Deploying to Google Cloud

It's quite straightforward to push Docker images to Google Container Registry and deploy to GKE clusters with rok8s-scripts.

Google Cloud Credentials

To connect to Google Cloud from a CI workflow, a GCP service account (opens new window) is recommended. To create a service account:

• Go to the service accounts page (opens new window)
• Choose a name for the account (e.g. rok8s-scripts) and hit "Create"
• For "Service account permissions", choose Kubernetes Engine Developer and hit "continue"
• Click "Create Key" and choose "JSON"

To load the JSON credentials into a rok8s-scripts CI workflow, they'll need to be base64 encoded. This can be accomplished with a command like this:

cat downloaded_google_credentials.json | base64 -w 0

With those credentials in base64 format, you'll need to add them as a protected environment variable in your CI tool of choice. This environment variable needs to be named GCLOUD_KEY, and contain a base64 encoded copy of GCP Service Account credentials. It's important that this value is not checked into your codebase, as the credentials could potentially provide a great deal of access into your systems.

Google Cloud Configuration

With the credentials properly configured, there are a few more environment variables that need to be set:

1. GCP_PROJECT The Google Cloud project that you'll be working with.

2. GOOGLE_APPLICATION_CREDENTIALS A path for the decoded $GCLOUD_KEY to be stored. A simple value like /tmp/gcloud_key.json is generally sufficient here.

Optional Google Cloud Configuration

You can optionally specify a registry to use when using the prepare-gcloud script:

1. GCP_IMG_REPO_LOCATION The registry location to pass to the gcloud auth configure-docker command. If omitted the google default is used. See here. (opens new window)

Google Kubernetes Engine Configuration

The configuration above is sufficient to connect to a GCP project and pull or push images, but more configuration is required to connect to a GKE cluster.

1. CLUSTER_NAME The name of the GKE cluster to connect to.

Regional Clusters

1. GCP_REGIONAL_CLUSTER Set this variable to true if the cluster to connect to is regional.

2. GCP_REGION The GCP region the cluster is in.

Zonal Clusters

1. GCP_ZONE The GCP zone the cluster is in.

Connecting to Google Cloud

With the above environment variables in place, it's time to run a script to pull it all together and configure the Google Cloud CLI:

prepare-gcloud