Skip to main content
Mount a GPFS file system to access shared data through a local directory on a compute node. You can start from either the file system page or the instance page in the console.

Prerequisites

  • The file system is in the Running state and in the same region as the compute nodes. Cross-region mounting is not supported.
  • The compute nodes have network connectivity to the mount service, and the Cloud Assistant components are running.
  • The local mount path does not conflict with an existing mount. If you mount only a subdirectory of the file system, that subdirectory must already exist.
Before mounting, prepare a mount service, bind the file system, and add target compute nodes as clients. Reuse a suitable service if one exists. Otherwise, create one separately or through the instance mounting wizard. See Manage GPFS mount services.These configurations prepare the access environment. They do not mount the file system to a local directory on the compute nodes.

Mount from the file system page

  1. Log in to the GPU Cloud console.
  2. Go to GPFS > File System.
  3. Select the file system and click Mount.
  4. Select the CES instance and configure the paths.
  5. Check the instance and paths, click Confirm, and wait for mounting to complete.
  6. Find the mount path and log in to the instance to verify that you can access data through the target directory.

Mount from the instance page

  1. Go to Instance and select the target instances on the GPU Instance or CPU Instance tab.
    • For a single instance, select Storage Operations > Mount File System from its more menu in the actions column.
    • For multiple instances, select them and click Mount File System in the bottom actions area.
  2. In the GPFS Mount Service step of the Mount GPFS File System panel, select a mount service for the target instances. Create a service if no suitable one exists.
  3. Click Next and complete the remaining steps in the wizard.
  4. Find the mount path and log in to the instances to verify data access. If any operations failed, record the error details and follow the checks in Troubleshoot mount failures.
Mounting a file system to a CKS cluster node does not make it accessible to Pods automatically. See Mount persistent storage to a workload for workload storage configuration.

Unmount a file system

Unmount a file system when an instance no longer needs access through a mount directory. The following steps use the CES instance page.
Unmounting interrupts data access through the mount path. Stop related read and write tasks, ensure that no processes are using the directory, and check the selected instances and mount path before proceeding.
  1. On the GPU Instance or CPU Instance tab of the Instance page, select one or more instances.
  2. Select Unmount from the more menu in the bottom actions area.
  3. Configure and verify the parameters in the Umount File System panel.
  4. Click Umount and wait for the operation to complete.
  5. Find the mount paths again and verify that the target path is no longer mounted on the selected nodes.
Unmounting a file system, removing a client, and unbinding a file system are separate operations. To also clean up mount service configurations, manage the clients and file system bindings after unmounting.

Troubleshoot mount failures

If mounting fails, check the following: If the issue persists, contact technical support with the file system ID, mount service ID, instance ID, mount path, and error information.