Help & Documentation>Cloud Object Storage

Deploying with Tencent Cloud TKE

Last updated: 2023-09-13 16:23:56

On this page:

To deploy GooseFS using TKE, the open source Fluid component is required, which is available in the TKE application market. You can deploy as follows:
1. Use Fluid helm chart to deploy controllers.
2. Use kubectl to create the dataset and GooseFS runtime.

Considerations

1. Have a Tencent Cloud TKE cluster.
2. Have installed kubectl of v1.18 or a later version.

Installation

1. Navigate to the TKE Application Market, locate the Fluid application, and click to access the application details.
2. Click Create Application, enter a name, select the TKE cluster in the side panel, and in the parameter input box, change the true parameter to false (do not install Alluxio). Keep other configurations as default. Then click Create.
3. Verify the controller components. In the left Cluster, locate the corresponding cluster. If you see two controllers in the cluster's workloads, it indicates that the Fluid component has been installed successfully.

Operation Example

1. Cluster access authorization

[root@master01 run]# export KUBECONFIG=xxx/cls-xxx-config (Download the cluster credential to a directory from the TKE console.)
Note
The cluster API Server needs to have external network access enabled.

2. Create a UFS dataset (using COS as an example)

First, create a secret.yaml file for encryption. The template is as follows:
apiVersion: v1
kind: Secret
metadata:
  name: mysecret
stringData:
  fs.cosn.userinfo.secretKey: xxx
  fs.cosn.userinfo.secretId: xxx
Create secret:
[root@master01 ~]# kubectl apply  -f secret.yaml
secret/mysecret created
dataset.yaml template:
apiVersion: data.fluid.io/v1alpha1
kind: Dataset
metadata:
  name: slice1
spec:
  mounts:
  - mountPoint: cosn://{your bucket}
    name: slice1
    options:
      fs.cosn.bucket.region: ap-beijing
      fs.cosn.impl: org.apache.hadoop.fs.CosFileSystem
      fs.AbstractFileSystem.cosn.impl: org.apache.hadoop.fs.CosN
      fs.cosn.userinfo.appid: "${your appid}"
    encryptOptions:
      - name: fs.cosn.userinfo.secretKey
        valueFrom:
          secretKeyRef:
            name: mysecret
            key: fs.cosn.userinfo.secretKey
      - name: fs.cosn.userinfo.secretId
        valueFrom:
          secretKeyRef:
            name: mysecret
            key: fs.cosn.userinfo.secretId
Create a dataset:
[root@master01 run]# kubectl apply -f dataset.yaml 
dataset.data.fluid.io/slice1 created
Query the dataset status (NotBond):
[root@master01 run]# kubectl get dataset
NAME     UFS TOTAL SIZE   CACHED     CACHE CAPACITY   CACHED PERCENTAGE   PHASE   AGE
slice1 
                                                                 NotBound   11s

3. Creating the runtime

untime.yaml template:
apiVersion: data.fluid.io/v1alpha1
kind: GooseFSRuntime
metadata:
  name: slice1
spec:
  replicas: 1
  data:
    replicas: 1
  goosefsVersion:
    imagePullPolicy: Always
    image: ${img_uri}
    imageTag: ${tag}
  tieredstore:
    levels:
      - mediumtype: MEM
        path: /dev/shm
        quota: 1Gi
        high: "0.95"
        low: "0.7"
  properties:
    # goosefs user
    goosefs.user.file.writetype.default: MUST_CACHE
  master:
    replicas: 1
    journal:
      volumeType: hostpath
    jvmOptions:
      - "-Xmx40G"
      - "-XX:+UnlockExperimentalVMOptions"
      - "-XX:ActiveProcessorCount=8"
  worker:
    jvmOptions:
      - "-Xmx12G"
      - "-XX:+UnlockExperimentalVMOptions"
      - "-XX:MaxDirectMemorySize=32g"
      - "-XX:ActiveProcessorCount=8"
    resources:
      limits:
        cpu: 8
  fuse:
    imagePullPolicy: Always
    image: ${fuse_uri}
    imageTag: ${tag_num}
    env:
      MAX_IDLE_THREADS: "32"
    jvmOptions:
      - "-Xmx16G"
      - "-Xms16G"
      - "-XX:+UseG1GC"
      - "-XX:MaxDirectMemorySize=32g"
      - "-XX:+UnlockExperimentalVMOptions"
      - "-XX:ActiveProcessorCount=24"
    resources:
      limits:
        cpu: 16
    args:
      - fuse
      - --fuse-opts=kernel_cache,ro,max_read=131072,attr_timeout=7200,entry_timeout=7200,nonempty
Create runtime:
[root@master01 run]# kubectl apply -f runtime.yaml 
goosefsruntime.data.fluid.io/slice1 created
Check the GooseFS component status:
[root@master01 run]# kubectl get pods
NAME                               READY   STATUS      RESTARTS   AGE
slice1-fuse-xsvwj     1/1     Running   0          37s
slice1-master-0       2/2     Running   0          118s
slice1-worker-fzpdw   2/2     Running   0          37s

4. Load Data

dataload.yaml data loading component:
apiVersion: data.fluid.io/v1alpha1
kind: DataLoad
metadata:
  name: slice1-dataload
spec:
  dataset:
    name: slice1
    namespace: default
The dataset status is “Bond”, and 0% is cached.
[root@master01 run]# kubectl get dataset
NAME     UFS TOTAL SIZE   CACHED     CACHE CAPACITY   CACHED PERCENTAGE   PHASE   AGE
slice1   97.67MiB         0.00B    4.00GiB          0.0%                Bound   31m
Start loading data:
[root@master01 run]# kubectl apply -f dataload.yaml 
dataload.data.fluid.io/slice1-dataload created
View data loading progress:
[root@master01 run]# kubectl get dataset --watch
NAME     UFS TOTAL SIZE   CACHED     CACHE CAPACITY   CACHED PERCENTAGE   PHASE   AGE
slice1   97.67MiB         52.86MiB   4.00GiB          54.1%               Bound   39m
slice1   97.67MiB         53.36MiB   4.00GiB          54.6%               Bound   39m
slice1   97.67MiB         53.36MiB   4.00GiB          54.6%               Bound   39m
slice1   97.67MiB         53.87MiB   4.00GiB          55.2%               Bound   39m
slice1   97.67MiB         53.87MiB   4.00GiB          55.2%               Bound   39m
100% of the data is cached:
[root@master01 run]# kubectl get dataset --watch
NAME     UFS TOTAL SIZE   CACHED     CACHE CAPACITY   CACHED PERCENTAGE   PHASE   AGE
slice1   97.67MiB         97.67MiB   4.00GiB          100.0%              Bound   44m

5. Inspect Data

[root@master01 run]# kubectl get pods
NAME                               READY   STATUS      RESTARTS   AGE
slice1-dataload-loader-job-km6mg   0/1     Completed   0          12m
slice1-fuse-xsvwj                  1/1     Running     0          17m
slice1-master-0                    2/2     Running     0          19m
slice1-worker-fzpdw                2/2     Running     0          17m
Go to the GooseFS master container:
[root@master01 run]# kubectl exec -it slice1-master-0 -- /bin/bash
Defaulting container name to goosefs-master.
List GooseFS directories:
[root@VM-2-40-tlinux goosefs-1.0.0-SNAPSHOT-noUI-noHelm]# goosefs fs ls /slice1
          10240       PERSISTED 06-25-2021 16:45:11:809 100% /slice1/p1
              1       PERSISTED 05-24-2021 16:07:37:000  DIR /slice1/a
          10000       PERSISTED 05-26-2021 19:29:05:000  DIR /slice1/p2
View a specific file:
[root@VM-2-40-tlinux goosefs-1.0.0-SNAPSHOT-noUI-noHelm]# goosefs fs ls /slice1/a/
             12       PERSISTED 06-25-2021 16:45:11:809 100% /slice1/a/1.xt

中国站 - English