Distributed Home File System

Javadocs

What is this?

Inspired by Syncthing and Google Drive File Stream, this project aspires to combine the best of them into one complete solution for all the file synchronization needs: being decentralized like Syncthing and allowing you to stream your files like Google Drive File Stream

Run wrapper

Download latest build

This is a simple set of scripts that allows you to run/stop the DHFS server in the background, and update it.

Once unpacked, in the root folder (dhfs), there will be 3 folders:

• app contains the application
• data contains the filesystem data storage
• fuse is the default filesystem mount point (not on Windows, the default mount drive letter is Z)

Note that on Windows, the path to the root can not contain spaces.

How to use it?

General prerequisites

Java should be available as java in PATH, or with a correctly set JAVA_HOME (ignored on Windows), and Java 21 is required.

FUSE 2 userspace library also should be available:

• On Ubuntu libfuse2 package can be installed, or an analogous package for other distributions.

• On Windows, WinFsp should be installed.

• On macOS, macFUSE.

How to run it?

In the run-wrapper app folder, 3 scripts are available.

• run script starts the filesystem
• stop script stops it
• update script will update the filesystem to the newest available CI build

On Windows, Powershell versions of the scripts should be used. For them to work, it might be required to allow execution of unsigned scripts using set-executionpolicy unrestricted.

Additional options

Additional options for the filesystem can be specified in the extra-opts file in the same directory with the run scripts.

One line in the extra-opts file corresponds to one option passed to the JVM when starting the filesystem.

Some extra possible configuration options are:

• -Ddhfs.fuse.root= specifies the root where filesystem should be mounted. By default, it is the fuse path under the run-wrapper root. For windows, it should be a disk root, by default Z:\.
• -Ddhfs.objects.last-seen.timeout= specifies the period of time (in seconds) after which unavailable peers will be ignored for garbage collection and resynchronized after being reconnected. The default is 43200 (30 days), if set to -1, this feature is disabled.
• -Ddhfs.objects.autosync.download-all= specifies whether all objects (files and their data) should be downloaded to this peer. true or false, the default is false.
• -Ddhfs.objects.peerdiscovery.port= port to broadcast on and listen to for LAN peer discovery (default is 42262)
• -Ddhfs.objects.peerdiscovery.broadcast= whether to enable local peer discovery or not (default is true)
• -Dquarkus.http.port= HTTP port to listen on (default is 8080)
• -Dquarkus.http.ssl-port= HTTPS port to listen on (default is 8443)
• -Dquarkus.http.host= IP address to listen on (default is 0.0.0.0)
• -Ddhfs.peerdiscovery.static-peers= allows to manually specify a peer's address in format of peer id:http port:https port, for example -Ddhfs.peerdiscovery.static-peers=11000000-0000-0000-0000-000000000000:127.0.0.1:9010:9011

On Windows, the entire space for the filesystem should also be preallocated, the -Ddhfs.objects.persistence.lmdb.size= option controls the size (the value is in bytes), on Windows the default is 100 GB.

In case of errors, the standard output is redirected to quarkus.log in the app folder, on Windows the error output is separate.

How to connect to other peers?

Then, a web interface will be available at losthost:8080 (or whatever the HTTP port is), that can be used to connect with other peers. Peers on local network should be available to be connected to automatically.

Other notes

Running tests

To run LazyFS tests, LazyFS needs to be built: the git submodules need to be cloned and ./thirdparty/lazyfs/build.sh script needs to be run.

LazyFS tests were only tested on Linux.

Notice

This software was developed with the support of the Faculty of Information Technology, Czech Technical University in Prague, fit.cvut.cz