摘要：一旦計(jì)算存儲(chǔ)分離后，就可真正實(shí)現(xiàn)隨意漂移，對(duì)集群維護(hù)來(lái)說(shuō)是個(gè)極大的福音。不過(guò)目前面臨最大的挑戰(zhàn)可能還是分布式存儲(chǔ)的性能問(wèn)題。

kubernetes集群三步安裝

git clone https://github.com/rook/rook

cd cluster/examples/kubernetes/ceph
kubectl create -f operator.yaml 

查看operator是否成功：

[root@dev-86-201 ~]# kubectl get pod -n rook-ceph-system
NAME                                  READY   STATUS    RESTARTS   AGE
rook-ceph-agent-5z6p7                 1/1     Running   0          88m
rook-ceph-agent-6rj7l                 1/1     Running   0          88m
rook-ceph-agent-8qfpj                 1/1     Running   0          88m
rook-ceph-agent-xbhzh                 1/1     Running   0          88m
rook-ceph-operator-67f4b8f67d-tsnf2   1/1     Running   0          88m
rook-discover-5wghx                   1/1     Running   0          88m
rook-discover-lhwvf                   1/1     Running   0          88m
rook-discover-nl5m2                   1/1     Running   0          88m
rook-discover-qmbx7                   1/1     Running   0          88m

然后創(chuàng)建ceph集群：

kubectl create -f cluster.yaml

查看ceph集群：

[root@dev-86-201 ~]# kubectl get pod -n rook-ceph
NAME                               READY   STATUS    RESTARTS   AGE
rook-ceph-mgr-a-8649f78d9b-jklbv   1/1     Running   0          64m
rook-ceph-mon-a-5d7fcfb6ff-2wq9l   1/1     Running   0          81m
rook-ceph-mon-b-7cfcd567d8-lkqff   1/1     Running   0          80m
rook-ceph-mon-d-65cd79df44-66rgz   1/1     Running   0          79m
rook-ceph-osd-0-56bd7545bd-5k9xk   1/1     Running   0          63m
rook-ceph-osd-1-77f56cd549-7rm4l   1/1     Running   0          63m
rook-ceph-osd-2-6cf58ddb6f-wkwp6   1/1     Running   0          63m
rook-ceph-osd-3-6f8b78c647-8xjzv   1/1     Running   0          63m

參數(shù)說(shuō)明：

apiVersion: ceph.rook.io/v1
kind: CephCluster
metadata:
  name: rook-ceph
  namespace: rook-ceph
spec:
  cephVersion:
    # For the latest ceph images, see https://hub.docker.com/r/ceph/ceph/tags
    image: ceph/ceph:v13.2.2-20181023
  dataDirHostPath: /var/lib/rook # 數(shù)據(jù)盤(pán)目錄
  mon:
    count: 3
    allowMultiplePerNode: true
  dashboard:
    enabled: true
  storage:
    useAllNodes: true
    useAllDevices: false
    config:
      databaseSizeMB: "1024"
      journalSizeMB: "1024"

訪問(wèn)ceph dashboard:

[root@dev-86-201 ~]# kubectl get svc -n rook-ceph
NAME                      TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)          AGE
rook-ceph-mgr             ClusterIP   10.98.183.33             9283/TCP         66m
rook-ceph-mgr-dashboard   NodePort    10.103.84.48             8443:31631/TCP   66m  # 把這個(gè)改成NodePort模式
rook-ceph-mon-a           ClusterIP   10.99.71.227             6790/TCP         83m
rook-ceph-mon-b           ClusterIP   10.110.245.119           6790/TCP         82m
rook-ceph-mon-d           ClusterIP   10.101.79.159            6790/TCP         81m

然后訪問(wèn)https://10.1.86.201:31631 即可

管理賬戶admin,獲取登錄密碼：

kubectl -n rook-ceph get secret rook-ceph-dashboard-password -o yaml | grep "password:" | awk "{print $2}" | base64 --decode

apiVersion: ceph.rook.io/v1
kind: CephBlockPool
metadata:
  name: replicapool   # operator會(huì)監(jiān)聽(tīng)并創(chuàng)建一個(gè)pool，執(zhí)行完后界面上也能看到對(duì)應(yīng)的pool
  namespace: rook-ceph
spec:
  failureDomain: host
  replicated:
    size: 3
---
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
   name: rook-ceph-block    # 這里創(chuàng)建一個(gè)storage class, 在pvc中指定這個(gè)storage class即可實(shí)現(xiàn)動(dòng)態(tài)創(chuàng)建PV
provisioner: ceph.rook.io/block
parameters:
  blockPool: replicapool
  # The value of "clusterNamespace" MUST be the same as the one in which your rook cluster exist
  clusterNamespace: rook-ceph
  # Specify the filesystem type of the volume. If not specified, it will use `ext4`.
  fstype: xfs
# Optional, default reclaimPolicy is "Delete". Other options are: "Retain", "Recycle" as documented in https://kubernetes.io/docs/concepts/storage/storage-classes/
reclaimPolicy: Retain

在cluster/examples/kubernetes 目錄下，官方給了個(gè)worldpress的例子，可以直接運(yùn)行一下：

kubectl create -f mysql.yaml
kubectl create -f wordpress.yaml

查看PV PVC：

[root@dev-86-201 ~]# kubectl get pvc
NAME             STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS      AGE
mysql-pv-claim   Bound    pvc-a910f8c2-1ee9-11e9-84fc-becbfc415cde   20Gi       RWO            rook-ceph-block   144m
wp-pv-claim      Bound    pvc-af2dfbd4-1ee9-11e9-84fc-becbfc415cde   20Gi       RWO            rook-ceph-block   144m

[root@dev-86-201 ~]# kubectl get pv
NAME                                       CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS   CLAIM                    STORAGECLASS      REASON   AGE
pvc-a910f8c2-1ee9-11e9-84fc-becbfc415cde   20Gi       RWO            Retain           Bound    default/mysql-pv-claim   rook-ceph-block            145m
pvc-af2dfbd4-1ee9-11e9-84fc-becbfc415cde   20Gi       RWO            Retain           Bound    default/wp-pv-claim      rook-ceph-block            145m

看下yaml文件：

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: mysql-pv-claim
  labels:
    app: wordpress
spec:
  storageClassName: rook-ceph-block   # 指定storage class
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 20Gi  # 需要一個(gè)20G的盤(pán)

...

        volumeMounts:
        - name: mysql-persistent-storage
          mountPath: /var/lib/mysql
      volumes:
      - name: mysql-persistent-storage
        persistentVolumeClaim:
          claimName: mysql-pv-claim  # 指定上面定義的PVC

是不是非常簡(jiǎn)單。

要訪問(wèn)wordpress的話請(qǐng)把service改成NodePort類(lèi)型，官方給的是loadbalance類(lèi)型：

kubectl edit svc wordpress

[root@dev-86-201 kubernetes]# kubectl get svc
NAME              TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)        AGE
wordpress         NodePort    10.109.30.99           80:30130/TCP   148m

分布式存儲(chǔ)在容器集群中充當(dāng)非常重要的角色，使用容器集群一個(gè)非常重要的理念就是把集群當(dāng)成一個(gè)整體使用，如果你在使用中還關(guān)心單個(gè)主機(jī)，比如調(diào)度到某個(gè)節(jié)點(diǎn)，

掛載某個(gè)節(jié)點(diǎn)目錄等，必然會(huì)導(dǎo)致不能把云的威力百分之百發(fā)揮出來(lái)。 一旦計(jì)算存儲(chǔ)分離后，就可真正實(shí)現(xiàn)隨意漂移，對(duì)集群維護(hù)來(lái)說(shuō)是個(gè)極大的福音。

比如集群機(jī)器過(guò)保了需要下架，那么我們?cè)苹募軜?gòu)因?yàn)樗袞|西無(wú)單點(diǎn)，所以只需要簡(jiǎn)單驅(qū)逐改節(jié)點(diǎn)，然后下架即可，不用關(guān)心上面跑的是什么業(yè)務(wù)，不管是有狀態(tài)還是無(wú)

狀態(tài)的都可以自動(dòng)修復(fù)。 不過(guò)目前面臨最大的挑戰(zhàn)可能還是分布式存儲(chǔ)的性能問(wèn)題。 在性能要求不苛刻的場(chǎng)景下我是極推薦這種計(jì)算存儲(chǔ)分離架構(gòu)的。

探討可加QQ群：98488045