Increase the size of a PFS

Increase the size of a Parallel File System (PFS) when a training job or dataset outgrows it. The PFS keeps all of its data, and it can stay mounted while you increase it.

Before you begin

You need a role with Update permission on the Dataset service in IAM, and a PFS that reads OK in Parallel File System. Increase PFS Size is available in that status.

Increase the size

1. In Parallel File System, open the PFS’s Actions menu and select Increase PFS Size.
2. Set the new size in whole TB, from 1 TB up to your account’s limit, 250 TB by default. Choose a size larger than the current one. An increase is permanent: the PFS keeps its larger size from then on. (Optional) To go above your limit, ask for a higher one from this screen. The request opens a support ticket with E2E, and you can increase past the old limit once E2E raises it.
3. Tick “I acknowledge that increasing PFS size is irreversible and cannot be downgraded.”
4. Select Resize.

An instance using the PFS picks up the new size while it runs.

The PFS is Resizing until the new size is in place. TIR then emails whoever created the PFS to say whether the increase succeeded or failed. If it failed, the PFS returns to OK at its old size, and TIR opens a support ticket for it. Billing moves to the new size once the increase completes; for current prices, see E2E’s pricing page.

For help when Increase PFS Size is missing or an increase fails, see Troubleshooting TIR Parallel File System.

To see which instances and clusters have the PFS mounted, see See what uses a PFS.