Mounting Google Cloud Storage on Linux with rclone

This article shows how to mount a Google Cloud Storage (GCS) bucket as a folder on a Linux machine with rclone, and how to mount it again automatically after a reboot.

Costs. GCS charges for storage, operations and network egress, and a mount can generate many operations. In Ursa Major, Standard storage is recharged to a lab funding source (Tier 2), while archive classes such as Coldline fall under Tier 1 within limits. See KB005: Ursa Major service tiers and KB007: Tier 2 recharge workflow. Archive classes carry retrieval charges, so they are a poor fit for a mount you read often.

Prerequisites

A Linux machine where you can use FUSE (your workstation, or a VM in your project)
rclone installed (see rclone.org/install)
An existing GCS bucket
An account or service account with access to the bucket

Step 1: Configure an rclone remote

The simplest way is the interactive wizard:

rclone config

Choose n for a new remote and name it gc.
For the storage type, choose Google Cloud Storage (this is not Google Drive) (gcs).
Set credentials in one of these ways:
- On a Google Cloud VM: answer true to env_auth to use the VM's service account.
- With a service account key file: enter its path at service_account_file. Keep the key file private (chmod 600).
- As yourself: leave both blank and complete the browser login rclone starts. On a machine without a browser, rclone prints a command to run on a machine that has one.
Enter your project number if asked (needed only to list or create buckets).
If your bucket uses uniform bucket-level access, answer true to bucket_policy_only.
Accept the defaults for the rest, then save.

The result in ~/.config/rclone/rclone.conf looks similar to this (with a service account key):

[gc]
type = gcs
project_number = 123456789012
service_account_file = /home/yourname/.config/gcloud/my-lab-sa-key.json
bucket_policy_only = true

Check that it works:

rclone lsd gc:
rclone ls gc:my-lab-bucket

Step 2: Create a mount point

mkdir -p ~/gc

Step 3: Mount the bucket

rclone mount gc:my-lab-bucket ~/gc --vfs-cache-mode writes --daemon --log-file ~/gc-rclone.log

--vfs-cache-mode writes lets applications save files normally. --daemon runs the mount in the background and returns you to the prompt; messages go to the log file.

To unmount:

fusermount3 -u ~/gc

(On older systems use fusermount -u.)

Step 4: Mount automatically at startup (optional)

Add an @reboot entry to your crontab. Use the full path to rclone, because cron does not load your usual PATH:

(crontab -l 2>/dev/null; echo "@reboot $(command -v rclone) mount gc:my-lab-bucket $HOME/gc --vfs-cache-mode writes --daemon --log-file $HOME/gc-rclone.log") | crontab -

A systemd user service is a more robust option if you manage the machine yourself.

Things to know

A mounted bucket is slower than a local disk, especially with many small files. For bulk transfers use rclone copy or rclone sync instead.
Objects in GCS cannot be edited in place; changing a file uploads the whole file again.
Google also offers Cloud Storage FUSE (gcsfuse), which is another way to mount a bucket on Linux.

For rclone's full GCS options, see the rclone Google Cloud Storage documentation. Questions: research-computing@ucr.edu.

Owner: Research Computing Reviewed: 4 Oct 2026