Skip to main content
A GPFS mount service configures the client environment that compute nodes need to access a file system. You can create a service, bind a file system, and add clients separately, or complete these tasks in the instance mounting wizard.
The operation durations on this page are estimates. Check the corresponding asynchronous task for actual progress.

​
Prerequisites

  • Identify the compute nodes that need GPFS access and their region.
  • Ensure that the compute nodes can reach the mount service over the network and that the Cloud Assistant components are running.
  • Before binding a file system, ensure that it is in the Running state and in the same region as the compute nodes.

​
Create a mount service

  1. Log in to the GPU Cloud console.
  2. Go to GPFS > Mount Service and click Create Mount Service.
  3. Configure the mount service.
  4. Click Confirm and wait for the client environment to be configured. Creating a mount service takes approximately 10 minutes.
  5. Check the service status in the mount service list. Click the service name to view its clients, file systems, and mount counts.

​
Bind a file system

  1. Find the target service in the mount service list and select Bind File System from its more menu.
  2. Check the mount service information and select the file system in Volume.
  3. Click Confirm and wait for the binding to complete. Binding takes approximately five minutes. Verify that the file system is associated with the service.

​
Manage clients

A client is a compute node that accesses GPFS through a mount service. Adding a client prepares its access environment. To access data through a local directory, you must also mount the GPFS file system.

​
Add clients

  1. Click Client Management in the target mount service’s actions column.
  2. Click Add Client.
  3. Check the mount service and VPC information, then select one or more nodes in Client Nodes. We recommend adding 20 to 50 nodes per request. The system processes nodes in batches of 20, with each batch taking approximately 20 minutes.
  4. Click Confirm and wait for the GPFS client environment to be configured on the nodes. Check their status in the client list.

​
Remove clients

Removing a client affects the node’s access to GPFS through the mount service. Stop related read and write tasks and unmount the associated file systems from the node before proceeding.
  1. Select the clients you no longer need on the client management page.
  2. Click Remove Client, verify the selected nodes, and confirm as prompted. The system processes nodes in batches of 20, with each batch taking approximately five minutes.
  3. Verify that the nodes no longer appear in the client list.

​
Unbind a file system

Unbind a file system when you no longer need to access it through a particular mount service.
Stop tasks that access the file system through the mount service and unmount it from the related compute nodes before unbinding it.
  1. Return to the mount service list and select Unbind File System from the target service’s more menu.
  2. Check the file system binding and confirm as prompted.
  3. Wait for the operation to complete and verify that the file system is no longer associated with the mount service. Unbinding takes approximately five minutes.

​
Delete a mount service

Deleting a mount service affects GPFS access for compute nodes that depend on it. Ensure that the service is no longer in use, unmount the associated file systems, and check its clients and file system bindings before proceeding.
  1. Find the service you no longer need in the mount service list and select Delete from its more menu.
  2. Check the target service and confirm as prompted.
  3. Verify that the service no longer appears in the list.

​
View asynchronous tasks

Use asynchronous tasks to check the progress and results of mount service operations.
  1. Go to GPFS > Async Task and locate the operation by its description and time.
  2. Check the task status. Click the task ID to view the status, start and end times, and errors for each execution step.
  3. If the task fails or partially fails, check the failed steps and affected nodes. Identify which nodes have completed the operation and which still require attention.
If you cannot resolve the issue, contact technical support with the task ID, mount service ID, related resource IDs, and error information.