Mount datasets and file systems on an instance

Mount a dataset, a Shared File System (SFS), or, in Delhi, a Parallel File System (PFS) on an instance, and its data appears in a folder inside the instance. Mounted storage holds its data in its own product, outside the instance, so the data survives restarts and stops and stays when you delete the instance. Every kind except a disk dataset can be mounted on several instances at once, so several instances can work with the same large data from one place.

Before you begin

• An instance in the project and region where your storage is. To change the datasets on an existing instance, your role in the project must let you update instances.
• The dataset, SFS, or PFS, created and ready. Each tab on the instance’s page has a button that creates one: Create Dataset, Create SFS, or Create PFS. A disk dataset is TIR Block Storage, a dataset in an EOS bucket is TIR Object Storage, and the file systems are the TIR Shared File System and the Parallel File System. Each is billed as its own product. See E2E’s pricing page.
• Your work saved to /home/jovyan. Mounting or unmounting storage on a running instance restarts it, which ends every process in it and clears everything outside the workspace and mounted storage.

Choose what to mount

Each kind of storage suits different work:

Storage	Appears at	Access
Dataset in an EOS bucket	/datasets/<dataset-name>	Read-only, on many instances at once
Disk dataset	/datasets/<dataset-name>	Read-write, on one instance at a time
Shared File System	The mount path you choose	Read-write, on many instances at once
Parallel File System (Delhi)	The mount path you choose	Read-write, or read-only if you tick Read Only; on many instances at once

An EOS dataset suits many instances that only read the same data. A disk dataset gives one instance storage of its own to write to. An SFS lets several instances read and write the same files. In Delhi, choose a PFS for data larger than an SFS holds: a PFS holds 1,000 GB to 250,000 GB, an SFS 250 GB to 64,000 GB. Several instances can mount the same PFS, each read-write or read-only.

Mount storage when you create an instance

When you create an instance, open Storage > Other storage on the Create Instance form. Pick a Shared File System, Datasets, or, in Delhi, a Parallel File System, and give each file system a mount path. A Parallel File System can be mounted Read Only. Selected datasets are mounted at /datasets when the instance starts. See Create an instance for the rest of the form.

Mount or unmount a dataset

You can mount a dataset once it is ready.

1. Open the instance’s page and select the Associated Datasets tab, which has every dataset in the project and region.
2. Tick each dataset to mount, and clear each one to unmount.
3. Click Update, tick Are you sure you want to continue updating datasets?, and click Update again.

Each mounted dataset appears at /datasets/<dataset-name>, where <dataset-name> is the dataset’s name. On a running instance, the instance restarts to apply the change. On a stopped instance, the change applies at the next start.

A disk dataset can be mounted on one instance at a time. To move one to this instance, unmount it from the other instance first. You can tick it here once it is unmounted.

To see what an EOS dataset holds, click its name.

Mount or unmount a Shared File System

Mounting or unmounting an SFS restarts a running instance, and packages you installed yourself need installing again afterwards.

1. Open the instance’s page and select the Shared File-System tab.
2. In the SFS’s row, type a mount path, then click Mount.
3. In Mounting SFS, tick the box to confirm, then select Continue.

To unmount, click Unmount in the SFS’s row, tick the box, and select Continue.

Mount or unmount a Parallel File System

You can mount a PFS in Delhi once its storage status is ready.

As with an SFS, mounting or unmounting a PFS restarts a running instance.

1. Open the instance’s page and select the Parallel File-System tab.
2. In the PFS’s row, type a mount path. To mount it read-only, tick Read Only; leave it clear for read-write.
3. Click Mount, tick the box to confirm, and select Continue.

To unmount, click Unmount in the PFS’s row, tick the box, and select Continue.

Mount path rules

The mount path for an SFS or a PFS must:

• start with / and end with a folder name
• use only letters, digits, hyphens, and underscores in each part, with a single / between parts
• be a path that only this file system uses on the instance

The paths /home/jovyan, /mnt/local, /datasets, and anything under /datasets/ are taken by the instance’s own storage, so pick another one. A path inside the workspace, such as /home/jovyan/shared, is allowed and puts the file system next to your workspace files.

Check the mount

On the tab you used, the dataset or file system is now Mounted. Once the instance is Running again, list the folder from a terminal in JupyterLab or over SSH:

ls /datasets/<dataset-name>

Replace <dataset-name> with the dataset’s name. For an SFS or a PFS, list the mount path you typed instead. If a mount is refused, see A mount or unmount is refused.

To see how mounted storage compares with the workspace and the rest of the container, read Where your files live on an instance.