浅析云原生分布式文件系统JuiceFS-天翼云开发者社区

1 JuiceFS 技术架构

JuiceFS 文件系统由三个部分组成：

JuiceFS 客户端（Client）：所有文件读写，以及碎片合并、回收站文件过期删除等后台任务，均在客户端发生。客户端需要同时与对象存储和元数据引擎打交道。客户端支持多种接入方式：

通过 FUSE：JuiceFS 文件系统能够以 POSIX 兼容的方式挂载到服务器，将海量云端存储直接当做本地存储来使用。
通过 Hadoop Java SDK：JuiceFS 文件系统能够直接替代 HDFS，为 Hadoop 提供低成本的海量存储。
通过 Kubernetes CSI 驱动：JuiceFS 文件系统能够直接为 Kubernetes 提供海量存储。
通过 S3 网关：使用 S3 作为存储层的应用可直接接入，同时可以使用 AWS CLI、s3cmd、MinIO client 等工具访问 JuiceFS 文件系统。
通过 WebDAV 服务：以 HTTP 协议，以类似 RESTful API 的方式接入 JuiceFS 并直接操作其中的文件。

数据存储（Data Storage）：文件将会被切分上传至对象存储服务。JuiceFS 支持几乎所有的公有云对象存储，同时也支持 OpenStack Swift、Ceph、MinIO 等私有化的对象存储。

元数据引擎（Metadata Engine）：用于存储文件元数据（metadata），包含以下内容：

常规文件系统的元数据：文件名、文件大小、权限信息、创建修改时间、目录结构、文件属性、符号链接、文件锁等。
文件数据的索引：文件的数据分配和引用计数、客户端会话等。

JuiceFS 采用多引擎设计，目前已支持 Redis、TiKV、MySQL/MariaDB、PostgreSQL、SQLite 等作为元数据服务引擎。

2 JuiceFS CSI 架构

JuiceFS CSI 驱动遵循 CSI 规范，实现了容器编排系统与 JuiceFS 文件系统之间的接口。在 Kubernetes 下，JuiceFS 可以用持久卷（PersistentVolume）的形式提供给 Pod 使用。

JuiceFS CSI 驱动包含以下组件：JuiceFS CSI Controller（StatefulSet）以及 JuiceFS CSI Node Service（DaemonSet），你可以方便的用 kubectl 查看：

 

root@node1:~# kubectl get pod -l app.kubernetes.io/name=juicefs-csi-driver
NAME                       READY   STATUS    RESTARTS   AGE
juicefs-csi-controller-0   4/4     Running   0          26m
juicefs-csi-controller-1   4/4     Running   0          26m
juicefs-csi-node-5qr7b     3/3     Running   0          26m

CSI 默认采用容器挂载（Mount Pod）模式，也就是让 JuiceFS 客户端运行在独立的 Pod 中，其架构如下：

采用独立 Mount Pod 来运行 JuiceFS 客户端，并由 CSI Node Service 来管理 Mount Pod 的生命周期。这样的架构提供如下好处：

多个 Pod 共用 PV 时，不会新建 Mount Pod，而是对已有的 Mount Pod 做引用计数，计数归零时删除 Mount Pod。
CSI 驱动组件与客户端解耦，方便 CSI 驱动自身的升级。详见。

在同一个节点上，一个 PVC 会对应一个 Mount Pod。而使用了相同 PV 的容器，则可以共享一个 Mount Pod。PVC、PV、Mount Pod 之间的关系如下图所示：

3 使用方式

你可以以静态配置和动态配置的方式来使用 CSI 驱动。

3.1 静态配置

静态配置方式最为简单直接，需要 Kubernetes 管理员创建 PersistentVolume（PV）以及文件系统认证信息（以 Kubernetes Secret 形式保存），然后用户创建 PersistentVolumeClaim（PVC），在定义中绑定该 PV，最后在 Pod 定义中引用该 PVC。资源间关系如下图所示：

一般在以下场景使用静态配置：

你在 JuiceFS 中已经存储了大量数据，想要直接在 Kubernetes 容器中访问；
对 CSI 驱动功能做简单验证；

3.2 动态配置

考虑到静态配置的管理更加复杂，规模化使用 CSI 驱动时，一般会以「动态配置」方式使用，管理员不再需要手动创建 PV，同时实现应用间的数据隔离。这种模式下，管理员会负责创建一个或多个 StorageClass，用户只需要创建 PVC，指定 StorageClass，并且在 Pod 中引用该 PVC，CSI 驱动就会按照 StorageClass 中配置好的参数，为你自动创建 PV。资源间关系如下：

以容器挂载模式为例，从创建到使用的流程大致如下：

用户创建 PVC，指定已经创建好的 StorageClass；
CSI Controller 负责在 JuiceFS 文件系统中做初始化，默认以 PV ID 为名字创建子目录，同时创建对应的 PV。该过程所需的配置，都在 StorageClass 中指定或引用；
Kubernetes (PV Controller 组件) 将上述用户创建的 PVC 与 CSI Controller 创建的 PV 进行绑定，此时 PVC 与 PV 的状态变为「Bound」；
用户创建应用 Pod，声明使用先前创建的 PVC；
CSI Node Service 负责在应用 Pod 所在节点创建 Mount Pod；
Mount Pod 启动，执行 JuiceFS 客户端挂载，将挂载点暴露给宿主机，路径为 /var/lib/juicefs/volume/[pv-name]；
CSI Node Service 等待 Mount Pod 启动成功后，将 PV 对应的 JuiceFS 子目录 bind 到容器内，路径为其声明的 VolumeMount 路径；
Kubelet 启动应用 Pod。

4 部署 JuiceFS

元数据采用 Redis，数据存储使用 MinIO。Redis、MinIO 和 JuiceFS CSI 都统一采用 Helm 方式来部署。

4.1 部署 Redis

执行如下步骤添加 bitnmai 源，拉取 redis-cluster charts包：

 

root@node1:~# 
root@node1:~# helm search repo bitnami |grep redis
bitnami/redis                               	18.3.3       	7.2.3        	Redis(R) is an open source, advanced key-value ...
bitnami/redis-cluster                       	9.1.3        	7.2.3        	Redis(R) is an open source, scalable, distribut...
root@node1:~# helm search repo bitnami |grep redis-cluster
bitnami/redis-cluster                       	9.1.3        	7.2.3        	Redis(R) is an open source, scalable, distribut...
root@node1:~# 
root@node1:~# help pull bitnami/redis-cluster --version 9.1.3
root@node1:~# tar -xvzf redis-cluster-9.1.3.tgz
root@node1:~# cd redis-cluster/

其中，values.yaml 示例：

 

root@node1:~/redis-cluster# cat values.yaml |grep -v "#"


global:
  imageRegistry: "dockerhub.enlightencloud.local"
  imagePullSecrets: []
  storageClass: "local"
  redis:
    password: "Moya8788"


nameOverride: ""
fullnameOverride: ""
clusterDomain: cluster.local
commonAnnotations: {}
commonLabels: {}
extraDeploy: []

diagnosticMode:
  enabled: false
  command:
    - sleep
  args:
    - infinity

image:
  registry: dockerhub.enlightencloud.local
  repository: bitnami/redis-cluster
  tag: 7.2.3-debian-11-r1
  digest: ""
  pullPolicy: IfNotPresent
  pullSecrets: []
  debug: false
networkPolicy:
  enabled: false
  allowExternal: true
  ingressNSMatchLabels: {}
  ingressNSPodMatchLabels: {}

serviceAccount:
  create: false
  name: ""
  annotations: {}
  automountServiceAccountToken: false

rbac:
  create: false
  role:
    rules: []
podSecurityContext:
  enabled: true
  fsGroup: 1001
  sysctls: []
podDisruptionBudget: {}
minAvailable: ""
maxUnavailable: ""
containerSecurityContext:
  enabled: true
  runAsUser: 1001
  runAsNonRoot: true
  privileged: false
  readOnlyRootFilesystem: false
  allowPrivilegeEscalation: false
  capabilities:
    drop: ["ALL"]
  seccompProfile:
    type: "RuntimeDefault"
usePassword: true
password: ""
existingSecret: ""
existingSecretPasswordKey: ""
usePasswordFile: false
tls:
  enabled: false
  authClients: true
  autoGenerated: false
  existingSecret: ""
  certificatesSecret: ""
  certFilename: ""
  certKeyFilename: ""
  certCAFilename: ""
  dhParamsFilename: ""
service:
  ports:
    redis: 6379
  nodePorts:
    redis: ""
  extraPorts: []
  annotations: {}
  labels: {}
  type: ClusterIP
  clusterIP: ""
  loadBalancerIP: ""
  loadBalancerSourceRanges: []
  externalTrafficPolicy: Cluster
  sessionAffinity: None
  sessionAffinityConfig: {}
  headless:
    annotations: {}
persistence:
  enabled: true
  path: /bitnami/redis/data
  subPath: ""
  storageClass: ""
  annotations: {}
  accessModes:
    - ReadWriteOnce
  size: 8Gi
  matchLabels: {}
  matchExpressions: {}

persistentVolumeClaimRetentionPolicy:
  enabled: false
  whenScaled: Retain
  whenDeleted: Retain

volumePermissions:
  enabled: false
  image:
    registry: dockerhub.enlightencloud.local
    repository: bitnami/os-shell
    tag: 11-debian-11-r91
    digest: ""
    pullPolicy: IfNotPresent
    pullSecrets: []
  containerSecurityContext:
    enabled: true
    runAsUser: 0
    privileged: false
  resources:
    limits: {}
    requests: {}
podSecurityPolicy:
  create: false


redis:
  command: []
  args: []
  updateStrategy:
    type: RollingUpdate
    rollingUpdate:
      partition: 0

  podManagementPolicy: Parallel
  hostAliases: []
  hostNetwork: false
  useAOFPersistence: "yes"
  containerPorts:
    redis: 6379
    bus: 16379
  lifecycleHooks: {}
  extraVolumes: []
  extraVolumeMounts: []
  customLivenessProbe: {}
  customReadinessProbe: {}
  customStartupProbe: {}
  initContainers: []
  sidecars: []
  podLabels: {}
  priorityClassName: ""
  defaultConfigOverride: ""
  configmap: ""
  extraEnvVars: []
  extraEnvVarsCM: ""
  extraEnvVarsSecret: ""
  podAnnotations: {}
  resources:
    limits: {}
    requests: {}
  schedulerName: ""
  shareProcessNamespace: false
  livenessProbe:
    enabled: true
    initialDelaySeconds: 5
    periodSeconds: 5
    timeoutSeconds: 5
    successThreshold: 1
    failureThreshold: 5
  readinessProbe:
    enabled: true
    initialDelaySeconds: 5
    periodSeconds: 5
    timeoutSeconds: 1
    successThreshold: 1
    failureThreshold: 5
  startupProbe:
    enabled: false
    path: /
    initialDelaySeconds: 300
    periodSeconds: 10
    timeoutSeconds: 5
    failureThreshold: 6
    successThreshold: 1
  podAffinityPreset: ""
  podAntiAffinityPreset: soft
  nodeAffinityPreset:
    type: ""
    key: ""
    values: []
  affinity: {}
  nodeSelector: {}
  tolerations: []
  topologySpreadConstraints: []


updateJob:
  activeDeadlineSeconds: 600
  command: []
  args: []
  hostAliases: []
  helmHook: post-upgrade
  annotations: {}
  podAnnotations: {}
  podLabels: {}
  extraEnvVars: []
  extraEnvVarsCM: ""
  extraEnvVarsSecret: ""
  extraVolumes: []
  extraVolumeMounts: []
  initContainers: []
  podAffinityPreset: ""
  podAntiAffinityPreset: soft
  nodeAffinityPreset:
    type: ""
    key: ""
    values: []
  affinity: {}
  nodeSelector: {}
  tolerations: []
  priorityClassName: ""
  resources:
    limits: {}
    requests: {}


cluster:
  init: true
  nodes: 6
  replicas: 1
  externalAccess:
    enabled: false
    hostMode: false
    service:
      disableLoadBalancerIP: false
      loadBalancerIPAnnotaion: ""
      type: LoadBalancer
      port: 6379
      loadBalancerIP: []
      loadBalancerSourceRanges: []
      annotations: {}
  update:
    addNodes: false
    currentNumberOfNodes: 6
    currentNumberOfReplicas: 1
    newExternalIPs: []


metrics:
  enabled: false
  image:
    registry: dockerhub.enlightencloud.local
    repository: bitnami/redis-exporter
    tag: 1.55.0-debian-11-r2
    digest: ""
    pullPolicy: IfNotPresent
    pullSecrets: []
  resources: {}
  extraArgs: {}
  extraEnvVars: []
  podAnnotations:
    prometheus.io/scrape: "true"
    prometheus.io/port: "9121"
  podLabels: {}
  containerSecurityContext:
    enabled: false
    allowPrivilegeEscalation: false
  serviceMonitor:
    enabled: false
    namespace: ""
    interval: ""
    scrapeTimeout: ""
    selector: {}
    labels: {}
    annotations: {}
    jobLabel: ""
    relabelings: []
    metricRelabelings: []
  prometheusRule:
    enabled: false
    additionalLabels: {}
    namespace: ""
    rules: []
  priorityClassName: ""
  service:
    type: ClusterIP
    clusterIP: ""
    loadBalancerIP: ""
    annotations: {}
    labels: {}


sysctlImage:
  enabled: false
  command: []
  registry: 
  repository: bitnami/os-shell
  tag: 11-debian-11-r91
  digest: ""
  pullPolicy: IfNotPresent
  pullSecrets: []
  mountHostSys: false
  containerSecurityContext:
    enabled: true
    runAsUser: 0
    privileged: true
  resources:
    limits: {}
    requests: {}

执行如下命令部署 Redis 集群：

 

root@node1:~/redis-cluster# helm install juicefs-redis .
NAME: juicefs-redis
LAST DEPLOYED: Wed Nov 15 16:24:40 2023
NAMESPACE: default
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
CHART NAME: redis-cluster
CHART VERSION: 9.1.3
APP VERSION: 7.2.3** Please be patient while the chart is being deployed **


To get your password run:
    export REDIS_PASSWORD=$(kubectl get secret --namespace "default" juicefs-redis-redis-cluster -o jsonpath="{.data.redis-password}" | base64 -d)

You have deployed a Redis&reg; Cluster accessible only from within you Kubernetes Cluster.INFO: The Job to create the cluster will be created.To connect to your Redis&reg; cluster:

1. Run a Redis&reg; pod that you can use as a client:
kubectl run --namespace default juicefs-redis-redis-cluster-client --rm --tty -i --restart='Never' \
 --env REDIS_PASSWORD=$REDIS_PASSWORD \
--image dockerhub.enlightencloud.local/bitnami/redis-cluster:7.2.3-debian-11-r1 -- bash

2. Connect using the Redis&reg; CLI:

redis-cli -c -h juicefs-redis-redis-cluster -a $REDIS_PASSWORD
root@node1:~/redis-cluster# 
root@node1:~/redis-cluster# export REDIS_PASSWORD=$(kubectl get secret --namespace "default" juicefs-redis-redis-cluster -o jsonpath="{.data.redis-password}" | base64 -d)
root@node1:~/redis-cluster# kubectl run --namespace default juicefs-redis-redis-cluster-client --rm --tty -i --restart='Never' \
>  --env REDIS_PASSWORD=$REDIS_PASSWORD \
> --image dockerhub.enlightencloud.local/bitnami/redis-cluster:7.2.3-debian-11-r1 -- bash
If you don't see a command prompt, try pressing enter.
I have no name!@juicefs-redis-redis-cluster-client:/$ redis-cli -c -h juicefs-redis-redis-cluster -a $REDIS_PASSWORD
Warning: Using a password with '-a' or '-u' option on the command line interface may not be safe.
juicefs-redis-redis-cluster:6379> 
juicefs-redis-redis-cluster:6379> config get databases
1) "databases"
2) "1"
juicefs-redis-redis-cluster:6379> select 0
OK
juicefs-redis-redis-cluster:6379> keys

查看 Redis 资源：

 

root@node1:~# kubectl get pods |grep redis
juicefs-redis-redis-cluster-0                                   1/1     Running   1 (101m ago)   103m
juicefs-redis-redis-cluster-1                                   1/1     Running   1 (101m ago)   103m
juicefs-redis-redis-cluster-2                                   1/1     Running   1 (101m ago)   103m
juicefs-redis-redis-cluster-3                                   1/1     Running   1 (101m ago)   103m
juicefs-redis-redis-cluster-4                                   1/1     Running   0              103m
juicefs-redis-redis-cluster-5                                   1/1     Running   0              102m
juicefs-redis-redis-cluster-client                              1/1     Running   0              100m
root@node1:~/redis-cluster# 
root@node1:~# kubectl get services |grep redis
juicefs-redis-redis-cluster            ClusterIP   10.233.34.103   <none>        6379/TCP             103m
juicefs-redis-redis-cluster-headless   ClusterIP   None            <none>        6379/TCP,16379/TCP   103m

4.2 部署 MinIO

执行如下步骤添加 minio 源，并拉取 minio charts包：

 

root@node1:~# helm repo add minio xxx
root@node1:~# helm search repo minio 
NAME         	CHART VERSION	APP VERSION	DESCRIPTION                                       
aliyun/minio 	0.5.5        	           	Distributed object storage server built for clo...
bitnami/minio	12.10.0      	2023.11.11 	MinIO(R) is an object storage server, compatibl...
minio/minio  	8.0.10       	master     	High Performance, Kubernetes Native Object Storage
root@node1:~# helm pull minio/minio --version 8.0.10
root@node1:~# tar -xvzf minio-8.0.10.tgz
root@node1:~# cd minio/

其中，values.yaml 示例：

 

root@node1:~/minio# cat values.yaml |grep -v "#"
nameOverride: ""

fullnameOverride: ""

clusterDomain: cluster.local

image:
  repository: dockerhub.enlightencloud.local/minio/minio
  tag: RELEASE.2021-02-14T04-01-33Z
  pullPolicy: IfNotPresent

mcImage:
  repository: dockerhub.enlightencloud.local/minio/mc
  tag: RELEASE.2021-02-14T04-28-06Z
  pullPolicy: IfNotPresent

helmKubectlJqImage:
  repository: dockerhub.enlightencloud.local/minio/helm-kubectl-jq
  tag: 3.1.0
  pullPolicy: IfNotPresent

mode: standalone

additionalLabels: []

additionalAnnotations: []

extraArgs: []

DeploymentUpdate:
  type: RollingUpdate
  maxUnavailable: 0
  maxSurge: 100%

StatefulSetUpdate:
  updateStrategy: RollingUpdate

priorityClassName: ""

accessKey: "Moya8788"
secretKey: "Moya8788"
certsPath: "/etc/minio/certs/"
configPathmc: "/etc/minio/mc/"
mountPath: "/export"

existingSecret: ""

bucketRoot: ""

drivesPerNode: 1
replicas: 4
zones: 1

tls:
  enabled: false
  certSecret: ""
  publicCrt: public.crt
  privateKey: private.key

trustedCertsSecret: ""

persistence:
  enabled: true

  existingClaim: ""

  storageClass: "local"
  VolumeName: ""
  accessMode: ReadWriteOnce
  size: 50Gi

  subPath: ""


service:
  type: ClusterIP
  clusterIP: ~
  port: 9000
  nodePort: 32000

  externalIPs: []

  annotations: {}


imagePullSecrets: []

ingress:
  enabled: false
  labels: {}

  annotations: {}
  path: /
  hosts:
    - chart-example.local
  tls: []

nodeSelector: {}
tolerations: []
affinity: {}

securityContext:
  enabled: true
  runAsUser: 1000
  runAsGroup: 1000
  fsGroup: 1000

podAnnotations: {}

podLabels: {}

resources:
  requests:
    memory: 4Gi

defaultBucket:
  enabled: false
  name: bucket
  policy: none
  purge: false

buckets: []

makeBucketJob:
  podAnnotations:
  annotations:
  securityContext:
    enabled: false
    runAsUser: 1000
    runAsGroup: 1000
    fsGroup: 1000
  resources:
    requests:
      memory: 128Mi

updatePrometheusJob:
  podAnnotations:
  annotations:
  securityContext:
    enabled: false
    runAsUser: 1000
    runAsGroup: 1000
    fsGroup: 1000

s3gateway:
  enabled: false
  replicas: 4
  serviceEndpoint: ""
  accessKey: ""
  secretKey: ""

azuregateway:
  enabled: false
  replicas: 4


gcsgateway:
  enabled: false
  replicas: 4
  gcsKeyJson: ""
  projectId: ""


nasgateway:
  enabled: false
  replicas: 4
  pv: ~

environment: {}

networkPolicy:
  enabled: false
  allowExternal: true

podDisruptionBudget:
  enabled: false
  maxUnavailable: 1

serviceAccount:
  create: true
  name:

metrics:
  serviceMonitor:
    enabled: false
    additionalLabels: {}
    relabelConfigs: {}

etcd:
  endpoints: []
  pathPrefix: ""
  corednsPathPrefix: ""
  clientCert: ""
  clientCertKey: ""

执行如下命令部署 MinIO：

root@node1:~/minio# helm install juicefs-minio .

查看 MinIO 资源：

 

root@node1:~/minio# kubectl get pods |grep minio
juicefs-minio-58bfcf5574-5nrwh                                  1/1     Running   0              75m
root@node1:~/minio# 
root@node1:~/minio# kubectl get service |grep minio
juicefs-minio                          ClusterIP   10.233.61.184   <none>        9000/TCP             75m

4.3 部署 JuiceFS CSI

执行如下命令添加 juicefs 的源，下载 jucefs-csi-driver charts：

 

root@node1:~# helm repo add juicefs xxx
root@node1:~# helm pull juicefs/juicefs-csi-driver --version 0.18.1
root@node1:~# tar -xvzf juicefs-csi-driver-0.18.1.tgz
root@node1:~# cd juicefs-csi-driver/

其中，values.yaml 示例：

 

root@node1:~/juicefs-csi-driver# cat values.yaml |grep -v "#"


image:
  repository: dockerhub.enlightencloud.local/juicedata/juicefs-csi-driver
  tag: "v0.22.1"
  pullPolicy: ""

sidecars:
  livenessProbeImage:
    repository: dockerhub.enlightencloud.local/k8scsi/livenessprobe
    tag: "v1.1.0"
    pullPolicy: ""
  nodeDriverRegistrarImage:
    repository: dockerhub.enlightencloud.local/k8scsi/csi-node-driver-registrar
    tag: "v2.1.0"
    pullPolicy: ""
  csiProvisionerImage:
    repository: dockerhub.enlightencloud.local/k8scsi/csi-provisioner
    tag: "v1.6.0"
    pullPolicy: ""
  csiResizerImage:
    repository: dockerhub.enlightencloud.local/k8scsi/csi-resizer
    tag: "v1.0.1"
    pullPolicy: ""

imagePullSecrets: []

mountMode: mountpod

hostAliases: []

kubeletDir: /var/lib/kubelet

jfsMountDir: /var/lib/juicefs/volume
jfsConfigDir: /var/lib/juicefs/config

immutable: false

dnsPolicy: ClusterFirstWithHostNet
dnsConfig:
  {}

serviceAccount:
  controller:
    create: true
    annotations: {}
    name: "juicefs-csi-controller-sa"
  node:
    create: true
    annotations: {}
    name: "juicefs-csi-node-sa"

controller:
  enabled: true
  leaderElection:
    enabled: true
    leaderElectionNamespace: ""
    leaseDuration: ""
  provisioner: false
  replicas: 2
  resources:
    limits:
      cpu: 1000m
      memory: 1Gi
    requests:
      cpu: 100m
      memory: 512Mi
  terminationGracePeriodSeconds: 30
  affinity: {}
  nodeSelector: {}
  tolerations:
    - key: CriticalAddonsOnly
      operator: Exists
  service:
    port: 9909
    type: ClusterIP
  priorityClassName: system-cluster-critical
  envs:

node:
  enabled: true
  hostNetwork: false
  resources:
    limits:
      cpu: 1000m
      memory: 1Gi
    requests:
      cpu: 100m
      memory: 512Mi
  storageClassShareMount: false
  mountPodNonPreempting: false
  terminationGracePeriodSeconds: 30
  affinity: {}
  nodeSelector: {}
  tolerations:
    - key: CriticalAddonsOnly
      operator: Exists
  priorityClassName: system-node-critical
  envs: []
  updateStrategy:
    rollingUpdate:
      maxUnavailable: 50%

defaultMountImage:
  ce: ""
  ee: ""

webhook:
  certManager:
    enabled: true
  caBundlePEM: |

  crtPEM: |

  keyPEM: |

  timeoutSeconds: 5
  FailurePolicy: Fail

storageClasses:
- name: "juicefs-sc"
  enabled: true
  reclaimPolicy: Delete
  allowVolumeExpansion: true

  backend:
    name: "liusu-juicefs"
    metaurl: "redis://:Moya8788@juicefs-redis-redis-cluster:6379/0"
    storage: "minio"
   
    token: ""
    accessKey: "Moya8788"
    secretKey: "Moya8788"
    envs: ""
    configs: ""
    trashDays: ""
    formatOptions: ""

  mountOptions:

  pathPattern: ""

  cachePVC: ""

  mountPod:
    resources:
      limits:
        cpu: 5000m
        memory: 5Gi
      requests:
        cpu: 1000m
        memory: 1Gi
    image: ""

执行如下命令来部署 juicefs-csi-driver：

root@node1:~/juicefs-csi-driver# helm install liusu-juicefs .

查看资源：

 

root@node1:~/juicefs-csi-driver# kubectl get pods |grep csi
juicefs-csi-controller-0                                        4/4     Running   0              71m
juicefs-csi-controller-1                                        4/4     Running   0              70m
juicefs-csi-node-5qr7b                                          3/3     Running   0              71m
root@node1:~/juicefs-csi-driver# kubectl get storageclasses.storage.k8s.io -o wide |grep juice
juicefs-sc        csi.juicefs.com    Delete          Immediate              true                   71m
root@node1:~/juicefs-csi-driver#  
root@node1:~/juicefs-csi-driver# kubectl get storageclasses.storage.k8s.io -o wide juicefs-sc -o yaml
allowVolumeExpansion: true
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  annotations:
    meta.helm.sh/release-name: liusu-juicefs
    meta.helm.sh/release-namespace: default
    storageclass.kubesphere.io/allow-clone: "true"
    storageclass.kubesphere.io/allow-snapshot: "true"
  creationTimestamp: "2023-11-15T09:08:04Z"
  labels:
    app.kubernetes.io/instance: liusu-juicefs
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/name: juicefs-csi-driver
    app.kubernetes.io/version: 0.22.1
    helm.sh/chart: juicefs-csi-driver-0.18.1
  name: juicefs-sc
  resourceVersion: "16500"
  uid: a1d42963-2e80-43fa-a275-fd162032d408
parameters:
  csi.storage.k8s.io/controller-expand-secret-name: juicefs-sc-secret
  csi.storage.k8s.io/controller-expand-secret-namespace: default
  csi.storage.k8s.io/node-publish-secret-name: juicefs-sc-secret
  csi.storage.k8s.io/node-publish-secret-namespace: default
  csi.storage.k8s.io/provisioner-secret-name: juicefs-sc-secret
  csi.storage.k8s.io/provisioner-secret-namespace: default
  juicefs/mount-cpu-limit: 5000m
  juicefs/mount-cpu-request: 1000m
  juicefs/mount-memory-limit: 5Gi
  juicefs/mount-memory-request: 1Gi
provisioner: csi.juicefs.com
reclaimPolicy: Delete
volumeBindingMode: Immediate

4.4 验证结果

4.4.1 创建 PVC

如下所示创建了 juicefs 类型的 pvc：

 

root@node1:~/juicefs-csi-driver# kubectl get pvc  |grep liusu-juicefs
liusu-juicefs-pvc1                         Bound    pvc-157aa1a1-1e85-4396-a44e-7ea9d4a66b6e   10Gi       RWX            juicefs-sc     63m
liusu-juicefs-pvc2                         Bound    pvc-02a38e1b-e2d0-44c3-9917-090c008904cd   10Gi       RWX            juicefs-sc     62m
liusu-juicefs-pvc3                         Bound    pvc-a45192ad-f882-446f-b3a2-70cbfb8565cd   10Gi       RWX            juicefs-sc     62m

4.4.2 创建 Pod

使用 centos 容器镜像创建一个工作负载，挂载上述步骤创建的 pvc，挂载至容器的 /data 目录下，启动命令执行一个脚本向文件中循环打印日期：

 

root@node1:~/juicefs-csi-driver# kubectl get pods liusu-juicefs-test1-76bdcc77f7-8657p -o yaml
apiVersion: v1
kind: Pod
metadata:
  annotations:
    cni.projectcalico.org/containerID: f4862b25dad5b674a35c8c535220ab07a4da0034173dc1b30a90dfd9c6ad00f2
    cni.projectcalico.org/podIP: 10.233.90.39/32
    cni.projectcalico.org/podIPs: 10.233.90.39/32
    juicefs-mountpod-pvc-157aa1a1-1e85-4396-a44e-7ea9d4a66b6e: default/juicefs-node1-pvc-157aa1a1-1e85-4396-a44e-7ea9d4a66b6e-cfsuav
    kubesphere.io/creator: admin
    kubesphere.io/imagepullsecrets: '{}'
    logging.kubesphere.io/logsidecar-config: '{}'
  creationTimestamp: "2023-11-15T09:19:43Z"
  generateName: liusu-juicefs-test1-76bdcc77f7-
  labels:
    app: liusu-juicefs-test1
    juicefs-uniqueid: ""
    pod-template-hash: 76bdcc77f7
  name: liusu-juicefs-test1-76bdcc77f7-8657p
  namespace: default
  ownerReferences:
  - apiVersion: apps/v1
    blockOwnerDeletion: true
    controller: true
    kind: ReplicaSet
    name: liusu-juicefs-test1-76bdcc77f7
    uid: ced8e092-65a9-41b0-ac13-7e3be08ac647
  resourceVersion: "19965"
  uid: 8bb042a2-38b1-492b-9951-2d5012785028
spec:
  containers:
  - command:
    - sh
    - -c
    - while true; do echo $(date -u) >> /data/out1.txt; sleep 5; done
    image: centos:7.9.2009
    imagePullPolicy: IfNotPresent
    name: container-t0tqdm
    resources: {}
    terminationMessagePath: /dev/termination-log
    terminationMessagePolicy: File
    volumeMounts:
    - mountPath: /data
      name: volume-8qnjf8
    - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
      name: kube-api-access-w45h8
      readOnly: true
  dnsPolicy: ClusterFirst
  enableServiceLinks: true
  nodeName: node1
  preemptionPolicy: PreemptLowerPriority
  priority: 0
  restartPolicy: Always
  schedulerName: default-scheduler
  securityContext: {}
  serviceAccount: default
  serviceAccountName: default
  terminationGracePeriodSeconds: 30
  tolerations:
  - effect: NoExecute
    key: node.kubernetes.io/not-ready
    operator: Exists
    tolerationSeconds: 300
  - effect: NoExecute
    key: node.kubernetes.io/unreachable
    operator: Exists
    tolerationSeconds: 300
  volumes:
  - name: volume-8qnjf8
    persistentVolumeClaim:
      claimName: liusu-juicefs-pvc1
  - name: kube-api-access-w45h8
    projected:
      defaultMode: 420
      sources:
      - serviceAccountToken:
          expirationSeconds: 3607
          path: token
      - configMap:
          items:
          - key: ca.crt
            path: ca.crt
          name: kube-root-ca.crt
      - downwardAPI:
          items:
          - fieldRef:
              apiVersion: v1
              fieldPath: metadata.namespace
            path: namespace
status:
  conditions:
  - lastProbeTime: null
    lastTransitionTime: "2023-11-15T09:19:43Z"
    status: "True"
    type: Initialized
  - lastProbeTime: null
    lastTransitionTime: "2023-11-15T09:20:44Z"
    status: "True"
    type: Ready
  - lastProbeTime: null
    lastTransitionTime: "2023-11-15T09:20:44Z"
    status: "True"
    type: ContainersReady
  - lastProbeTime: null
    lastTransitionTime: "2023-11-15T09:19:43Z"
    status: "True"
    type: PodScheduled
  containerStatuses:
  - containerID: docker://8af50aa5d547dc5dec1a6deaca900311051af2b70ba86de62f72382421433706
    image: centos:7.9.2009
    imageID: docker-pullable://centos@sha256:be65f488b7764ad3638f236b7b515b3678369a5124c47b8d32916d6487418ea4
    lastState: {}
    name: container-t0tqdm
    ready: true
    restartCount: 0
    started: true
    state:
      running:
        startedAt: "2023-11-15T09:20:44Z"
  hostIP: 172.24.21.221
  phase: Running
  podIP: 10.233.90.39
  podIPs:
  - ip: 10.233.90.39
  qosClass: BestEffort
  startTime: "2023-11-15T09:19:43Z"

4.4.3 查看元数据和对象存储

在 redis 客户端 Pod 里面查看 0 号数据里面已经保存了元数据信息：

 

juicefs-redis-redis-cluster:6379> select 0
OK
juicefs-redis-redis-cluster:6379> keys *
 1) "{0}dirQuotaUsedInodes"
 2) "{0}lastCleanupFiles"
 3) "{0}dirQuota"
 4) "{0}i2"
 5) "{0}setting"
 6) "{0}delSlices"
 7) "{0}sessionInfos"
 8) "{0}nextchunk"
 9) "{0}d1"
10) "{0}c3_0"
11) "{0}totalInodes"
12) "{0}allSessions"
13) "{0}dirDataLength"
14) "{0}d2"
15) "{0}x1"
16) "{0}dirUsedInodes"
17) "{0}i9223372032828243968"
18) "{0}dirQuotaUsedSpace"
19) "{0}nextinode"
20) "{0}usedSpace"
21) "{0}i3"
22) "{0}lastCleanupSessions"
23) "{0}nextsession"
24) "{0}nextCleanupSlices"
25) "{0}dirUsedSpace"
26) "{0}lastCleanupTrash"
27) "{0}i1"

查看存储：

 

root@node1:~# kubectl get pvc -A |grep juicefs-minio
default                        juicefs-minio                              Bound    pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c   50Gi       RWO            local          87m
root@node1:~# 
root@node1:~# cd /var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c/
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c# 
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c# 
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c# ls -l
total 4
drwxr-sr-x 3 liusu liusu 4096 11月 15 17:20 test008
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c# cd test008/
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c/test008# 
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c/test008# ls -l
total 4
drwxr-sr-x 4 liusu liusu 4096 11月 15 17:26 liusu-juicefs
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c/test008# cd liusu-juicefs/
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c/test008/liusu-juicefs# 
root@node1:/var/openebs/local/pvc-c6cd65fa-5191-4dba-84d4-a392399b2c0c/test008/liusu-juicefs# ls -l
total 12
drwxr-sr-x 3 liusu liusu 4096 11月 15 17:20 chunks
-rw-r--r-- 1 liusu liusu   36 11月 15 17:20 juicefs_uuid
drwxr-sr-x 2 liusu liusu 4096 11月 15 17:26 meta

已经将数据分按 Chunk 保存在了 MinIO 中，这个环境中 MinIO使用的是 OpenEBS 类型的 Local PV。并且 MinIO 采用的是单节点部署方式，后续需要在验证 MinIO 为集群版部署的场景。

应用商城

合作伙伴

开发者

支持与服务

了解天翼云

浅析云原生分布式文件系统JuiceFS

1 JuiceFS 技术架构

2 JuiceFS CSI 架构

3 使用方式

3.1 静态配置

3.2 动态配置

4 部署 JuiceFS

4.1 部署 Redis

4.2 部署 MinIO

4.3 部署 JuiceFS CSI

4.4 验证结果

4.4.1 创建 PVC

4.4.2 创建 Pod

4.4.3 查看元数据和对象存储

活动

应用商城

合作伙伴

开发者

支持与服务

了解天翼云

浅析云原生分布式文件系统JuiceFS

1 JuiceFS 技术架构

2 JuiceFS CSI 架构

3 使用方式

3.1 静态配置

3.2 动态配置

4 部署 JuiceFS

4.1 部署 Redis

4.2 部署 MinIO

4.3 部署 JuiceFS CSI

4.4 验证结果

4.4.1 创建 PVC

4.4.2 创建 Pod

4.4.3 查看元数据和对象存储