Skip to content

TUTORIAL.md

Latest commit

聽

History

History
61 lines (39 loc) 路 2.37 KB

TUTORIAL.md

File metadata and controls

61 lines (39 loc) 路 2.37 KB
Raw

Deploy Manifest on GCP

This walkthrough deploys Manifest on Google Cloud with Cloud Run, Cloud SQL for PostgreSQL, a private Cloud Storage bucket mounted for request recordings, and Secret Manager. DeployStack asks for a few settings, writes terraform.tfvars, and runs Terraform in the selected project.

Prerequisites

Pick the Google Cloud project you want to deploy into and make sure billing is enabled. This stack creates paid resources, including Cloud SQL.

Enable required APIs

gcloud services enable \
  iam.googleapis.com \
  run.googleapis.com \
  sqladmin.googleapis.com \
  secretmanager.googleapis.com \
  storage.googleapis.com

Run the installer

DeployStack reads <walkthrough-editor-open-file filePath="deploy/gcp/deploystack.json">deploystack.json</walkthrough-editor-open-file>, prompts for project, region, service name, image, and sizing, then runs Terraform.

deploystack install

The first deploy usually takes 10-15 minutes because Cloud SQL needs time to provision. The Terraform config also patches BETTER_AUTH_URL to the final Cloud Run URL after the service is created. Request recordings are written through a Cloud Storage volume mount at /data/request-recordings; the Cloud Run service account receives object read/write/delete access only for that bucket.

Open Manifest

terraform output service_url
terraform output recording_bucket

Open the URL and create the first account. The first account becomes the admin.

To verify the deployment:

curl -sSf "$(terraform output -raw health_check_url)"

Tearing it down

deploystack uninstall

Cloud SQL deletion protection is enabled by default. Set database_deletion_protection=false and apply before uninstalling. The recording bucket also protects stored objects from accidental deletion: export or delete its contents before uninstalling, or keep the bucket and remove it from Terraform state if the recordings must survive the rest of the stack.

You're done