Cloud Parallel File Storage:Mount a file system using a CPFS-NFS client (recommended)

Usage notes

Step 1: Download and install the CPFS-NFS client

This client is open source. For the source code or to compile it manually, visit the GitHub repository.

1. Log on to the ECS instance on which you want to mount the CPFS file system. See Select a method to connect to an ECS instance.

2. Download and install the CPFS-NFS client.

   Alibaba Cloud Linux

   1. Download the CPFS-NFS client.

      wget https://cpfs-hangzhou-nfs-client.oss-cn-hangzhou.aliyuncs.com/aliyun-alinas-utils-latest.al.noarch.rpm

   2. Install the CPFS-NFS client.

      sudo yum install aliyun-alinas-utils-*.rpm

   CentOS

   1. Download the CPFS-NFS client.

      wget https://cpfs-hangzhou-nfs-client.oss-cn-hangzhou.aliyuncs.com/aliyun-alinas-utils-latest.el.noarch.rpm

   2. Install the CPFS-NFS client.

      sudo yum install aliyun-alinas-utils-*.rpm

   Ubuntu

   1. Download the CPFS-NFS client.

      sudo wget https://cpfs-hangzhou-nfs-client.oss-cn-hangzhou.aliyuncs.com/aliyun-alinas-utils-latest.deb -O /tmp/aliyun-alinas-utils-latest.deb

   2. Install the CPFS-NFS client.

      sudo apt-get update
      sudo apt-get install /tmp/aliyun-alinas-utils-latest.deb

   Debian

   1. Download the CPFS-NFS client.

      sudo wget https://cpfs-hangzhou-nfs-client.oss-cn-hangzhou.aliyuncs.com/aliyun-alinas-utils-latest.deb -O /tmp/aliyun-alinas-utils-latest.deb

   2. Install the CPFS-NFS client.

      sudo apt-get update
      sudo apt-get install /tmp/aliyun-alinas-utils-latest.deb

   SUSE

   1. Download the CPFS-NFS client.

      wget https://cpfs-hangzhou-nfs-client.oss-cn-hangzhou.aliyuncs.com/aliyun-alinas-utils-latest.lp15.x86_64.rpm

   2. Install the CPFS-NFS client.

      sudo zypper --no-gpg-checks install -y aliyun-alinas-utils-*.rpm

3. Check whether the installation is successful.

   which mount.cpfs-nfs

   A successful response is shown in the following example:

   /usr/sbin/mount.cpfs-nfs

Step 2: Mount the file system using the CPFS-NFS client

1. Mount the CPFS file system.

   1. Scenario 1: General mount (default path)

      This scenario applies to most environments where AppArmor, SELinux, or special path access restrictions are not enabled.

      sudo mount -t cpfs-nfs -o vers=3,rsize=1048576,wsize=1048576,hard,timeo=600,retrans=2,noresvport file-system-id.region.cpfs.aliyuncs.com:/share/path /mnt

   2. Scenario 2: Mount in a high-security environment (such as systems with SELinux or AppArmor enabled)

      If security hardening is enabled on your operating system through policies such as AppArmor or SELinux, the default mount might cause access exceptions. In this case, specify a path explicitly using hp_config_dir. When you run the command, replace <custom_directory> with a path that your system's security policy allows.

      sudo mount -t cpfs-nfs -o vers=3,rsize=1048576,wsize=1048576,hard,timeo=600,retrans=2,noresvport,hp_config_dir=<custom_directory> file-system-id.region.cpfs.aliyuncs.com:/share/path /mnt

2. Mount command parameters

   Parameter	Description
   file-system-id.region.cpfs.aliyuncs.com:/share/path /mnt	Represents <mount_path> <local_path_to_mount_on_the_server>. Replace them as needed. Mount path: The mount path of the export directory. In the File Storage NAS console, go to the File System list page and click Manage next to the target file system. On the Protocol Service page, click Export Directory in the Actions column. In the Export Directory panel that appears, obtain the mount path. Example: cpfs-196f91a8e58b****-195ceeac7b6ac****.cn-chengdu.cpfs.aliyuncs.com:/share/fileset Local path to mount on the server: The root directory (/) or any subdirectory (for example, /mnt) of the Linux ECS instance. If it is a subdirectory, make sure the subdirectory exists.
   vers	The file system version. CPFS supports mounting file systems only with the NFSv3 protocol.
   Mount options	Separate multiple mount options with commas (,). The options are described as follows: rsize: Defines the size of the data block for reading data between the client and the file system. Recommended value: 1048576. wsize: Defines the size of the data block for writing data between the client and the file system. Recommended value: 1048576. Note To change the I/O size parameters (rsize and wsize), use the maximum value (1048576) whenever possible to avoid performance degradation. hard: If the CPFS file system is temporarily unavailable, local applications that use a file on the file system stop and wait until the file system is back online. Enabling this parameter is recommended. timeo: The time in tenths of a second that the CPFS-NFS client waits for a response before it retries a request to the file system. Recommended value: 600 (60 seconds). Note If you must change the timeout parameter (timeo), use a value of 150 or greater. The unit of the timeo parameter is a tenth of a second, so 150 represents 15 seconds. retrans: The number of times the CPFS-NFS client retries a request. Recommended value: 2. noresvport: Uses a new TCP port on network reconnection to ensure that the connection is not interrupted during network fault recovery. Enabling this parameter is recommended. Important Do not use the soft option. It poses a data consistency threat. If you use the `soft` option, you assume all associated threats. Avoid setting any other mount options that differ from the default values. Changing the read or write buffer sizes or disabling attribute caching can degrade performance. hp_config_dir: (Optional) Explicitly specifies the directory for the haproxy configuration file. Set this parameter if your operating system has Security Enforcement (such as AppArmor or SELinux) or other high-security policies that restrict processes to reading configurations or running only in specific paths. Path selection recommendation: Select a suitable directory from your system's security policy whitelist and ensure the path is authorized by the security module. Logic: The CPFS client first tries to use the directory specified by this parameter. If not specified, it attempts to automatically parse the security policy whitelist. If parsing fails, it falls back to the default path /var/run/cpfs.

3. Run the mount -l command to view the mount result.

   After a successful mount, run the df -h command to view the capacity information of the current file system.

4. (Optional) Configure automatic mount on startup

   To prevent mount information from being lost after the ECS instance restarts, configure the /etc/fstab file. If you used the hp_config_dir parameter during the manual mount, also configure it here.

   1. Open the /etc/fstab configuration file and add the mount configuration.

      • Configuration for a general environment:

        file-system-id.region.nas.aliyuncs.com:/share/path /mnt cpfs-nfs vers=3,rsize=1048576,wsize=1048576,hard,timeo=600,retrans=2,_netdev,noresvport 0 0

      • Example configuration for a high-security environment (using a custom directory):

        Replace <custom_directory> with a path that your system's security policy allows.

        file-system-id.region.nas.aliyuncs.com:/share/path /mnt cpfs-nfs vers=3,rsize=1048576,wsize=1048576,hard,timeo=600,retrans=2,_netdev,noresvport,hp_config_dir=<custom_directory> 0 0

      • Other parameter descriptions

        Parameter	Description
        _netdev	This parameter prevents the client from mounting the file system before the network is ready.
        0 (the first item after noresvport)	A non-zero value indicates that the file system must be backed up by dump. For CPFS file systems, this value is 0 by default.
        0 (the second item after noresvport)	This value indicates the order in which fsck checks the file system at startup. For CPFS file systems, this value is 0 by default, which means fsck does not run at startup.

   2. Run the reboot command to restart the ECS instance.

      Before restarting the ECS instance, confirm that the manual mount was successful to prevent the instance from failing to restart. If the automatic mount is configured successfully, run the df -h command to view the mounted CPFS file system after the ECS instance restarts.

Client configuration

Configuration file

[DEFAULT]
logging_level = INFO
logging_max_bytes = 1048576
logging_file_count = 10

[mount]
stunnel_debug_enabled = false

# Validate the certificate hostname on mount. This option is not supported by certain stunnel versions.
stunnel_check_cert_hostname = false

# Use OCSP to check certificate validity. This option is not supported by certain stunnel versions.
stunnel_check_cert_validity = false
proxy_port = 12049

cpfs_proxy_addr = 127.0.1.255
cpfs_proxy_port_min = 30000
cpfs_proxy_port_max = 60000

[mount-watchdog]
poll_interval_sec = 1
unmount_grace_period_sec = 30
dns_refresh_interval = 60

[client-tool]

haproxy configuration

global
    maxconn 4096

defaults
    mode    tcp
    balance leastconn
    timeout client      60s
    timeout server      60s
    timeout connect     3s
    retries 3

frontend cpfs2049
    bind 127.0.1.255:30000
    default_backend bk2049

backend bk2049
    server cpfs_primary 172.27.1.189:2049 maxconn 2048 check port 2049 inter 2s fall 5 rise 30 on-marked-up shutdown-backup-sessions
    server cpfs_backup  172.27.0.214:2049 maxconn 2048 check port 2049 inter 2s fall 5 rise 30 backup

FAQ

• What do I do if the 'Failed to resolve domain: file-system-id.region.cpfs.aliyuncs.com' error is returned when I mount a CPFS file system?

• What are the differences between a CPFS-POSIX client and a CPFS-NFS client?

• How to remove residual configuration for an uninstalled ECS instance?

• What do I do if the '[FATAL] B: Could not connect to B via ssh' error is returned when I mount a CPFS file system?

References

• CPFS data flow

• Data monitoring

• Access control

• Unmount a file system