASE.2020-04-04T05-39-31Z
        # Unfortunately you must manually define each server. Perhaps autodiscovery via DNS can be implemented in the future.
        args:
        - server
        - http://dev-learn-7{7...9}/mnt/disk{1...2}/minio/minio1/data
        ports:
        - containerPort: 9000
        volumeMounts:
        - name: storage
          mountPath: /mounts/minio1/

• minio-distributed-headless-service.yaml

apiVersion: v1
kind: Service
metadata:
  name: minio
  labels:
    app: minio
spec:
  publishNotReadyAddresses: true
  clusterIP: None
  ports:
    - port: 9000
      name: minio
  selector:
    app: minio

注意：集群模式的minio至少需要四块硬盘，否则是无法启动的，所以，这里：
- http://dev-learn-7{7...9}/mnt/disk{1...3}/minio/minio1/data
总数必须超过4个， 我这里是6块硬盘。

磁盘挂载方式我选择将/mnt/disk1-2/minio/minio1/data/以bind的方式挂载到/mounts/minio1/mnt/disk1-2/minio/minio1/data/下，从而方便将主机的文件系统映射到container里。每个节点都执行

[root@dev-learn-77 ~]# mkdir -p  /mounts/minio1/mnt/disk1/minio/minio1/data 
[root@dev-learn-77 ~]# mkdir -p  /mounts/minio1/mnt/disk2/minio/minio1/data 
[root@dev-learn-77 ~]# 
[root@dev-learn-77 ~]# mkdir -p  /mnt/disk1/minio/minio1/data/
[root@dev-learn-77 ~]# mkdir -p  /mnt/disk2/minio/minio1/data/
[root@dev-learn-77 ~]# 
[root@dev-learn-77 ~]# mount --bind /mnt/disk1/minio/minio1/data/ /mounts/minio1/mnt/disk1/minio/minio1/data/
[root@dev-learn-77 ~]# mount --bind /mnt/disk2/minio/minio1/data/ /mounts/minio1/mnt/disk2/minio/minio1/data/
[root@dev-learn-77 ~]# 
[root@dev-learn-77 ~]# echo "/mnt/disk1/minio/minio1/data    /mounts/minio1/mnt/disk1/minio/minio1/data    none    bind    0    0">>/etc/fstab
[root@dev-learn-77 ~]# echo "/mnt/disk2/minio/minio1/data    /mounts/minio1/mnt/disk2/minio/minio1/data    none    bind    0    0">>/etc/fstab

然后将挂载信息，写入/etc/fstab中

echo "/mnt/disk1/minio/minio1/data    /mounts/minio1/mnt/disk1/minio/minio1/data    none    bind    0    0">>/etc/fstab
echo "/mnt/disk2/minio/minio1/data    /mounts/minio1/mnt/disk2/minio/minio1/data    none    bind    0    0">>/etc/fstab

创建minio集群

[root@dev-learn-77 minio]# kubectl label node dev-learn-77 minio-server=true
node/dev-learn-77 labeled
[root@dev-learn-77 minio]# kubectl label node dev-learn-78 minio-server=true
node/dev-learn-78 labeled
[root@dev-learn-77 minio]# kubectl label node dev-learn-79 minio-server=true
node/dev-learn-79 labeled

[root@dev-learn-77 minio]# kubectl create -f minio-distributed-headless-service.yaml 
service/minio created
[root@dev-learn-77 minio]# kubectl create -f minio-distributed-daemonset.yaml 
daemonset.apps/minio created
[root@dev-learn-77 minio]# kubectl get service
NAME         TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)    AGE
kubernetes   ClusterIP   10.96.0.1    <none>        443/TCP    5d5h
minio        ClusterIP   None         <none>        9000/TCP   43s
[root@dev-learn-77 minio]# kubectl get daemonset
NAME    DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR       AGE
minio   2         2         1       2            1           minio-server=true   44s
[root@dev-learn-77 minio]# 
[root@dev-learn-77 minio]# kubectl get pod -o wide
NAME          READY   STATUS    RESTARTS   AGE    IP             NODE           NOMINATED NODE   READINESS GATES
minio-2cgbg   1/1     Running   0          109s   172.22.21.79   dev-learn-79   <none>           <none>
minio-cxdzl   1/1     Running   0          109s   172.22.21.78   dev-learn-78   <none>           <none>
[root@dev-learn-77 minio]# 

但是发现只有78和79参与了调度，77master节点并没有pod的启动。
因为出于安全考虑，k8s默认禁止master参与调度普通pod
执行如下命令解除限制：

[root@dev-learn-77 minio]# kubectl taint node dev-learn-77 node-role.kubernetes.io/master-
node/dev-learn-77 untainted

然后就可以正常调度三个pod，作为集群模式

[root@dev-learn-77 minio]# kubectl get pod -o wide
NAME          READY   STATUS    RESTARTS   AGE     IP             NODE           NOMINATED NODE   READINESS GATES
minio-2cgbg   1/1     Running   0          11m     172.22.21.79   dev-learn-79   <none>           <none>
minio-5jzql   1/1     Running   0          6m45s   172.22.21.77   dev-learn-77   <none>           <none>
minio-cxdzl   1/1     Running   0          11m     172.22.21.78   dev-learn-78   <none>           <none>
[root@dev-learn-77 minio]# 

然后浏览器输入三个节点中任意一个ip:9000，即可访问:
用户名密码是daemonset中设置的minio/minio123

之所以没用kubenetes的service，是因为环境比较简陋，没有DNS，所以即使使用service，也只能在集群内部使用，外面的网络是无法访问的，所以出次下册。

另一方面存储使用的是主机文件系统，当然也可以使用PV/PVC的方式，更好不过了。

因为没有使用service，所以负载均衡和高可用，需要另作打算，将在下一篇文章中讲述。