Learn how to run your application locally, deploy it, and test on App Engine.
To test your application's functionality before deploying, run your application
in your local environment with the development tools that you usually use.
We recommend that you use standard Python tools, such as
virtualenv to create
isolated environments and
pytest to run unit tests and integration tests.
We do not recommend that you depend on tools like
dev_appserver, the local development
server provided with the Google Cloud SDK. However, if you are
migrating an existing App Engine application from Python 2 to Python 3, we have
dev_appserver to facilitate this process. For all other local development
scenarios, we recommend standard Python testing approaches.
For example, you can usually run a Flask application with Flask's development server using:
Django applications can be started using:
python manage.py runserver
To simulate a production App Engine environment, you can run the full Web Server
Gateway Interface (WSGI) server locally. To do this, use the same command
specified as entrypoint in your
app.yaml, for example:
gunicorn -b :$PORT main:app
Before you begin
Before you can deploy your application:
- The Owner of the GCP project must enable App Engine.
- You must ensure that your user account includes the required privileges.
Deploying your application
Deploy your application to App Engine using the
gcloud app deploy command.
To programmatically deploy your apps, use the Admin API.
Deploying a service
You deploy your application to App Engine by deploying versions of your application's services and each of their configuration files.
To deploy a version of your application's service, run the following command from the directory where the app.yaml file of your service is located:
gcloud app deploy
Specifying no files with the command deploys only the
app.yaml file in your
current directory. By default, the
deploy command generates a unique ID for the
version that you deploy, deploys the version to the GCP project you configured the
gcloud tool to use, and routes all traffic to the new version.
You can change the default behavior of the command by targeting specific files or including additional parameters:
- To deploy the other configuration files of your service, you must target and deploy each file separately. For example:
gcloud app deploy cron.yaml gcloud app deploy dispatch.yaml gcloud app deploy index.yaml
- To specify a custom version ID, use the
- To prevent traffic from being automatically routed to the new version, use the
- To deploy to a specific GCP project, use the
For example, to deploy the service defined by app.yaml to a specific GCP project, assign it a custom version ID, and prevent traffic from being routed to the new version:
gcloud app deploy --project PROJECT_ID --version VERSION_ID --no-promote
For more information about this command, see the
gcloud app deployreference.
Deploying multiple services
You use the same deployment command for deploying or updating the multiple services that make up your application.
To deploy multiple services, you must separately deploy each service's
app.yaml file. For example:
gcloud app deploy service1/app.yaml gcloud app deploy service2/app.yaml
You can specify multiple files with a single deploy command:
gcloud app deploy service1/app.yaml service2/app.yaml
Requirements for deploying multiple services
- You must initially deploy a version of your application to the
defaultservice before you can create and deploy subsequent services.
- The ID of each of your services must be specified in their corresponding
app.yamlconfiguration files. To specify the service ID, include the
serviceelement definition in each configuration file. By default, excluding this element definition from your configuration file deploys the version to the
You can use a
.gcloudignore file to specify files and directories not to
upload to GCP when you deploy your services. This is useful for ignoring build
artifacts and other files that do not need to be uploaded with your deployment.
Managing build images
Each time you deploy a new version, a container image is created using the Cloud Build service. That container image then runs in the App Engine standard environment.
Built container images are stored in the
app-engine folder in Container Registry. You can download these images to keep
or run elsewhere. Once deployment is complete, App Engine no longer needs the
container images. Note that they are not automatically deleted, so to avoid
reaching your storage quota, you can safely delete any images you don't need.
For more information about managing images in Container Registry, see the
Container Registry documentation.
Viewing your application
After you deploy your application to App Engine, you can run the following command
to launch your browser and view it at
gcloud app browse
Testing on App Engine before shifting traffic
Before configuring a new version to receive traffic, you can test it on App
Engine. For example, to test a new version of your
Deploy your new version and include the
gcloud app deploy --no-promote
Access your new version by navigating to the following URL:
Now you can test your new version in the App Engine runtime environment. You can debug your application by viewing its logs. For more information, see Writing Application Logs.
Requests sent to
http://YOUR_PROJECT_ID.appspot.comare be routed to the version previously configured to receive traffic.
When you want to send traffic to the new version, use the GCP Console to migrate traffic:
Select the version you just deployed and click Migrate traffic.
You can use the same process to test new versions of other services by replacing
default in the URL with your service's name:
For more information about targeting specific services and versions, see How Requests are Routed.
Using the local development server
The Google Cloud SDK includes a local development server named
dev_appserver that you can run locally
to simulate your application running in production App Engine. This development
server partially simulates the environment in which your application runs.
Running the local development server
After you create the
configuration file for your app, you can start the local development server with
dev_appserver.py command to run your app locally.
To start the local development server:
dev_appserver.pycommand from the root of your project directory. Specify your project ID and path to your
dev_appserver.py --application=PROJECT_ID app.yaml
To change the port, include the
dev_appserver.py --application=PROJECT_ID app.yaml --port=9999
To learn more about the
dev_appserver.pycommand options, see Local Development Server Options.
As the local development server starts, it sets up a development environment that pre-installs the dependencies found in your
The local development server is now running and listening for requests. Visit http://localhost:8080/ in your web browser to see the app in action.
If you specified a custom port with the
--portoption, remember to open your browser to that port.
To stop the local server from the command line, press
Control-Con your keyboard.
Detecting application runtime environment
To determine whether your code is running in production or in the local
development server, you can check the
GAE_ENV environment variable:
if os.getenv('GAE_ENV', '').startswith('standard'): # Production in the standard environment else: # Local execution.
Using the local development server with Google Cloud services
You can integrate
dev_appserver with other Google Cloud components.
Cloud client libraries
Many Google Cloud client libraries depend on the presence of the
GOOGLE_CLOUD_PROJECT environment variable, which should be your Google Cloud Project ID. You can find its value by issuing the
gcloud config list project command or looking at your Google Cloud Console.
To ensure that this environment variable is set correctly during local development, initialize
dev_appserver using the
--application=PROJECT_ID parameter as shown in the example above.
The local development server automatically installs dependencies found in your
dev_appserver also allows you to test functionality that is configured via
app.yaml. For example, you can test your app's ability to serve static files. When
dev_appserver is running, any changes to
app.yaml automatically restarts your app to reflect these changes. This may result in a temporary delay as dependencies are downloaded and installed.
Instance management and routing in the development server
Discovering instance addresses
The local development server creates all manual scaling instances at startup. Instances for automatic and basic scaling services are managed dynamically. The server assigns a port to each service, and clients can depend on the server to load-balance and select an instance automatically. The port assignments for addressing each service appear in the server's log message stream.
Here are the ports for an app that defines three services (the scaling type of each service is not relevant):
INFO Starting module "default" running at: http://localhost:8084 INFO Starting module "service1" running at: http://localhost:8082 INFO Starting module "service2" running at: http://localhost:8083
When you use a service's address (for example
http://localhost:8082/), the server selects (or creates) an instance of the service and send the request to that instance.
The server assigns unique ports to each instance of a service. You can use the admin server to discover these ports. There is a unique port for the admin server, which appears in the message log:
INFO Starting admin server at: http://localhost:8000
This address takes you to the admin server console. Click on Instances to see the dynamic state of your app's instances
A separate entry appears for each manual and basic instance. The instance numbers are links with unique port addresses for each instance. Click on the link to send a request directly to that instance.
If your app includes a
dispatch.yaml file, the log messages stream includes a dispatcher port:
INFO Starting dispatcher running at: http://localhost:8080
Requests to this port are routed according to the rules in the dispatch file.
The server does not support
dispatch.yaml file rules that include hostnames
url: "customer1.myapp.com/*"). Rules with relative path patterns
url: "*/fun") do work, so you can use URLs like
to reach instances. The server reports an error in the log stream if you try
to start an application with a
dispatch.yaml file that contains host-based rules.