In this brief tutorial you will:
- install ESP prerequisites
- build ESP
- execute ESP unit and integration tests
- run ESP locally with an example Node.js Bookstore backend application
On Mac OS X, install:
On Linux, install:
# Software packages needed for building ESP
sudo apt-get install -y \
g++ git openjdk-8-jdk openjdk-8-source \
pkg-config unzip uuid-dev zip zlib1g-dev
# Software packages needed for building ESP
sudo apt-get install libtool m4 autotools-dev automake
# Software packages needed for running ESP tests
sudo apt-get install libio-socket-ssl-perl
ESP is built using Bazel build tool. Install Bazel version 0.5.4, following the Bazel documentation.
Note: Bazel is under active development and from time to time, ESP continuous integration systems are upgraded to a new version of Bazel. Currently, ESP requires Bazel 0.5.4.
The version of Bazel used by ESP continuous integration systems can be found in
the linux-install-software
script variable BAZEL_VERSION=<SHA>
.
Clone the ESP GitHub repository, initialize Git submodules, and build ESP using Bazel. Detailed instructions for building ESP on Ubuntu 16.04 can be found in the document.
# Clone the ESP repository
git clone https://github.com/cloudendpoints/esp
cd esp
# Initialize Git submodules
git submodule update --init --recursive
# Build ESP binary
bazel build //src/nginx/main:nginx-esp
# Run ESP unit and integration tests
bazel test //src/... //third_party:all
The ESP binary location is:
./bazel-bin/src/nginx/main/nginx-esp
For the remainder of the tutorial we'll use Shell environment variable
ESP_BINARY
to store the location of the ESP binary:
ESP="$(git rev-parse --show-toplevel)"
ESP_BINARY="${ESP}/bazel-bin/src/nginx/main/nginx-esp"
Make sure you have Node.js installed before proceeding. To install Node.js on Linux, you can download the binary distribution and unpack it:
cd /usr/local
sudo tar --strip-components 1 -xzf /path/to/node/tar/file
The example backend application is a simple bookstore service. Start the application using npm:
cd "${ESP}/test/bookstore"
npm install # one-time installation of dependencies
npm test # run bookstore unit tests
# Run the bookstore backend
npm start
# Call the backend (this calls the backend directly)
curl http://localhost:8080/shelves
# Create a directory for Nginx log files
mkdir -p "${TMPDIR}/esp/logs"
# Start ESP
"${ESP_BINARY}" \
-p "${TMPDIR}/esp" \
-c "${ESP}/src/nginx/conf/esp.conf"
Now we are ready to call the the backend via Extensible Service Proxy, using curl:
curl -v http://localhost:8090/shelves
curl -v http://localhost:8090/shelves/1
curl -v http://localhost:8090/shelves/1/books
The calls will fail because ESP is not yet completely configured (we'll do that next) but you will be able to see the calls registered in the log files:
cat "${TMPDIR}/esp/access.log"
cat "${TMPDIR}/esp/error.log"
Shutdown ESP by running:
"${ESP_BINARY}" \
-p "${TMPDIR}/esp" \
-c "${ESP}/src/nginx/conf/esp.conf" \
-s quit
In the previous part of the tutorial we ran ESP locally, without Google Cloud Endpoints integration. In the next section we'll enable integration with Google Cloud Endpoints and use ESP to manage your local Bookstore API.
If you don't have a project, create one in Google Developer Console.
In the ${ESP}/src/nginx/conf/bookstore.json
file, replace the two occurrences
of MY_PROJECT_ID
with the ID of your Google cloud project. These two
occurrences are at the beginning of the bookstore.json
configuration file:
- the
producerProjectId
value - the service
name
value
{
"name": "MY_PROJECT_ID.appspot.com",
"producerProjectId": "MY_PROJECT_ID",
"apis": [
...
With your bookstore.json configuration file updated, deploy the configuration to Google Cloud Endpoints:
gcloud components update
gcloud service-management deploy \
--project=MY_PROJECT_ID \
"${ESP}/src/nginx/conf/bookstore.json"
In order to integrate with Google Cloud Endpoints, ESP needs to be able to send authenticated requests to the Google Cloud Endpoints service control. Service control enables logging, reporting of API related metrics, validation of API keys, etc.
When ESP executes in a Google Compute Engine virtual machine, it will use the virtual machine's service account to authenticate calls to the Endpoints service control. To enable this integration for ESP executing locally, we will give ESP the service account credentials from your project.
You can create a service account for ESP to use with your cloud project in the
Service Accounts
page of the Developer Console. Download the JSON file with the private key
to ${ESP}/src/nginx/conf/service_account.json
and uncomment the
servicecontrol_secret
entry in ${ESP}/src/nginx/conf/esp.conf
:
endpoints {
on;
api bookstore.json;
servicecontrol_secret service_account.json;
}
Once you have deployed the Bookstore API configuration to Google Cloud Endpoints and provided a service account credentials to ESP, you can restart ESP:
# Start ESP
"${ESP_BINARY}" \
-p "${TMPDIR}/esp" \
-c "${ESP}/src/nginx/conf/esp.conf"
First, we call an API method which doesn't require an API key. The method lists all available shelves in the bookstore:
curl -v http://localhost:8090/shelves
The response should include:
{"shelves":[
{"name":"shelves/1","theme":"Fiction"},
{"name":"shelves/2","theme":"Fantasy"}
]}
Next, we call a method which is configured to require an API key; a method which returns books on a bookstore shelf:
curl http://localhost:8090/shelves/1/books
This call fails with a message: Method doesn't allow unregistered callers
.
To authenticate the method call, create a Browser or a Server API key in the
Credentials page of the
Developer Console. Repeat the call with the generated API key:
curl http://localhost:8090/shelves/1/books?key=<YOUR_API_KEY>
This time, the call successfully returns the list of all fiction books:
{ "books": [
{ "name":"shelves/1/books/3",
"author":"Neal Stephenson",
"title":"REAMDE"
}
]}
Open ${ESP}/src/nginx/conf/bookstore.json
and replace the (initially empty)
authentication
section in the global scope with:
"authentication": {
"providers": [
{
"id": "test-auth-provider",
"issuer": "[email protected]",
"jwksUri": "https://www.googleapis.com/service_accounts/v1/jwk/[email protected]"
}
],
"rules": [
{
"requirements": [
{
"audiences": "test-esp-audience",
"providerId": "test-auth-provider"
}
],
"selector": "GetShelf"
}
]
},
This tells ESP that the GetShelf
operation is authenticated and
configures authentication parameters:
- who should be the issuer of the auth tokens (
issuer
), - where to get the public keys of the issuer to validate the signature of
the auth tokens (
jwksUri
), - what audiences the tokens should be issued for (
audiences
).
Reload the nginx configuration:
"${ESP_BINARY}" \
-c "${ESP}/src/nginx/conf/esp.conf" \
-p "${TMPDIR}/esp" \
-s reload
Call the updated endpoint without a token and observe that the request fails with a JWT validation failure:
curl http://localhost:8090/shelves/1
Now generate a JWT auth token using the ${ESP}/client/custom/gen-auth-token.sh
utility and the private key of this particular issuer in the
${ESP}/client/custom/esp-test-client-secret-jwk.json
file:
"${ESP}/client/custom/gen-auth-token.sh" \
-s "${ESP}/client/custom/esp-test-client-secret-jwk.json" \
-a "test-esp-audience"
And finally call the authenticated endpoint with the token.
curl http://localhost:8090/shelves/1?key=<YOUR_API_KEY> \
-H "Authorization: Bearer <GENERATED_TOKEN>"
Visit the Endpoints page, in the Google cloud developer console, select your project and click on the Bookstore API. There you can find API metrics and links to logs that correspond to the API calls you just made using curl.
Congratulations on completing the ESP tutorial! You have successfully:
- built ESP from source
- ran all unit and integration tests locally
- ran ESP locally with a local Bookstore backend
- integrated your local ESP with Endpoins Service Control to enable Endpoints features (logging, monitoring, API Key validation).
To learn more about Endpoints, you can visit our Google Cloud Endpoints documentation which includes more examples and information on using Endpoints with Google Compute Engine, Google Container Engine, and Google App Engine Flex.