Introduction
Of course, one of the primary features of Laravel Vapor is atomic, zero-downtime deployments. Unlike many other features of Vapor, deployments are only initiated via the Vapor CLI. During deployment, Vapor will run thebuild
steps of your vapor.yml
file on the local machine that the deployment is running on. This might be your personal machine or a continuous integration platform.
Initiating Deployments
To initiate a deployment, execute thedeploy
CLI command from the root of your application:
vapor deploy production
, Vapor will look for a vapor.yml
manifest file. However, you can specify a custom manifest using the --manifest
option:
AWS Lambda has strict limitations on the size of applications running within the environment. If your application exceeds this limit, you may take advantage of Vapor’s Docker based deployments. Docker based deployments allow you to package and deploy applications up to 10GB in size.
Build Hooks
You may define build hooks for an environment using thebuild
key within your vapor.yml
file. These commands are executed on the local machine that the deployment is running on and may be used to prepare your application for deployment. During deployment, your application is built within a temporary .vapor
directory that is created by the CLI and all of your build
commands will run within that temporary directory:
vapor.yml
Deploy Hooks
You may define deployment hooks for an environment using thedeploy
key within your vapor.yml
file. These commands are executed against the deployed environment before it is activated for general availability. If any of these commands fail, the deployment will not be activated.
vapor.yml
Reviewing Output / Logs
When a deployment hook fails, you may review the output / logs via the Vapor UI’s deployment detail screen. Also, if you are deploying your application using thevapor deploy
command, the CLI output will contain the failing hook output. Of course, you may review the output at any time using the hook:output
command:
hook:log
command:
Native Vapor runtimes do not support squashed migrations. If you wish to utilize this feature of Laravel, you may take advantage of Vapor’s Docker based deployments
Assets
During deployment, Vapor will automatically extract all of the assets in your Laravel project’spublic
directory and upload them to S3. In addition, Vapor will create a AWS CloudFront (CDN) distribution to distribute these assets efficiently around the world.
Because all of your assets will be served via S3 / CloudFront, you should always generate URLs to these assets using Laravel’s asset
helper. Vapor injects an ASSET_URL
environment variable which Laravel’s asset
helper will use when constructing your URLs:
Custom Asset Domains
If you would like to use your own domain to serve your assets, you may do so by attaching a custom asset domain to your project. First, you should ensure a DNS zone exists for the domain and that a certificate is issued for the domain in theus-east-1
region.
Next, set the asset-domain
option of your vapor.yml
file to your chosen domain:
Referencing Assets From JavaScript
If you are referencing your project’s public assets in your JavaScript code, you may generate URLs to these assets using Vapor’s NPM package. This package includes aVapor.asset
helper that will behave like Laravel’s asset
helper but on the client. To get started, install the laravel-vapor
NPM package:
app.js
file, initialize the global Vapor JavaScript object:
Vapor.asset
helper in your JavaScript code:
If you are using Laravel Vite with your project, you only need to utilize the
asset
helper when you are referencing assets you don’t want bundled, such as those that already live in your public directory.If you want to use the asset
helper with your Vite project, you will also need to specify the base URL for assets in your application’s entry point, for example in your resources/js/app.js
, like so: Vapor.withBaseAssetUrl(import.meta.env.VITE_VAPOR_ASSET_URL)
Asset Compilation
If you are compiling your application’s assets in one of the build steps listed in yourvapor.yml
configuration file, you may need your build script to be able to access your environment variables. An excellent example of this requirement is instantiating a frontend SDK such as Pusher.
Build steps are executed in the environment where a deployment is initiated - typically, this will be your local machine or your CI pipeline. As such, you must ensure the required environment variables are available in that environment.
To assist with this, Vapor will attempt to load variables first from .env.[environment]
(e.g. .env.staging
). If that file does not exist, Vapor will attempt to load variables from the .env
file. You should ensure one of these file contains all of the environment variables needed by that environment’s build script.
When using CI platforms, you may not have access to the environment files as these are typically omitted from version control. In this scenario, your CI provider will typically provide a mechanism for injecting variables into the build pipeline. For instance, with GitHub Actions, your GitHub Action configuration might look like the following:
When using Vite and running
npm run build
, Vite will always use the .env.production
file if it is present, even if you are deploying a different environment. If you maintain multiple environment files in your deployment environment, you should set the Vite build mode explicitly:Code Splitting / Dynamic Imports With Mix
If you are taking advantage of JavaScript dynamic imports and code splitting in your project via Laravel Mix, you will need to let Webpack know where the child chunks will be loaded from for each deployment. To accomplish this, you can take advantage of theASSET_URL
variable that Laravel Vapor injects into your environment during your build step:
Hot Module Replacement With Mix
If you are using code splitting and “hot module replacement” with Laravel Mix during local development, you will need to use themix
helper locally and the asset
helper when deploying to Vapor:
URLs Within CSS
Sometimes, your CSS may need to reference asset URLs, such as abackground-image
property that references an image via URL. Obviously, you are not able to use the PHP asset
helper within your CSS. For this reason, Vapor will automatically prepend the correct asset base URL to all relative URLs in your CSS during the build process. After your build steps have executed, Vapor performs this action against any CSS files in your application’s public
directory (including directories nested under public
).
Root Domain Assets
Some applications, such as PWAs, may need to serve certain assets from the root domain. If your application has this need, you can define an array of assets that should be served from your application’s root domain via theserve_assets
configuration option within your vapor
configuration file:
vapor
configuration file, you can publish it using the vendor:publish
Artisan command:
Due to the serverless nature of applications powered by Vapor, assets served from the root domain are not cacheable at the client-side and they are served using Laravel routes. Therefore, you should only serve assets that absolutely must be served from the root domain as there is a slight performance penalty for doing so.
Dot Files As Assets
Typically, “dot” files are not uploaded to the AWS CloudFront CDN by Vapor. However, if you need the public directory’s dot files to be uploaded as assets, you should set thedot-files-as-assets
key to true
in your vapor.yml
file:
Root Domain robots.txt
By default, Vapor will redirect requests to your application’srobots.txt
file to the S3 asset bucket or CloudFront’s asset URL. If you would like the robots.txt
file to be served directly from your application, you may set the redirect_robots_txt
configuration value to false
within your application’s config/vapor.php
configuration file:
vapor
configuration file, you can publish it using the vendor:publish
Artisan command:
Redeploying
Sometimes you may need to simply redeploy a given environment without rebuilding it. For example, you may wish to do this after updating an environment variable. To accomplish this, you may use the Vapor UI or theredeploy
CLI command:
Rollbacks
To rollback to a previous deployment, you may select the deployment in the Vapor UI and click the “Rollback To” button, or you may use therollback
CLI command. The rollback
command’s “—select” option will allow you to select which deployment to rollback to from a list of recent deployments. If the rollback
command is executed without this option, it will simply rollback to the most previous successful deployment:
When rolling back to a previous deployment, Vapor will use the environment’s variables and secrets as they existed at the time the deployment you’re rolling back to was originally deployed.
Deploying From CI
So far, we have discussed deploying Vapor projects from your local command line. However, you may also deploy them from a CI platform of your choice. Since the Vapor CLI client is part of your Composer dependencies, you may simply execute thevapor deploy
command in your CI platform’s deployment pipeline.
In order to authenticate with Vapor from your CI platform, you will need to add a VAPOR_API_TOKEN
environment variable to your CI build environment. You may generate an API token in your Vapor API settings dashboard.
Git Commit Information
Some CI platforms expose the Git commit information as environment variables during your build. You may pass this information to thevapor deploy
command. For example, if using CodeShip:
VAPOR_COMMIT_HASH
.
Example With GitHub Actions
If your application uses GitHub Actions as its CI platform, the following guidelines will assist you in configuring Vapor deployments so that your application is automatically deployed when someone pushes a commit to themaster
branch:
-
First, add the
VAPOR_API_TOKEN
environment variable to your “GitHub > Project Settings > Secrets” settings so that GitHub can authenticate with Vapor while running actions. -
Next, create a
deploy.yml
file within theyour-project/.github/workflows
directory. The file should have the following contents:
- Finally, you can edit the
deploy.yml
file to fit your application’s deployment needs, as it may require a different PHP version or a library likenpm
. Once you are done, commit and push thedeploy.yml
file tomaster
so GitHub Actions can run the first deployment job.
Docker Arm Runtime
Since GitHub actions run on thex86_64
architecture, they are unable to natively compile arm64
based images. In order to compile the docker-arm
runtime, the arm64
platform can be emulated using QEMU. To get started, add the following to your deploy.yml
, directly after the “Checkout Code” step:
Example With Chipper CI
If your application uses Chipper CI as its CI platform, the following guidelines will assist you in configuring Vapor deployments so that your application is automatically deployed when someone pushes a commit to themaster
branch:
-
First, add the
VAPOR_API_TOKEN
environment variable to your “Chipper CI > Project Settings > Project Environment Variables” settings so that Chipper CI can authenticate with Vapor while running your build pipeline. -
Then, on the “Build Pipeline” dashboard, add a step with the name
Deploy
and the following content:
- Next, commit and push any change to the
master
branch so that Chipper CI will deploy your application.
If you plan to use Docker based runtimes to run your Vapor application, you must purchase a Chipper CI paid plan and enable Docker in each project that contains a Docker based Vapor application.