gcsfuse is a user-space file system for interacting with Google Cloud Storage.

Current status

Please treat gcsfuse as beta-quality software. Use it for whatever you like, but be aware that bugs may lurk, and that we reserve the right to make small backwards-incompatible changes.

The careful user should be sure to read semantics.md for information on how gcsfuse maps file system operations to GCS operations, and especially on surprising behaviors. The list of open issues may also be of interest.

Installing

See installing.md for full installation instructions for Linux and Mac OS X.

Mounting

Prerequisites

Before invoking gcsfuse, you must have a GCS bucket that you want to mount. If your bucket doesn't yet exist, create one using the Google Developers Console.

Make sure the Google Cloud Storage JSON API is enabled.

GCS credentials are automatically loaded using Google application default credentials, or a JSON key file can be specified explicitly using --key-file. If you haven't already done so, the easiest way to set up your credentials for testing is to run the gcloud tool:

    gcloud auth login

See mounting.md for more information on credentials.

Invoking gcsfuse

To mount a bucket using gcsfuse over an existing directory /path/to/mount, invoke it like this:

gcsfuse my-bucket /path/to/mount

Important: You should run gcsfuse as the user who will be using the file system, not as root. Do not use sudo.

The gcsfuse tool will exit successfully after mounting the file system. Unmount in the usual way for a fuse file system on your operating system:

umount /path/to/mount         # OS X
fusermount -u /path/to/mount  # Linux

If you are mounting a bucket that was populated with objects by some other means besides gcsfuse, you may be interested in the --implicit-dirs flag. See the notes in semantics.md for more information.

See mounting.md for more detail, including notes on running in the foreground and fstab compatiblity.

Performance

Latency and rsync

Writing files to and reading files from GCS has a much higher latency than using a local file system. If you are reading or writing one small file at a time, this may cause you to achieve a low throughput to or from GCS. If you want high throughput, you will need to either use larger files to smooth across latency hiccups or read/write multiple files at a time.

Note in particular that this heavily affects rsync, which reads and writes only one file at a time. You might try using gsutil -m rsync to transfer multiple files to or from your bucket in parallel instead of plain rsync with gcsfuse.

Rate limiting

If you would like to rate limit traffic to/from GCS in order to set limits on your GCS spending on behalf of gcsfuse, you can do so:

The flag --limit-ops-per-sec controls the rate at which gcsfuse will send requests to GCS.
The flag --limit-bytes-per-sec controls the egress bandwidth from gcsfuse to GCS.

All rate limiting is approximate, and is performed over an 8-hour window. By default, requests are limited to 5 per second. There is no limit applied to bandwidth by default.

GCS round trips

By default, gcsfuse uses two forms of caching to save round trips to GCS, at the cost of consistency guarantees. These caching behaviors can be controlled with the flags --stat-cache-capacity, --stat-cache-ttl and --type-cache-ttl. See semantics.md for more information.

Timeouts

If you are using FUSE for macOS, be aware that by default it will give gcsfuse only 60 seconds to respond to each file system operation. This means that if you write and then flush a large file and your upstream bandwidth is insufficient to write it all to GCS within 60 seconds, your gcsfuse file system may become unresponsive. This behavior can be tuned using the daemon_timeout mount option. See issue #196 for details.

Downloading object contents

Behind the scenes, when a newly-opened file is first modified, gcsfuse downloads the entire backing object's contents from GCS. The contents are stored in a local temporary file whose location is controlled by the flag --temp-dir. Later, when the file is closed or fsync'd, gcsfuse writes the contents of the local file back to GCS as a new object generation.

Files that have not been modified are read portion by portion on demand. gcsfuse uses a heuristic to detect when a file is being read sequentially, and will issue fewer, larger read requests to GCS in this case.

The consequence of this is that gcsfuse is relatively efficient when reading or writing entire large files, but will not be particularly fast for small numbers of random writes within larger files, and to a lesser extent the same is true of small random reads. Performance when copying large files into GCS is comparable to gsutil (see issue #22 for testing notes). There is some overhead due to the staging of data in a local temporary file, as discussed above.

Note that new and modified files are also fully staged in the local temporary directory until they are written out to GCS due to being closed or fsync'd. Therefore the user must ensure that there is enough free space available to handle staged content when writing large files.

Other performance issues

If you notice otherwise unreasonable performance, please file an issue.

Support

gcsfuse is open source software, released under the Apache license. It is distributed as-is, without warranties or conditions of any kind.

For support, visit Server Fault. Tag your questions with gcsfuse and google-cloud-platform, and make sure to look at previous questions and answers before asking a new one. For bugs and feature requests, please file an issue.

Versioning

gcsfuse version numbers are assigned according to Semantic Versioning. Note that the current major version is 0, which means that we reserve the right to make backwards-incompatible changes.

vrobot/gcsfuse