How Fleet works

Fleet stores a snapshot of Docker Images in its own database, consisting of metadata deemed most pertinent to both the users of the images, and the repository owner. It will synchronise with Docker Hub over a set interval in order to update its stored data.

It then displays this snapshot data on its own status page as a useful list, containing links to each repository and image owned by the repository owner. Each image also contains a status which is managed by the repository owner, who can define images as either Stable or Unstable. This is designed to quickly help users know when an image is undergoing a state of instability which is known by the owner.

Why a snapshot?

In short, Docker Hub's API is very slow. It would not be a good long-term solution to just proxy the responses from Docker Hub and translate the data into something considered useful by downstream clients. By caching the image information in its own database, Fleet is able to more efficiently return the status data for each image and repository. In doing so, it is also able to provide more concise data, such as image versions, as part of the primary response, rather than requiring users to make a separate call.

As an example comparison between obtaining all image name, pull and version information for all LinuxServer images from Docker Hub, and then obtaining that same data via Fleet's API:

Capabilities

Fleet has the ability to display images with a particular state, which provides contextual information to visitors of the application's main page.

Hidden

If an image is hidden, it will not be displayed as part of the main list, nor will it be returned as part of any API calls. This also means that the pull count of a hidden image is not included.

Unstable

Marks an image as having issues known by the maintainer. A useful state to assign to an image if the latest build (or builds) are causing downstream breakages. This may also be useful if an upstream dependency or application is causing breakages in the image directly.

Deprecated

If the maintainer of the image, or upstream application no longer wishes to provide support, or if the image has reached its end-of-life (or has been superseded by another), marking an image as deprecated will ensure users are made aware that no further updates will be supplied, and should stop using it. Deprecation notices are also provided to give context.

API

Fleet exposes a single API endpoint which can be used to obtain image list and pull count information for all relevant images maintained by the repository

{
    "status": "OK",
    "data" {
        "totalPullCount": 1862494227,
        "repositories": {
            "lsiobase": [
                {
                    "name": "alpine",
                    "pullCount": 4275970,
                    "version": "3.6",
                    "stable": true
                },
                {
                    "name": "alpine.arm64",
                    "pullCount": 66234,
                    "version": "edge",
                    "stable": true
                },
                ...
            ],
            "linuxserver": [
                {
                    "name": "airsonic",
                    "pullCount": 4608329,
                    "version": "v10.2.1",
                    "stable": true
                },
                {
                    "name": "apache",
                    "pullCount": 3011699,
                    "version": "latest",
                    "stable": true
                },
                ...
            ]
            ...
        }
    }
}

Running Fleet

Grab the latest Fleet release from GitHub.

SQL

Fleet stores its data in a MariaDB database which you need to provide. In order for the application to manage its tables and procedures, the user you provide it needs to have the relevant GRANT permissions to the fleet database. The following script should be sufficient to get the initial database set up.

CREATE SCHEMA `fleet`;
CREATE USER 'fleet_user' IDENTIFIED BY 'supersecretpassword';
GRANT ALL ON `fleet`.* TO 'fleet_user';

The username and password that you define must then be provided as part of Fleet's configuration.

Configuration File

All primary configuration for Fleet at runtime is loaded in via a fleet.properties file. This can be located anywhere on the file system, and is loaded in via a Runtime argument:

# Runtime
fleet.app.port=8080
fleet.refresh.interval=60
​
# If set to DATABASE, fleet.admin.username and fleet.admin.password are not used.
fleet.admin.authentication.type=PROPERTIES|DATABASE
fleet.admin.secret=<your_secret_string>
​
# User for management of images and repositories
fleet.admin.username=test
fleet.admin.password=test
​
# Database Connectivity
fleet.database.driver=org.mariadb.jdbc.Driver
fleet.database.url=jdbc:mariadb://<IP_OR_URL>:3306/fleet
fleet.database.username=<fleet_sql_user>
fleet.database.password=<fleet_sql_password>
​
# Docker Hub
fleet.dockerhub.username=<username_for_your_dockerhub_account>
fleet.dockerhub.password=<password_for_your_dockerhub_account>

All configuration can be loaded either via the config file, via JVM arguments, or via the system environment. Fleet will first look in the configuration file, then JVM runtime, and finally in the system environment. It will load the first value it finds, which can be useful when needing to override specific properties.

As well as the base configuration file, Fleet also supports some runtime arguments by means of the -D flag. These can be used to direct Fleet to behave in a specific way at runtime.