Updated on 2025-01-10 GMT+08:00

Mounting an SFS Turbo File System to Linux ECSs

After creating a file system, you need to mount it to cloud servers so that they can share the file system. This section describes how to mount a file system to ECSs.

  • The operations of mounting a file system to BMSs and containers on CCE are the same as those of ECSs. To mount a file system to CCE, see Storage or Storage (FlexVolume). Then complete the deployment on the CCE console.
  • The mount operation may vary depending on the server OS. Perform the mount operation based on your server OS.

Notes and Constraints

The constraints only apply to local paths (mount points) and do not affect other files or directories.

  • SFS Turbo file systems cannot be mounted to Windows ECSs.
  • The atime, ctime, and mtime attributes of a local path (root directory of the mount point) are the current time of the file system server. Each time a root directory attribute is queried, the current time of the server is returned.
  • Metadata of the local paths (mount points) cannot be modified. Specifically, the following operations cannot be performed on the local paths' metadata:

    - touch: Update file access time and modification time.

    - rm: Delete files or directories.

    - cp: Replicate files or directories.

    - mv: Move files or directories.

    - rename: Rename files or directories.

    - chmod: Modify permissions on files or directories.

    - chown: Change the owners of files or directories.

    - chgrp: Change the group of a file or directory.

    - ln: Create hard links.

    - link: Create hard links.

    - unlink: Delete hard links.

Prerequisites

  • You have checked the type of the OS on each ECS. Different OSs require different commands to install the NFS client.
  • You have created an SFS Turbo file system and obtained its shared path from the file system list.
  • At least an ECS that is in the same VPC as the file system is available.
  • The IP address of the DNS server for resolving the SFS Turbo file system domain name has been configured on the ECSs. For details, see Binding an EIP.

Mounting an SFS Turbo File System

  1. Log in to the ECS console.

    1. Log in to the console and select a region and a project.
    2. Choose Compute > Elastic Cloud Server to go to the ECS console.

  2. Log in to the ECS as user root. You can log in using the console or a remote access tool (such as PuTTY).

  3. Install the NFS client.

    1. Check whether the NFS software package is installed.
      • On CentOS, Red Hat, Oracle Enterprise Linux, SUSE, EulerOS, Fedora, or OpenSUSE, run the following command:
        rpm -qa|grep nfs
      • On Debian or Ubuntu, run the following command:
        dpkg -l nfs-common

      If a command output similar to the following is displayed, the NFS software package has been installed and you can go to 4. If no such command output is displayed, execute the installation command.
      • On CentOS, Red Hat, EulerOS, Fedora, or Oracle Enterprise Linux, the command output is as follows:
        libnfsidmap
        nfs-utils
      • On SUSE or OpenSUSE, the command output is as follows:
        nfsidmap
        nfs-client
      • On Debian or Ubuntu, the command output is as follows:
        nfs-common
    2. Install the NFS software package.

      The following commands require that the ECSs be connected to the Internet. Or, the installation will fail.

      • On CentOS, Red Hat, EulerOS, Fedora, or Oracle Enterprise Linux, run the following command:
        sudo yum -y install nfs-utils
      • On Debian or Ubuntu, run the following command:
        sudo apt-get install nfs-common
      • On SUSE or OpenSUSE, run the following command:
        zypper install nfs-client

  4. If the shared path consists of the file system domain name, run the following command to check whether the domain name can be resolved:

    nslookup <File system domain name>
    • The file system domain name is in the format of xxx.sfsturbo.internal (variable xxx is the file system ID). You can obtain the domain name from the shared path in the file system list.
    • If the nslookup command cannot be used, you can run yum install bind-utils to install the bind-utils software package.

  5. Mount the NFS file system.

    • To mount the file system root directory, run the following commands:
    mkdir <Local path>
    mount -t nfs -o vers=3,timeo=600,noresvport,nolock,tcp <Shared path> <Local path>
    • To mount a file system subdirectory, run the following commands:
    mkdir <Local path>/<Subdirectory>
    mount -t nfs -o vers=3,timeo=600,noresvport,nolock,tcp <File system domain name or IP address>:/<Subdirectory> <Local path>
    • You can obtain the domain name or IP address from the shared path in the file system list.
    • If any other resources, such as a disk, have been mounted on the desired local path, create a new path. (NFS clients do not refuse repeated mounts. If there are repeated mounts, information of the last successful mount is displayed.)
    • A file system can only be mounted to the ECSs that are in the same VPC as the file system.
    Table 1 describes the mount command variables.
    Table 1 Mount command variables

    Parameter

    Description

    <Local path>

    A local directory on the ECS used to mount the file system, for example, /local_path.

    <Shared path>

    For an SFS Turbo Standard-Enhanced, Standard, Performance-Enhanced, or Performance file system, the format is File system IP address:/, for example, 192.168.0.0:/.

    For an SFS Turbo 20 MB/s/TiB, 40 MB/s/TiB, 125 MB/s/TiB, 250 MB/s/TiB, 500 MB/s/TiB, 1,000 MB/s/TiB, or HPC Cache file system, the format is File System domain name:/, for example, xxx.sfsturbo.internal:/.

    Figure 1 shows an example.

    NOTE:
    • Variable x is a digit or letter.
    • If the shared path is too long to display completely, you can adjust the column width.

    vers

    File system version. Only NFSv3 is supported currently, so the value is fixed to 3.

    timeo

    Waiting time before the NFS client retransmits a request. The unit is 0.1 second. The recommended value is 600.

    noresvport

    Whether the NFS client uses a new TCP port when it re-establishes a network connection to the NFS server.

    It is strongly recommended that you specify noresvport, which ensures that your file system remains uninterrupted after a network reconnection or recovery.

    lock/nolock

    Whether to use the NLM protocol to lock files on the server. If nolock is specified, the lock is valid only for applications on the same client. It is invalid for applications on any other clients. nolock is recommended. If this parameter is not specified, lock is used. Then, other clients cannot write data to the file system.

    tcp/udp

    Protocol used by the NFS client to send requests to the server. You can use either UDP or TCP.

    General Purpose File System does not support UDP. Therefore, you need to specify tcp for general purpose file systems.

    Figure 1 Shared Path

    When mounting the file system, you can add performance optimization options described in Table 2. Use commas (,) to separate them. A command example is provided as follows:

    mount -t nfs -o vers=3,timeo=600,nolock,rsize=1048576,wsize=1048576,hard,retrans=3,tcp,noresvport,ro,async,noatime,nodiratime <Shared path> <Local path>
    Table 2 Performance optimization options

    Option

    Description

    rsize

    Maximum number of bytes in each read request that the client can receive when reading data from a file on the server. The actual data size is less than or equal to this parameter setting. The value of rsize must be a positive integral multiple of 1024. Specified values less than 1024 are automatically replaced with 4096, and values greater than 1048576 are automatically replaced with 1048576. By default, this parameter is set through a negotiation between the server and the client.

    You are advised to set this parameter to the maximum value 1048576.

    wsize

    Maximum number of bytes in each write request that the client can send when writing data to a file on the server. The actual data size is less than or equal to this parameter setting. The value of wsize must be a positive integral multiple of 1024. Specified values less than 1024 are automatically replaced with 4096, and values greater than 1048576 are automatically replaced with 1048576. By default, this parameter is set through a negotiation between the server and the client.

    You are advised to set this parameter to the maximum value 1048576.

    soft/hard

    soft indicates soft mounts. With soft specified, if an NFS request times out, the client returns an error to the calling program. hard indicates hard mounts. With hard specified, if an NFS request times out, the client continues to request until the request is successful.

    The default value is hard.

    retrans

    Number of retransmission times before the client returns an error. The recommended value is 1.

    tcp/udp

    If mountproto is not specified, the client will mount the file system using UDP first. If the UDP network cannot be connected, the client will mount the file system using TCP after freezing for several seconds.

    The security group of the file system does not allow inbound traffic over the UDP port, so you need to specify mountproto=tcp.

    ro/rw

    • ro indicates that the file system is mounted as read-only.
    • rw indicates that the file system is mounted as read/write.

    The default value is rw. If neither ro nor rw is specified, the file system will be mounted as read/write.

    noresvport

    Whether the NFS client uses a new TCP port when it re-establishes a network connection to the NFS server.

    It is strongly recommended that you specify noresvport, which ensures that your file system remains uninterrupted after a network reconnection or recovery.

    sync/async

    sync indicates that data is written to the server immediately. async indicates that data is first written to the cache and then to the server.

    async is recommended. Synchronous writes require that an NFS server returns a success message after all data is written to the server, which brings long latency.

    noatime

    If you do not need to record the file access time, set this parameter. This prevents overheads caused by frequent access to modify the time.

    nodiratime

    If you do not need to record the directory access time, set this parameter. This prevents overheads caused by frequent access to modify the time.

    You are advised to use the default values for the parameters with no recommendations provided.

  6. View the mounted file system.

    mount -l
    If the command output contains the following information, the file system has been mounted. You can access the file system to read or write data from the ECS.
    <Shared path> on </local_path> type nfs (rw,vers=3,timeo=600,nolock,addr=)

    The maximum size of a file that can be written to an SFS Turbo file system is 32 TB, and that for an SFS Turbo Enhanced file system is 320 TB.

  1. (Optional) Configure the fstab file to configure file system auto mount upon system startup.

    After a client ECS is restarted, it loses the file system mount information. You can configure auto mount in the fstab file to ensure that the ECS automatically mounts the file system when it restarts.

    1. Open the /etc/fstab file.
    vi /etc/fstab
    At the end of the file, add the file system information, for example:
    <Shared path> </local_path> nfs vers=3,timeo=600,noresvport,nolock,tcp 0 0

    Replace <Shared path> and </local_path> with actual values. You can obtain the shared path on the file system details page. Each record in the /etc/fstab file represents a mount. Each record has six fields, as described in Table 3.

    Table 3 Mount fields

    Field

    Description

    <Shared path>

    The address of the file system to be mounted. Set it to the shared path in the mount command in 5.

    </local_path>

    A directory created on the ECS for mounting the file system. Set it to the local path in the mount command in 5.

    nfs

    The file system or partition mount type. Set it to nfs.

    vers=3,timeo=600,noresvport,nolock,tcp

    Mount options. Use commas (,) to separate multiple options.

    • vers: The file system version. Value 3 indicates NFSv3.
    • timeo: The waiting time before the NFS client retransmits a request. The unit is 0.1 second. The recommended value is 600.
    • noresvport: Whether the NFS client uses a new TCP port when it re-establishes a network connection to the NFS server. It is strongly recommended that you specify noresvport, which ensures that your file system remains uninterrupted after a network reconnection or recovery.
    • nolock: Whether to use the NLM protocol to lock files on the server. If nolock is specified, the lock is valid only for applications on the same client. It is invalid for applications on any other clients. nolock is recommended. If this parameter is not specified, lock is used. Then, other clients cannot write data to the file system.
    • tcp: The TCP transmission protocol.

    0

    Choose whether to use dump to back up the file system.

    • 0: Dump backup is not used.
    • An integer greater than zero means that the file system is backed up. A smaller value has a higher priority.

    0

    Choose whether to use fsck to check file systems when the ECS is starting and specify the sequence for checking file systems.

    • 0: File systems are not checked.
    • By default, this field is set to 1 for the root directory. The values for other directories start from 2, and one with a smaller integer is checked earlier than one with a larger integer.

    For optimal system performance, configure file system information based on the mount example provided. If needed, you can customize certain mount options. Note that the customization may affect system performance.

    1. Press Esc, enter :wq, and press Enter to save and exit.
    1. (Optional) View the updated content of the /etc/fstab file.
    cat /etc/fstab
    Figure 2 Updated file content
    If auto mount fails due to a network issue, add the sleep option and a time in front of the mount command in the rc.local file, and mount the file system after the NFS service is started.
    sleep 10s && sudo mount -t nfs -o vers=3,timeo=600,noresvport,nolock,tcp <Shared path> </local_path>