Space Delivery

Repo management for Nethsecurity installations.

Something went wrong with some packages? Go straight to the faulty packages section. 🚀

Application Structure

The application is structured as follows:

• nginx: Nginx frontend that handles the http requests and proxies them to the php application.
• php: Main application that handles the logic and dispatches the jobs.
• scheduler: Scheduler that handles cron jobs for the php application.
• worker: Worker that handles the jobs dispatched by the php application.

The storage can be configured to use different disks, however a local shared storage is mandatory due to the sqlite database that php, scheduler and worker containers share.

Application Behavior

Base file request:

sequenceDiagram
    client ->> nginx/php: Request file
    database ->> nginx/php: Get repository
    opt repository not found
        nginx/php ->> client: 404 Not Found
    end
    database ->> nginx/php: Check cache
    alt cache miss
        nginx/php ->> my: Ask for authorization
        my ->> nginx/php: 
        alt authorized
            nginx/php ->> database: Save cache
        else not authorized
            nginx/php ->> client: 403 Forbidden
        end
    end
    nginx/php ->> filesystem: Check if file exists
    alt file does not exists
        nginx/php ->> client: 404 Not Found
    else filesystem supports temporary links
        nginx/php ->> client: Redirect to filesystem link 
        client ->> filesystem: Request file
    else manual download
        filesystem ->> nginx/php: 
        nginx/php ->> client: Download file
    end

Loading

Job dispatch sequence:

sequenceDiagram
    participant php
    participant scheduler
    participant database
    
    scheduler ->> database: Dispatch job
    php ->> database: Dispatch job
    database ->> worker: Loads job
    worker ->> worker: Run job
    worker ->> database: Clear job

Loading

Development Setup

Prerequisites

• Docker
• Docker Compose

Podman and Podman Compose can be used as an alternative to Docker and Docker Compose, however no deep testing has been done with these tools.

Environment

Copy the .env.example file to .env and edit the entries as needed.

Most of the environment variables are self-explanatory and there's no need to change their defaults unless explicitly told so. However, there are a few that you might want to change:

• APP_TIMEZONE: The timezone to use for the app. Even if the container inherits the host's timezone, it's recommended to set this value to avoid any issues.

• FILESYSTEM_DISK: Disk to use for I/O ops (cloning repositories and snapshots), defaults to local (storage/app directory). If you want to use a different disk, you need to set the corresponding values for the disk you want to use.

  For example, you can directly connect to a DO Space by filling up the AWS_* values with the corresponding values from the DO Space.

    FILESYSTEM_DISK=s3
    AWS_ACCESS_KEY_ID=your_access_key
    AWS_SECRET_ACCESS_KEY=your_secret_key
    AWS_DEFAULT_REGION=region of the bucket
    AWS_BUCKET=name of the bucket
    AWS_ENDPOINT=https://<region of the bucket>.digitaloceanspaces.com

  Additional docs can be found in Laravel Documentation.

• UID: The user ID for the development environment, set this before running any other command, if this value changes you will need to run the command under Build images again.

• GID: The group ID for the development environment, set this before running any other command, if this value changes you will need to run the command under Build images again.

Build images

To build the development images, you just run the following command:

docker compose build

Setup development environment

Now we just miss a few steps that will need to be run only once:

docker compose run --rm php php artisan key:generate

Running the development server

You're almost there! Run the following command to start up all the needed services:

docker compose up

You can find the app running at http://localhost:8080.

Running commands

To run any commands inside the development environment, you need to get to the shell using:

docker compose exec app bash

Running tests

Software is being tested using PestPHP. To run the tests, you can use the provided command inside the development environment:

php artisan test

Build

Prerequisites

• Docker Bake

Build the production image

GitHub Actions takes care of the deployment of the images to the registry, however if you want to build the production image yourself follow the instructions below.

docker buildx bake -f docker-bake.hcl production

You will find the images tagged as ghcr.io/nethserver/parceler-*:latest.

Production

The production environment is composed by the following services:

• nginx: nginx frontend that handles http requests.
• php: php-fpm that runs a Parceler instance.
• scheduler: scheduler dispatcher for worker.
• worker: worker that handles all jobs sent to queues.

It's advised to use a reverse proxy to handle the SSL termination and load balancing.

Parceler Configuration

The parceler service is being configured through an environment file, you can find the example in .env.prod.example. While some of the values are self-explanatory, there are a few that you need to manually set:

• APP_KEY: The application key, you can generate one using the development environment using php artisan key:generate --show.
• APP_URL: The full URL where the application is reached from, while most of the functionalities will work with a wrong value, the url generation is based off this value.
• FILESYSTEM_DISK: Disk to use during production, works same as development, more info in the development setup.

Container Configuration

Now that parceler is out the way, there's additional configuration needed for the containers to run properly, here's container specific configuration:

nginx needs variables to wait for the php container to be ready before starting, you can set the following:

• FPM_HOST: The host where the php-fpm service is running
• FPM_PORT: The port where the php-fpm service is running

worker and scheduler need the variables to wait for the php container to be ready (and hence prepped the environment):

• PHP_HOST: The host where the php-fpm service is running
• PHP_PORT: The port where the php-fpm service is running

Persistent data

There's a sqlite database being stored in the /var/www/html/storage (for php, worker and scheduler containers) directory when running the service, you need to make sure that this directory is persistent across restarts, otherwise you will lose reference to endpoints and snapshots (or files, if you're using the local disk).

First deploy

To deploy the service, you can find the related files needed in the deploy directory:

• compose: Contains the docker-compose.yml file, this file can deploy the full stack once copied to the server and then using docker compose up -d, remember to copy the .env production file to the same directory where you store the compose file.
• systemd: Uses systemd and podman to deploy the service, you can find the service files in the systemd. Use the v4 or v5 files, depending on the podman version you want to deploy. Remember to copy the .env production file to the %S/parceler.env directory.

Additional Configuration

If you're using rclone to sync the repositories, you can add the configuration file to the container by adding additional environment variables, documentation can be found in the documentation.

Usage

Maintenance mode

To avoid any issues with the files served by the service, if you are operating with the files, you can put the service in maintenance mode, this will prevent any new requests from being processed and will return a 503 status code. Be aware that even crons and queues will stop working, to force queues you can resort to this command. To enable maintenance mode, you can use the following command:

php artisan down

To disable maintenance mode, you can use the following command:

php artisan up

Additional configuration can be provided to the application, such as automatic redirects or have a token that allows the access. Additional configuration can be found in documentation.

Adding a repository

To add a repository, you need to enter to the php container and run the following command:

php artisan repository:create

The command will guide you through the process of adding a repository, here's the fields that will be asked:

• name: name of the repository, will be used to identify the repository under the path repositories/{community|enterprise}/{repository_name},
• command: the command the worker will run to sync the repository it can be anything available in the container. Save the content of the repository under the path source/{repository_name} in the disk you're using. (e.g. if you're using the local disk, save the content of the repository under storage/app/source/repository_name). rclone binary is available in the container, to add configuration file follow the Additional Configuration section.
• source_folder: if repository files are stored in a subfolder, you can specify it here, otherwise leave it empty.
• delay: how many days the upstream files are delayed.

Once the repository is added, a sync job will be created and the worker will start syncing the repository.

List repositories

To list all the repositories, you can use the php artisan repository:list command.

php artisan repository:list

Repo Sync

Repository syncs are dispatched by the scheduler daily. If you want to manually sync a repository, you can use the php artisan repository:sync {repository_name} command.

php artisan repository:sync {repository_name}

Freezing repositories

Freezing a repository will prevent the system to use the normal defer release using the snapshots. This won't halt syncs, so you can freeze the repository to avoid a faulty package, you then unfreeze it to skip faulty packages. To freeze a repository, you can use the php artisan repository:freeze {repository_name} command.

php artisan repository:freeze {repository_name}

Advanced usage can be achieved by providing a custom path like so:

php artisan repository:freeze {repository_name} {path}

Please remember:

• The path is relative to the storage disk.
• The path won't be validated, so make sure it's correct.

Unfreezing repositories

To unfreeze a repository, you can use the php artisan repository:unfreeze {repository_name} command.

php artisan repository:unfreeze {repository_name}

Force release of snapshots

The system will automatically use the oldest snapshot available inside the delay timer given for each repo. However, if you need to push forward the release of some packages you can just delete the older snapshots through the filesystem used. The system will automatically adapt to the new oldest snapshot available.

Listing files in a repository

Since there's a possibility to have a remote disk as a repo, the following command will list the files in the directory that is currently being served. The output is grep friendly.

php artisan repository:files {repository_name}

You can specify the path to list the files in a specific directory.

php artisan repository:files {repository_name} {path}

This is useful when looking for a specific package that is causing issues like this:

php artisan repository:files {repository_name} . | grep {package_name}

Using . as path everything on the storage will be listed, allowing to find for each snapshot the package that has been released. Remember that source/{repository_name} folder is always the latest sync, while snapshots/{repository_name} are the daily syncs.

List repository snapshots

To list all the snapshots of a repository, you can use the php artisan snapshot:list {repository_name} command, you'll be provided the folder that are snapshotted and which one is currently being served.

php artisan repository:snapshots {repository_name}

List of behaviours in case of distribution of faulty packages

The following list is a guide on how to handle the distribution of faulty packages, to find which of the snapshots has a faulty package go to the list repository files section.

• Community repository has a faulty package: Try to fix the issue before the daily sync happens, otherwise follow through in the next steps.

• Faulty package is in the snapshot X: You can safely delete the faulty snapshot, Parceler will use the oldest snapshot available inside the delay timer given for each repo.

• Faulty package is in snapshot X, but community has a fix: You can manually sync the repository to get the latest snapshot (if you don't have it already), then delete all the snapshots with the faulty package, Parceler will have the same behaviour as in the previous step.

• Faulty package is in snapshot X, with no fix available: You can freeze the repository to avoid the faulty package, then when a fix is eventually released you can unfreeze the repository to skip the faulty package.

FAQ

How do I know when the latest sync happened?

You can check the timestamp of the latest snapshot for each repository, otherwise you can check the worker container logs.