Laravel Reverb now Available

Laravel Reverb now Available

Introduction

Laravel Reverb brings blazing-fast and scalable real-time WebSocket communication directly to your Laravel application and provides seamless integration with Laravel’s existing suite of event broadcasting tools.

Installation

Laravel Reverb requires PHP 8.2+.

You may install Reverb using the install:broadcasting Artisan command:

php artisan install:broadcasting

Configuration

Behind the scenes, the install:broadcasting Artisan command will run the reverb:install command, which will install Reverb with a sensible set of default configuration options. If you would like to make any configuration changes, you may do so by updating Reverb's environment variables or by updating the config/reverb.php configuration file.

Application Credentials

In order to establish a connection to Reverb, a set of Reverb "application" credentials must be exchanged between the client and server. These credentials are configured on the server and are used to verify the request from the client. You may define these credentials using the following environment variables:

REVERB_APP_ID=my-app-id
REVERB_APP_KEY=my-app-key
REVERB_APP_SECRET=my-app-secret

Allowed Origins

You may also define the origins from which client requests may originate by updating the value of the allowed_origins configuration value within the apps section of the config/reverb.php configuration file. Any requests from an origin not listed in your allowed origins will be rejected. You may allow all origins using *:

'apps' => [
    [
        'id' => 'my-app-id',
        'allowed_origins' => ['laravel.com'],
        // ...
    ]
]

Additional Applications

Typically, Reverb provides a WebSocket server for the application in which it is installed. However, it is possible to serve more than one application using a single Reverb installation.

For example, you may wish to maintain a single Laravel application which, via Reverb, provides WebSocket connectivity for multiple applications. This can be achieved by defining multiple apps in your application's config/reverb.php configuration file:

'apps' => [
    [
        'id' => 'my-app-one',
        // ...
    ],
    [
        'id' => 'my-app-two',
        // ...
    ],
],

SSL

In most cases, secure WebSocket connections are handled by the upstream web server (Nginx, etc.) before the request is proxied to your Reverb server.

However, it can sometimes be useful, such as during local development, for the Reverb server to handle secure connections directly. If you are using Laravel Herd's secure site feature or you are using Laravel Valet and have run the secure command against your application, you may use the Herd / Valet certificate generated for your site to secure your Reverb connections. To do so, set the REVERB_HOST environment variable to your site's hostname or explicitly pass the hostname option when starting the Reverb server:

php artisan reverb:start --host="0.0.0.0" --port=8080 --hostname="laravel.test"

Since Herd and Valet domains resolve to localhost, running the command above will result in your Reverb server being accessible via the secure WebSocket protocol (wss) at wss://laravel.test:8080.

You may also manually choose a certificate by defining tls options in your application's config/reverb.php configuration file. Within the array of tls options, you may provide any of the options supported by PHP's SSL context options:

'options' => [
    'tls' => [
        'local_cert' => '/path/to/cert.pem'
    ],
],

Running the Server

The Reverb server can be started using the reverb:start Artisan command:

php artisan reverb:start

By default, the Reverb server will be started at 0.0.0.0:8080, making it accessible from all network interfaces.

If you need to specify a custom host or port, you may do so via the --host and --port options when starting the server:

php artisan reverb:start --host=127.0.0.1 --port=9000

Alternatively, you may define REVERB_SERVER_HOST and REVERB_SERVER_PORT environment variables in your application's .env configuration file.

Debugging

To improve performance, Reverb does not output any debug information by default. If you would like to see the stream of data passing through your Reverb server, you may provide the --debug option to the reverb:start command:

php artisan reverb:start --debug

Restarting

Since Reverb is a long-running process, changes to your code will not be reflected without restarting the server via the reverb:restart Artisan command.

The reverb:restart command ensures all connections are gracefully terminated before stopping the server. If you are running Reverb with a process manager such as Supervisor, the server will be automatically restarted by the process manager after all connections have been terminated:

php artisan reverb:restart

Running Reverb in Production

Due to the long-running nature of WebSocket servers, you may need to make some optimizations to your server and hosting environment to ensure your Reverb server can effectively handle the optimal number of connections for the resources available on your server.

If your site is managed by Laravel Forge, you may automatically optimize your server for Reverb directly from the "Application" panel. By enabling the Reverb integration, Forge will ensure your server is production-ready, including installing any required extensions and increasing the allowed number of connections.

Open Files

Each WebSocket connection is held in memory until either the client or server disconnects. In Unix and Unix-like environments, each connection is represented by a file. However, there are often limits on the number of allowed open files at both the operating system and application level.

Operating System

On a Unix-based operating system, you may determine the allowed number of open files using the ulimit command:

ulimit -n

This command will display the open file limits allowed for different users. You may update these values by editing the /etc/security/limits.conf file. For example, updating the maximum number of open files to 10,000 for the forge user would look like the following:

# /etc/security/limits.conf
forge        soft  nofile  10000
forge        hard  nofile  10000

Event Loop

Under the hood, Reverb uses a ReactPHP event loop to manage WebSocket connections on the server. By default, this event loop is powered by stream_select, which doesn't require any additional extensions. However, stream_select is typically limited to 1,024 open files. As such, if you plan to handle more than 1,000 concurrent connections, you will need to use an alternative event loop not bound to the same restrictions.

Reverb will automatically switch to an ext-event, ext-ev, or ext-uv powered loop when available. All of these PHP extensions are available for install via PECL:

pecl install event
# or
pecl install ev
# or
pecl install uv

Web Server

In most cases, Reverb runs on a non web-facing port on your server. So, in order to route traffic to Reverb, you should configure a reverse proxy. Assuming Reverb is running on host 0.0.0.0 and port 8080 and your server utilizes the Nginx web server, a reverse proxy can be defined for your Reverb server using the following Nginx site configuration:

server {
    ...

    location / {
        proxy_http_version 1.1;
        proxy_set_header Host $http_host;
        proxy_set_header Scheme $scheme;
        proxy_set_header SERVER_PORT $server_port;
        proxy_set_header REMOTE_ADDR $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "Upgrade";

        proxy_pass http://0.0.0.0:8080;
    }

    ...
}

Typically, web servers are configured to limit the number of allowed connections in order to prevent overloading the server. To increase the number of allowed connections on an Nginx web server to 10,000, the worker_rlimit_nofile and worker_connections values of the nginx.conf file should be updated:

user forge;
worker_processes auto;
pid /run/nginx.pid;
include /etc/nginx/modules-enabled/*.conf;
worker_rlimit_nofile 10000;

events {
  worker_connections 10000;
  multi_accept on;
}

The configuration above will allow up to 10,000 Nginx workers per process to be spawned. In addition, this configuration sets Nginx's open file limit to 10,000.

Ports

Unix-based operating systems typically limit the number of ports that can be opened on the server. You may see the currently allowed range via the following command:

cat /proc/sys/net/ipv4/ip_local_port_range
# 32768    60999

The output above shows the server can handle a maximum of 28,231 (60,999 - 32,768) connections since each connection requires a free port. Although we recommend horizontal scaling to increase the number of allowed connections, you may increase the number of available open ports by updating the allowed port range in your server's /etc/sysctl.conf configuration file.

Process Management

In most cases, you should use a process manager such as a Supervisor to ensure the Reverb server is continually running. If you are using Supervisor to run Reverb, you should update the minfds setting of your server's supervisor.conf file to ensure the Supervisor is able to open the files required to handle connections to your Reverb server:

[supervisord]
...
minfds=10000

Scaling

If you need to handle more connections than a single server will allow, you may scale your Reverb server horizontally. Utilizing the publish/subscribe capabilities of Redis, Reverb is able to manage connections across multiple servers. When a message is received by one of your application's Reverb servers, the server will use Redis to publish the incoming message to all other servers.

To enable horizontal scaling, you should set the REVERB_SCALING_ENABLED environment variable to true in your application's .env configuration file:

REVERB_SCALING_ENABLED=true

Next, you should have a dedicated, central Redis server to which all of the Reverb servers will communicate. Reverb will use the default Redis connection configured for your application to publish messages to all of your Reverb servers.

Once you have enabled Reverb's scaling option and configured a Redis server, you may simply invoke the reverb:start command on multiple servers that are able to communicate with your Redis server. These Reverb servers should be placed behind a load balancer that distributes incoming requests evenly among the servers.