Deploying the API backend

App Engine

With the addition of a small configuration step (described in the following steps), deploying your API so that it is managed by Endpoints is the same as deploying any application to the App Engine flexible environment. Follow the App Engine documentation to:

• Organize your configuration files.
• Create the app.yaml configuration file
• If your application is based on microservices, see Deploying multiple service applications documentation for information on configuring the app.yaml files for each service.

You deploy your API to App Engine by using the gcloud app deploy command. This command automatically builds a container image by using the Container Builder service and then deploys that image to your App Engine flexible environment.

Before you deploy:

• The Owner of the Google Cloud project must create the App Engine application.
• Ensure that your user account includes the required privileges.

Compute Engine

For Endpoints to manage your API, you must install and configure ESP, as well as the backend server code for your API. You need to install Docker on your Compute Engine VM instance so that you can run the ESP Docker image that is freely available in Artifact Registry.

Before you deploy:

The following outlines at a high-level the steps you must take before you can deploy your API and ESP to Compute Engine. In general, you do all the steps that you normally would to run your backend server code on Compute Engine.

1. Create, configure, and start your VM instance. See the Compute Engine documentation.
2. Install Docker Enterprise Edition (EE) or Docker Community Edition (CE) on your VM instance. See Install Docker.
3. Create a Docker container for your backend server code. See Cloud Build documentation.
4. Push the container to Artifact Registry or another registry.
5. Make sure that you can successfully:
   • Connect to the VM instance.
   • Run the Docker image to start your backend server on the VM instance. See Docker run reference.
   • Send requests to your API.

GKE

When you create a cluster in the Google Cloud console, by default, the OAuth scopes that are granted to the cluster's service account include the scopes that Endpoints requires:

• Service Control: Enabled
• Service Management: Read Only

When you create a cluster by using the gcloud container clusters create command or by using a third-party configuration file, make sure you specify the following scopes:

• "https://www.googleapis.com/auth/servicecontrol"
• "https://www.googleapis.com/auth/service.management.readonly"

For more information, see What are access scopes?

Before you deploy:

With the addition of a small section to your Deployment manifest file, you can run the ESP Docker image on your container clusters along with your containerized application. The following outlines at a high-level the steps you must take before you can deploy your API with ESP to GKE. In general, you do all the steps that you normally would to run your backend server code on GKE.

1. Deploy your containerized application to the container clusters. The general steps as described in the GKE documentation are:
   1. Package your app into a Docker image.
   2. Upload the image to a registry.
   3. Create a container cluster.
   4. Deploy your app to the cluster.
   5. Expose your app to the internet.
2. Make sure that you can successfully:
   • Start your API's server.
   • Send requests to your API.

App Engine

To deploy the API and ESP to App Engine:

1. Get the service name of your API. This is the name that you specified in the host field of your OpenAPI document.
2. Edit the app.yaml file, and add a section called endpoints_api_service that contains the service name. You can use the app.yaml file from the tutorial as a model:
   Java

   endpoints_api_service:
     # The following values are to be replaced by information from the output of
     # 'gcloud endpoints services deploy openapi-appengine.yaml' command.
     name: ENDPOINTS-SERVICE-NAME
     rollout_strategy: managed

   Python

   endpoints_api_service:
     # The following values are to be replaced by information from the output of
     # 'gcloud endpoints services deploy openapi-appengine.yaml' command.
     name: ENDPOINTS-SERVICE-NAME
     rollout_strategy: managed

   Go

   endpoints_api_service:
     # The following values are to be replaced by information from the output of
     # 'gcloud endpoints services deploy openapi-appengine.yaml' command.
     name: ENDPOINTS-SERVICE-NAME
     rollout_strategy: managed

   PHP

   endpoints_api_service:
     # The following values are to be replaced by information from the output of
     # 'gcloud endpoints services deploy openapi-appengine.yaml' command. If you have
     # previously run the deploy command, you can list your existing configuration
     # ids using the 'configs list' command as follows:
     #
     #     gcloud endpoints configs list --service=YOUR-PROJECT-ID.appspot.com
     #
     name: ENDPOINTS-SERVICE-NAME
     rollout_strategy: managed

   Ruby

   endpoints_api_service:
     # The following values are to be replaced by information from the output of
     # 'gcloud endpoints services deploy openapi-appengine.yaml' command.
     name: ENDPOINTS-SERVICE-NAME
     rollout_strategy: managed

   NodeJS

   endpoints_api_service:
     # The following values are to be replaced by information from the output of
     # 'gcloud endpoints services deploy openapi-appengine.yaml' command.
     name: ENDPOINTS-SERVICE-NAME
     rollout_strategy: managed

   Replace ENDPOINTS-SERVICE-NAME with your API's service name. For example:

   endpoints_api_service:
     name: example-project-12345.appspot.com
     rollout_strategy: managed
     

   The rollout_strategy: managed option configures ESP to use the latest deployed service configuration. When you specify this option, up to 5 minutes after you deploy a new service configuration, ESP detects the change and automatically begins using it. We recommend that you specify this option instead of a specific configuration ID for ESP to use.

   If your application is based on microservices, you must include the endpoints_api_service section in every app.yaml file.

3. Save the app.yaml file (or files).
4. Deploy your backend code and ESP to App Engine:

   gcloud app deploy

Because you added the endpoints_api_service section to the app.yaml file, the gcloud app deploy command deploys and configures ESP in a separate container to your App Engine flexible environment. All request traffic is routed through ESP, and it proxies requests and responses to and from the container running your backend server code.

If you need to configure ESP to use a specific configuration ID:

1. In the endpoints_api_service section of your app.yaml file, add the config_id field and set it to a specific configuration ID.
2. Either remove rollout_strategy: managed or set rollout_strategy to fixed. The fixed option configures ESP to use the service configuration that you specified in config_id.
3. Redeploy your API and ESP: gcloud app deploy

We recommend that you don't keep ESP configured to use a specific configuration ID for very long because if you deploy an updated service configuration, you have to restart ESP to use the new configuration.

To remove the specific configuration ID:

1. Remove the config_idoption from the app.yamlfile.
2. Add the rollout_strategy: managed option.
3. Issue the gcloud app deploy command

When you use the rollout_strategy: managed option, don't include config_id: YOUR_SERVICE_CONFIG_ID in the app.yaml file. If you do, gcloud app deploy fails with the following error:

config_id is forbidden when rollout_strategy is set to "managed".

When you deploy your API to the App Engine flexible environment for the first time, there might be a delay as your virtual machine (VM) and other infrastructure are set up. For additional information, see Ensuring successful deployment in the App Engine documentation.

Compute Engine

To deploy your API with ESP to Compute Engine with Docker:

1. Connect to your VM instance. Replace INSTANCE_NAME with the name of your VM instance.

   gcloud compute ssh INSTANCE_NAME

2. Create your own container network called esp_net:

   sudo docker network create --driver bridge esp_net

3. Run an instance of the image of your backend server code and connect it to the esp_net container network:

   sudo docker run \
       --detach \
       --name=YOUR_API_CONTAINER_NAME \
       --net=esp_net \
       gcr.io/YOUR_PROJECT_ID/YOUR_IMAGE:1.0

   • Replace YOUR_API_CONTAINER_NAME with the name of your container.
   • Replace YOUR_PROJECT_ID with the Google Cloud project ID that you used when you pushed the image.
   • Replace YOUR_IMAGE with the name of your image.
4. Get the service name of your API. This is the name that you specified in the host field of your OpenAPI document.
5. Run an instance of the ESP Docker image:

   sudo docker run \
       --name=esp \
       --detach \
       --publish=80:8080 \
       --net=esp_net \
       gcr.io/endpoints-release/endpoints-runtime:1 \
       --service=SERVICE_NAME \
       --rollout_strategy=managed \
       --backend=YOUR_API_CONTAINER_NAME:8080

   • Replace SERVICE_NAME with the name of your service.
   • Replace YOUR_API_CONTAINER_NAME with the name of your API's container.

   The --rollout_strategy=managed option configures ESP to use the latest deployed service configuration. When you specify this option, up to 5 minutes after you deploy a new service configuration, ESP detects the change and automatically begins using it. We recommend that you specify this option instead of a specific configuration ID for ESP to use.

If you need to configure ESP to use a specific configuration ID:

1. Include the --version option and set it to a specific configuration ID.
2. Either remove the --rollout_strategy=managed option or set --rollout_strategy to fixed. The fixed option configures ESP to use the service configuration that you specified in --version.
3. Issue the docker run command again.

If you specify both --rollout_strategy=managed and the --version option, ESP starts with the configuration that you specified in --version, but then runs in managed mode and obtains the latest configuration.

We recommend that you don't keep ESP configured to use a specific configuration ID for very long because if you deploy an updated service configuration, you have to restart ESP to use the new configuration.

To remove the specific configuration ID:

1. In the ESP flags for docker run, remove the --version option.
2. Add the --rollout_strategy=managed option.
3. Issue the docker run command to restart ESP.

See ESP startup options for the full list of options that you can specify when starting ESP.

GKE

To deploy ESP to GKE:

1. Get the service name of your API (the name that you specified in the host field of your OpenAPI document).
2. Open your Deployment manifest file (referred to as the deployment.yaml file) and add the following to the containers section:

   containers:
   - name: esp
     image: gcr.io/endpoints-release/endpoints-runtime:1
     args: [
       "--http_port=8081",
       "--backend=127.0.0.1:8080",
       "--service=SERVICE_NAME",
       "--rollout_strategy=managed"
     ]

   Replace SERVICE_NAME with your API's service name.

   The --rollout_strategy=managed" option configures ESP to use the latest deployed service configuration. When you specify this option, up to 5 minutes after you deploy a new service configuration, ESP detects the change and automatically begins using it. We recommend that you specify this option instead of a specific configuration ID for ESP to use.

3. Start the Kubernetes service by using the kubectl createcommand:

   kubectl create -f deployment.yaml

If you need to configure ESP to use a specific configuration ID:

1. In your Deployment manifest file, add the --version option and set it to a specific configuration ID.
2. Either remove --rollout_strategy=managed or set --rollout_strategy to fixed. The fixed option configures ESP to use the service configuration that you specified in --version.
3. Start the Kubernetes service: kubectl create -f deployment.yaml

If you specify both --rollout_strategy=managed and the --version option, ESP starts with the configuration that you specified in --version but then runs in managed mode and obtains the latest configuration.

We recommend that you don't keep ESP configured to use a specific configuration ID for very long because if you deploy an updated service configuration, you have to restart ESP to use the new configuration.

To remove the specific configuration ID:

1. In your Deployment manifest file, remove the --version option.
2. Add the --rollout_strategy=managed.
3. Start the Kubernetes service: kubectl create -f deployment.yaml

See ESP startup options for the full list of options that you can specify when starting ESP.