This topic describes how to access JindoFS from a Filesystem in Userspace (FUSE) client. FUSE allows you to use JindoFS in block storage and cache mode (only JFS Scheme supported).

Prerequisites

An E-MapReduce (EMR) cluster is created. For more information, see Create a cluster.

Background information

FUSE is a method provided by the Linux kernel to mount a file system. You can use a FUSE client to map files in a JindoFS cluster to a local disk. Then, you can directly access data in the JindoFS cluster without the need to use the hadoop fs -ls jfs://<namespace>/ command.

Mount a directory

Note Perform the following steps on each node of the cluster in sequence.
  1. Log on to the master node of the cluster in SSH mode. For more information, see Connect to the master node of an EMR cluster in SSH mode.
  2. Run the following command to create a directory:
    mkdir /mnt/jfs
  3. Run the following command to mount the directory:
    jindofs-fuse /mnt/jfs

    In this example, the /mnt/jfs/ directory is mounted.

Read and write files

  1. Run the following command to query all the sub-directories under the /mnt/jfs/ directory:
    ls /mnt/jfs/
    All the namespaces that you have configured on the server are returned.
    test  testcache
  2. Run the following command to query all the files in the test namespace:
    ls /mnt/jfs/test/
  3. Run the following command to create a directory:
    mkdir /mnt/jfs/test/dir1
    ls /mnt/jfs/test/
  4. Run the following command to write data into a file:
    echo "hello world" > /tmp/hello.txt
    cp /tmp/hello.txt /mnt/jfs/test/dir1/
  5. Run the following command to read data from a file:
    cat /mnt/jfs/test/dir1/hello.txt
    The following information is returned:
    hello world
The following example shows how to read data from and write data into files in Python:
  1. Write data into a file in Python. Create the write.py script file, which contains the following content:
    #! /usr/bin/env python36
    with open("/mnt/jfs/test/test.txt",'w',encoding = 'utf-8') as f:
       f.write("my first file\n")
       f.write("This file\n\n")
       f.write("contains three lines\n")
  2. Read data from a file in Python. Create the read.py script file, which contains the following content:
    #! /usr/bin/env python36
    with open("/mnt/jfs/test/test.txt",'r',encoding = 'utf-8') as f:
        lines = f.readlines()
        [print(x, end = '') for x in lines]
    Run the following command to read the data that is written into the test.txt file:
    [hadoop@emr-header-1 ~]$ ./read.py
    The following information is returned:
    my first file
    This file

Unmount FUSE

Note Perform the following steps on each node of the cluster in sequence.
  1. Log on to the master node of the cluster in SSH mode. For more information, see Connect to the master node of an EMR cluster in SSH mode.
  2. Run the following command to unmount FUSE:
    umount jindofs-fuse

    If the target is busy error is returned, switch to another directory, stop all the programs that are reading data from and writing data into FUSE files, and then unmount FUSE.