gsdk-deploy

This tool provides an automated deployment in your Google Cloud project using a combination of Docker, Kubernetes, Cloud Endpoints, Global Load Balancer, etc. It's intended for high-availablity deployments across Google Cloud regions while providing the simplicity of Google AppEngine deployment.

Configuration:

A configuration file (yaml/json) is required to provide a minimum amount of information needed to deploy your image. Each build type/configuration is represented by a root object. That object describes the parameters required to deploy the service. Environment expansion is available in the following forms: $PLAIN, $(PLAIN), ${PLAIN}; however, this can make it more difficult to run locally. You should prefer specifying the values under one build name and include that in the other builds via the extends entry.

deploy.yaml

# =============================================================================
# = Deployment configuration
# =============================================================================
"dev":
  "name": "$(SERVICE_NAME)-$(BRANCH)"
  "google-project": "$GCLOUD_PROJECT"
  "host": "$(SERVICE_NAME)-$(BRANCH).example.com"
  "api-config": ["./service.pb", "./api_config.yaml"]
  "env":
    - name: NODE_ENV
      value: "$BRANCH"
    - name: "ENDPOINT_NAME"
    - name: "ENDPOINT_VERSION"
  "replicas": 2
  "clusters":
    - "us-central1-a"
"prod":
  "extends": "dev"
  "clusters":
    - "us-central1-a"
    - "us-east4-c"
    - "us-west1-c"
# =============================================================================

Required Configuration Properites:

• google-project - The Google Cloud Project you will be deploying to.
• name - The name of the Kubernetes service to use
• host - The hostname you will direct
• api-config - An array of file names that describes your api (open-api or grpc)
• clusters - An array of Kubernetes cluster names to use. These should either already exist, or simply include the desired regional zone name anywhere in the text. Missing clusters will be created.

Common Configuration Properites:

• extends - Copy all settings from another build configuration.
• port - default = 8080 The port number your container service listens on, defined as NODE_PORT inside the container.
• env - A list of {name,value} pairs to use for the docker image, alternatively specify only a name to copy from current environment.
• replicas - The number of pods you want to run in each region, usually 2 or more.
• nodePort - A port number in the range of 30000-32767 that should be unique to the service. One will be generated from the name if not specified.
• livenessProbe - default = "/_ah/health" The path of your service's health check.
• readinessProbe - default = "/_ah/health?isReady=true" The path of your service's ready check.

There are other options available that can be discovered by the following command:

gsdk-deploy -branch=dev test-config

Prerequisites:

1. First things first, you need a Google account, and a Cloud Project.
2. This tool was written on and for a linux machine, so if you have issues on other environments please feel free submit a pull request to address your specific operating system.
3. You must enable enpoints on your google project at: https://console.cloud.google.com/endpoints
4. Since this service configuration assumes end-to-end SSL, you will need a certifcate that matches the hostname from the configuration file you will provide. You can always generate a self-signed certificate for testing via the following command:

openssl req -x509 -nodes -days 1825 -newkey rsa:2048 \
    -subj /C=US/ST=Delaware/L=Dover/O=ops/OU=dev/CN=example.org \
    -keyout _path_ -out _path_

Usage:

Basic command-line tool invocation:

sudo npm install -g @vroomlabs/gsdk-deploy
gsdk-deploy [command-name] -build=[build-name] (-arg=value ...)

Alternatively you can run as a local dev-package:

npm install @vroomlabs/gsdk-deploy --save-dev
./node_modules/.bin/gsdk-deploy [command-name] -build=[build-name] (-arg=value ...)

Full Deploy Example

gsdk-deploy full-deploy --build=dev --image=user/build

Commands:

• test-config - Light check on configuration
• full-deploy - Normal full deployment, create clusters, networks, etc *optional (user/build) -image=[local docker tag name]
• list-images - Display the list of docker tags from related repo
• kube-images - Display the images running in kube for each cluster
• kube-history - Display kube rollout history for each clusters
• kube-status - Display the current rollout status for each cluster
• kube-rollback - Perform a kube rollback for each cluster
• deploy-api - Deploy only the Google Endpoints API configuration
• deploy-tag - Update to a specific docker image tag name in each cluster *required -tag=[docker tag name from repo]
• deploy-image - Update to a specific docker image uri in each cluster *required -image=[full uri and tag]
• remove-deployment - Removes kubernetes deployment and load balancer

Arguments:

• -branch=[name] - REQUIRED: name of config section from configuration file
• -config=[path] - Relative path to a yaml configuration file
• -image=[name] - default = user/build, specifies a docker image locally or remote
• -auth-env=[name] - The name of an environment variable with base64 json auth
• -auth-file=[path] - Relative path to json authentication file
• -artifacts=[path] - Relative path to an existing directory to store artifacts
• -log=[level] - Specifies log console level (warn,info,debug,verbose,silly)

Arguments may be specified with one or two dash prefixes - following the name and either : or = to separate the value.

Environment:

• Most configuration values can be substituted at runtime via $(NAME)
• GCLOUD_COMMAND, DOCKER_COMMAND, and KUBECTL_COMMAND will control the command-line to these commands. for example: export GCLOUD_COMMAND=sudo /opt/gsdk/gcloud
• AUTO_UPDATE_SDK set to true to auto-upgrade gcloud sdk
• The following variables are defined at runtime:
  • BRANCH - value of the -branch= argument
  • ENDPOINT_NAME - value of the endpoints api name
  • ENDPOINT_VERSION - value of the endpoints api version
  • ARTIFACTS - folder for build artifacts and log files
  • BUILD_TIME - full build timestamp in yyyy-MM-ddThh-mm-ss-sssZ
  • SERVICE_NAME - the name of the service from deploy.yaml
  • GCLOUD_PROJECT - the name of the google project from deploy.yaml
  • HOSTNAME - the host value from the deploy.yaml file
  • NODE_PORT - the configured or generated port for the kube service

CircleCI

For those of you using CircleCI, here is a basic setup:

1. Following the direction here https://circleci.com/docs/1.0/google-auth/ to create the GCLOUD_SERVICE_KEY environment variable.
2. Create and commit a deploy.yaml configuration file in your root directory.
3. Configure your build etc as follows:

dependencies:
  pre:
  - "npm install -g @vroomlabs/gsdk-deploy"
compile:
  override:
  - docker build -t user/build .
deployment:
  develop:
    branch: [develop]
    commands:
    - "gsdk-deploy full-deploy -auth-env:GCLOUD_SERVICE_KEY_DEV -image:user/build -branch:dev"