Deploying Marten projects
This section describes what's involved when it comes to deploying a Marten web application and highlights some important things to consider before performing deploys.
Overview
Each deployment pipeline is unique and will vary from one project to another. That being said, a few things and requirements will be commonly encountered when it comes to deploying a Marten project:
- installing your project's dependencies
- compiling your project's server and management CLI
- collecting your project's assets
- applying any pending migrations to your database
- starting the compiled server
Where, when, and how these steps are performed will vary from one project to another. Each of these steps is highlighted below along with some recommendations.
It should also be noted that a few guides highlighting common deployment strategies can be leveraged if necessary:
Installing dependencies
One of the first things you need to do when deploying a Marten project is to ensure that the dependencies of the projects are available. In this light, you can leverage the shards install command to install your project's Crystal dependencies (assuming that Crystal is installed on your destination machine). Obviously, your project may require the installation of other types of dependencies (such as Node.js dependencies for example), which you have to take care of as well.
Compiling your project
Your project server and management CLI need to be compiled to run your project's server and to execute additional deployment-related management commands (eg. to collect assets or apply migrations).
When it comes to your project server, you will usually need to compile the src/server.cr file (which is automatically created when generating new projects via the new management command). This can be achieved with the following command:
crystal build src/server.cr -o bin/server --release
In the above example, the server binary is compiled using the -o bin/server option, which ensures that the compiled binary will be named server and stored under the related bin folder. You should obviously adapt this to your production environment.
The management CLI is provided by the manage.cr file located at the root of your project. As usual, this file is also automatically generated for you when creating Marten projects through the use of the new management command. Compiling this binary can be done with the following command:
crystal build manage.cr -o bin/manage --release
The above compilation commands make use of the --release flag, which enables compiler optimizations. As a result, the compilation of the final binaries may take quite some time depending on your project. You can also avoid using --release if needed but technically performances could be impacted. See Release builds for more details on this subject.
Collecting assets
You need to ensure that your project and applications assets (eg. JavaScripts, CSS files, etc) are "collected" at deploy time so that they are placed at the final destination from which they will be served: this operation is made available through the use of the collectassets management command. This "destination" depends on your deployment strategy and your configured assets settings: it can be as simple as moving all these assets to a dedicated folder in your server (so that they can be served by your web server), or it can involve uploading these assets to an S3 or GCS bucket for example.
In order to collect assets at deploy time, you will want to use the compiled manage binary and run the collectassets management command (as mentioned previously) with the --no-input flag set in order to disable user prompts:
bin/manage collectassets --no-input
The assets handling documentation also provides a few guidelines on how to serve asset files in production that may be worth reading.
Applying migrations
Your projects will likely make use of models, which means that you will need to ensure that those are properly created at your configured database level by running the associated migrations.
To do so, you can use the compiled manage binary and run the migrate management command:
bin/manage migrate
Please refer to Migrations to learn more about model migrations.
Running the server
You can run the compiled Marten server using the following command (obviously the location of the binary depends on how the compilation was actually performed):
bin/server
It's important to note that the Marten server is intended to be used behind a reverse proxy such as Nginx or Apache: you will usually want to configure such reverse proxy so that it targets your configured Marten server host and port. In this light, you should ensure that your Marten server is not using the HTTP port 80 (instead it could use something like 8080 or 8000 for example).
Depending on your use cases, a reverse proxy will also allow you to easily serve other contents such as assets or uploaded files, and to use SSL/TLS.
It is possible to run multiple processes of the same server behind a reverse proxy such as Nginx. Indeed, each compiled server can accept optional parameters to override the host and/or port being used. These parameters are respectively --bind (or -b) and --port (or -p). For example:
bin/server -b 127.0.0.1
bin/server -p 8080
Additional tips
This section lists a few additional things to consider when deploying Marten projects.
Secure critical setting values
You should pay attention to the value of some of your settings in production environments.
Debug mode
You should ensure that the debug setting is always set to false in production environments. Indeed, the debug mode can help for development purposes because it outputs useful tracebacks and site-related information. But there is a risk that all this information leaks somewhere if you enable this mode in production.
Secret key
You should ensure that the value of the secret_key setting is not hardcoded in your config files. Indeed, this setting value must be kept secret and you should ensure that it's loaded dynamically instead. For example, this setting's value could be set in a dedicated environment variable (or dotenv file) and loaded as follows:
Marten.configure do |config|
config.secret_key = ENV.fetch("MARTEN_SECRET_KEY") { raise "Missing MARTEN_SECRET_KEY env variable" }
# [...]
end