1. Jindo 直接加速 OSS 配置环境变量 1 2 3 4 5 6 7 export ENDPOINT=oss-cn-beijing-internal.aliyuncs.com export BUCKET= export AK= export SK= export NAMESPACE=default export PVC=myoss-jindo export URI=datacenter 创建凭证 1 2 3 4 5 6 7 8 9 10 11 kubectl apply -f - <<EOF apiVersion: v1 kind: Secret metadata: name: ${PVC} namespace: ${NAMESPACE} type: Opaque stringData: fs.oss.accessKeyId: ${AK} fs.oss.accessKeySecret: ${SK} EOF 创建 Dataset 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 kubectl apply -f - <<EOF apiVersion: data.fluid.io/v1alpha1 kind: Dataset metadata: name: ${PVC} namespace: ${NAMESPACE} spec: mounts: - mountPoint: oss://${BUCKET}/${URI}/ options: fs.oss.endpoint:

1. 关于 JuiceFS 的缓存 在主机上，预热的缓存是直接放在主机上的。 在集群中，分为两级缓存: Worker，提供集群级别共享的缓存 Fuse，提供仅当前节点级别的缓存 2. 使用 JuiceFS 客户端预热数据 需要注意的是在 Fuse 层预热，仅对当前节点有效，如果需要预热整个集群，需要在

1. 创建 Dataset 1 2 3 4 5 6 7 8 9 10 kubectl apply -f - <<EOF apiVersion: v1 kind: Secret metadata: name: my-s3 type: Opaque stringData: aws.accessKeyId: xxx aws.secretKey: xxx EOF 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 kubectl apply -f - <<EOF apiVersion: data.fluid.io/v1alpha1 kind: Dataset metadata: name: my-s3 spec: mounts: - mountPoint: s3://BUCKET/ name: s3 options: alluxio.underfs.s3.endpoint: ks3-cn-beijing-internal.ksyun.com alluxio.underfs.s3.disable.dns.buckets: "false" encryptOptions: - name: aws.accessKeyId valueFrom: secretKeyRef: name: my-s3 key: aws.accessKeyId - name: aws.secretKey valueFrom: secretKeyRef: name: my-s3 key: aws.secretKey accessModes: - ReadWriteMany EOF 2. 创建 Runtime 1 2 3 4 5 6 7 8 9

1. 分析 Fluid 挂载 NFS 存储 查看 Fuse Pod 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 kubectl get pod nfs-demo-fuse-f9wg8 -oyaml apiVersion: v1 kind: Pod metadata: generateName: nfs-demo-fuse- spec: containers: - command: - /usr/local/bin/entrypoint.sh env: - name: FLUID_RUNTIME_TYPE value: thin - name: FLUID_RUNTIME_NS value: default - name: FLUID_RUNTIME_NAME value: nfs-demo - name: MOUNT_POINT value: /runtime-mnt/thin/default/nfs-demo/thin-fuse - name: MOUNT_OPTIONS value: ro image: fluidcloudnative/nfs:v0.1 imagePullPolicy: IfNotPresent lifecycle: preStop: exec: command: - sh - -c - umount /runtime-mnt/thin/default/nfs-demo/thin-fuse name: thin-fuse securityContext: privileged: true volumeMounts:

1. 打包 Fluid Runtime 镜像 创建 fluid_config_init.py 脚本 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 #!/usr/bin/env python import json rawStr = "" with open("/etc/fluid/config.json", "r") as f: rawStr = f.readlines() rawStr = rawStr[0] script = """ #!/bin/sh set -ex MNT_FROM=$mountPoint MNT_TO=$targetPath trap "umount ${MNT_TO}" SIGTERM mkdir -p ${MNT_TO} mount -t lustre -o relatime,flock ${MNT_FROM} ${MNT_TO} sleep inf """ obj = json.loads(rawStr) with open("mount-lustre.sh", "w") as f: f.write('mountPoint="%s"\n' % obj["mounts"][0]["mountPoint"]) f.write('targetPath="%s"\n' % obj["targetPath"]) f.write(script) 只需调整一下 mount 命令即可。 创建启动脚本 entrypoint.sh 1 2 3