当前位置:网站首页>云原生(三十四) | Kubernetes篇之平台存储系统实战

云原生(三十四) | Kubernetes篇之平台存储系统实战

2022-08-11 11:32:00 Lanson

Kubernetes平台存储系统实战

一、块存储(RDB)

RDB: RADOS Block Devices

RADOS: Reliable, Autonomic Distributed Object Store

不能是RWX模式

1、配置

RWO:(ReadWriteOnce)

参考文档:Ceph Docs[1]

常用块存储 。RWO模式;STS删除,pvc不会删除,需要自己手动维护

apiVersion: ceph.rook.io/v1
kind: CephBlockPool
metadata:
  name: replicapool
  namespace: rook-ceph
spec:
  failureDomain: host  #容灾模式,host或者osd
  replicated:
    size: 2  #数据副本数量
---
apiVersion: storage.k8s.io/v1
kind: StorageClass  #存储驱动
metadata:
   name: rook-ceph-block
# Change "rook-ceph" provisioner prefix to match the operator namespace if needed
provisioner: rook-ceph.rbd.csi.ceph.com
parameters:
    # clusterID is the namespace where the rook cluster is running
    clusterID: rook-ceph
    # Ceph pool into which the RBD image shall be created
    pool: replicapool

    # (optional) mapOptions is a comma-separated list of map options.
    # For krbd options refer
    # https://docs.ceph.com/docs/master/man/8/rbd/#kernel-rbd-krbd-options
    # For nbd options refer
    # https://docs.ceph.com/docs/master/man/8/rbd-nbd/#options
    # mapOptions: lock_on_read,queue_depth=1024

    # (optional) unmapOptions is a comma-separated list of unmap options.
    # For krbd options refer
    # https://docs.ceph.com/docs/master/man/8/rbd/#kernel-rbd-krbd-options
    # For nbd options refer
    # https://docs.ceph.com/docs/master/man/8/rbd-nbd/#options
    # unmapOptions: force

    # RBD image format. Defaults to "2".
    imageFormat: "2"

    # RBD image features. Available for imageFormat: "2". CSI RBD currently supports only `layering` feature.
    imageFeatures: layering

    # The secrets contain Ceph admin credentials.
    csi.storage.k8s.io/provisioner-secret-name: rook-csi-rbd-provisioner
    csi.storage.k8s.io/provisioner-secret-namespace: rook-ceph
    csi.storage.k8s.io/controller-expand-secret-name: rook-csi-rbd-provisioner
    csi.storage.k8s.io/controller-expand-secret-namespace: rook-ceph
    csi.storage.k8s.io/node-stage-secret-name: rook-csi-rbd-node
    csi.storage.k8s.io/node-stage-secret-namespace: rook-ceph

    # Specify the filesystem type of the volume. If not specified, csi-provisioner
    # will set default as `ext4`. Note that `xfs` is not recommended due to potential deadlock
    # in hyperconverged settings where the volume is mounted on the same node as the osds.
    csi.storage.k8s.io/fstype: ext4

# Delete the rbd volume when a PVC is deleted
reclaimPolicy: Delete
allowVolumeExpansion: true

二、STS案例实战

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: sts-nginx
  namespacedefault
spec:
  selector:
    matchLabels:
      app: sts-nginx # has to match .spec.template.metadata.labels
  serviceName"sts-nginx"
  replicas3 # by default is 1
  template:
    metadata:
      labels:
        app: sts-nginx # has to match .spec.selector.matchLabels
    spec:
      terminationGracePeriodSeconds: 10
      containers:
      - name: sts-nginx
        image: nginx
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: www
          mountPath/usr/share/nginx/html
  volumeClaimTemplates:
  - metadata:
      name: www
    spec:
      accessModes: [ "ReadWriteOnce" ]
      storageClassName"rook-ceph-block"
      resources:
        requests:
          storage: 20Mi
---
apiVersion: v1
kind: Service
metadata:
  name: sts-nginx
  namespacedefault
spec:
  selector:
    app: sts-nginx
  type: ClusterIP
  ports:
  - name: sts-nginx
    port80
    targetPort80
    protocol: TCP

测试: 创建sts、修改nginx数据、删除sts、重新创建sts。他们的数据丢不丢,共享不共享

三、文件存储(CephFS)

1、配置

常用 文件存储。 RWX模式;如:10个Pod共同操作一个地方

参考文档:Ceph Docs[2]

apiVersion: ceph.rook.io/v1
kind: CephFilesystem
metadata:
  name: myfs
  namespace: rook-ceph # namespace:cluster
spec:
  # The metadata pool spec. Must use replication.
  metadataPool:
    replicated:
      size: 3
      requireSafeReplicaSizetrue
    parameters:
      # Inline compression mode for the data pool
      # Further reference: https://docs.ceph.com/docs/nautilus/rados/configuration/bluestore-config-ref/#inline-compression
      compression_mode:
        none
        # gives a hint (%) to Ceph in terms of expected consumption of the total cluster capacity of a given pool
      # for more info: https://docs.ceph.com/docs/master/rados/operations/placement-groups/#specifying-expected-pool-size
      #target_size_ratio: ".5"
  # The list of data pool specs. Can use replication or erasure coding.
  dataPools:
    - failureDomain: host
      replicated:
        size: 3
        # Disallow setting pool with replica 1this could lead to data loss without recovery.
        # Make sure you're *ABSOLUTELY CERTAIN* that is what you want
        requireSafeReplicaSize: true
      parameters:
        # Inline compression mode for the data pool
        # Further reference: https://docs.ceph.com/docs/nautilus/rados/configuration/bluestore-config-ref/#inline-compression
        compression_mode:
          none
          # gives a hint (%) to Ceph in terms of expected consumption of the total cluster capacity of a given pool
        # for more info: https://docs.ceph.com/docs/master/rados/operations/placement-groups/#specifying-expected-pool-size
        #target_size_ratio: ".5"
  # Whether to preserve filesystem after CephFilesystem CRD deletion
  preserveFilesystemOnDelete: true
  # The metadata service (mds) configuration
  metadataServer:
    # The number of active MDS instances
    activeCount: 1
    # Whether each active MDS instance will have an active standby with a warm metadata cache for faster failover.
    # If false, standbys will be available, but will not have a warm cache.
    activeStandby: true
    # The affinity rules to apply to the mds deployment
    placement:
      #  nodeAffinity:
      #    requiredDuringSchedulingIgnoredDuringExecution:
      #      nodeSelectorTerms:
      #      - matchExpressions:
      #        - key: role
      #          operator: In
      #          values:
      #          - mds-node
      #  topologySpreadConstraints:
      #  tolerations:
      #  - key: mds-node
      #    operator: Exists
      #  podAffinity:
      podAntiAffinity:
        requiredDuringSchedulingIgnoredDuringExecution:
          - labelSelector:
              matchExpressions:
                - key: app
                  operator: In
                  values:
                    - rook-ceph-mds
            # topologyKey: kubernetes.io/hostname will place MDS across different hosts
            topologyKey: kubernetes.io/hostname
        preferredDuringSchedulingIgnoredDuringExecution:
          - weight: 100
            podAffinityTerm:
              labelSelector:
                matchExpressions:
                  - key: app
                    operator: In
                    values:
                      - rook-ceph-mds
              # topologyKey: */zone can be used to spread MDS across different AZ
              # Use <topologyKey: failure-domain.beta.kubernetes.io/zone> in k8s cluster if your cluster is v1.16 or lower
              # Use <topologyKey: topology.kubernetes.io/zone>  in k8s cluster is v1.17 or upper
              topologyKey: topology.kubernetes.io/zone
    # A key/value list of annotations
    annotations:
    #  key: value
    # A key/value list of labels
    labels:
    #  key: value
    resources:
    # The requests and limits set here, allow the filesystem MDS Pod(s) to use half of one CPU core and 1 gigabyte of memory
    #  limits:
    #    cpu: "500m"
    #    memory: "1024Mi"
    #  requests:
    #    cpu: "500m"
    #    memory: "1024Mi"
    # priorityClassName: my-priority-class
  mirroring:
    enabled: false

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: rook-cephfs
# Change "rook-ceph" provisioner prefix to match the operator namespace if needed
provisioner: rook-ceph.cephfs.csi.ceph.com
parameters:
  # clusterID is the namespace where operator is deployed.
  clusterID: rook-ceph

  # CephFS filesystem name into which the volume shall be created
  fsName: myfs
  # Ceph pool into which the volume shall be created
  # Required for provisionVolume: "true"
  pool: myfs-data0
  # The secrets contain Ceph admin credentials. These are generated automatically by the operator
  # in the same namespace as the cluster.
  csi.storage.k8s.io/provisioner-secret-name: rook-csi-cephfs-provisioner
  csi.storage.k8s.io/provisioner-secret-namespace: rook-ceph
  csi.storage.k8s.io/controller-expand-secret-name: rook-csi-cephfs-provisioner
  csi.storage.k8s.io/controller-expand-secret-namespace: rook-ceph
  csi.storage.k8s.io/node-stage-secret-name: rook-csi-cephfs-node
  csi.storage.k8s.io/node-stage-secret-namespace: rook-ceph

reclaimPolicy: Delete
allowVolumeExpansiontrue

2、测试

apiVersion: apps/v1
kind: Deployment
metadata:
  name:  nginx-deploy
  namespacedefault
  labels:
    app:  nginx-deploy
spec:
  selector:
    matchLabels:
      app: nginx-deploy
  replicas3
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
    type: RollingUpdate
  template:
    metadata:
      labels:
        app:  nginx-deploy
    spec:
      containers:
      - name:  nginx-deploy
        image:  nginx
        volumeMounts:
        - name: localtime
          mountPath/etc/localtime
        - name: nginx-html-storage
          mountPath/usr/share/nginx/html
      volumes:
        - name: localtime
          hostPath:
            path: /usr/share/zoneinfo/Asia/Shanghai
        - name: nginx-html-storage
          persistentVolumeClaim:
            claimName: nginx-pv-claim
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: nginx-pv-claim
  labels:
    app:  nginx-deploy
spec:
  storageClassName: rook-cephfs
  accessModes:
    - ReadWriteMany  ##如果是ReadWriteOnce将会是什么效果
  resources:
    requests:
      storage: 10Mi

测试,创建deploy、修改页面、删除deploy,新建deploy是否绑定成功,数据是否在

四、pvc扩容

参照CSI(容器存储接口)文档:

卷扩容:Ceph Docs[3]

动态卷扩容

之前创建storageclass的时候已经配置好了

测试:去容器挂载目录 curl -O 某个大文件 默认不能下载

修改原来的PVC,可以扩充容器。

注意,只能扩容,不能缩容

有状态应用(3个副本)使用块存储。自己操作自己的pvc挂载的pv;也不丢失

无状态应用(3个副本)使用共享存储。很多人操作一个pvc挂载的一个pv;也不丢失

  • 其他Pod可以对数据进行修改

  • MySQL 有状态做成主节点。。。MySQL - Master ---- pv

  • MySQL 无状态只读 挂载master的 pvc。

参考资料

[1]

Ceph Docs: https://www.rook.io/docs/rook/v1.6/ceph-block.html

[2]

Ceph Docs: https://rook.io/docs/rook/v1.6/ceph-filesystem.html

[3]

Ceph Docs: https://www.rook.io/docs/rook/v1.6/ceph-csi-drivers.html#dynamically-expand-volume

原网站

版权声明
本文为[Lanson]所创,转载请带上原文链接,感谢
https://mdnice.com/writing/138f532d8157462e9b85ccf753e3a857