How to Recover from a Partial Terraform Deployment on GCP: A Snowplow Quick Start Guide
Blog post from Snowplow
Deploying Snowplow on Google Cloud Platform (GCP) using Terraform can present challenges, particularly if the Cloud SQL Admin API is not enabled before running the deployment, leading to partial success and subsequent errors. The user encountered a problem when they attempted to re-deploy after enabling the API, resulting in errors due to existing resources being unrecognized by Terraform because of a missing or corrupted state file. Best practices to recover from such issues include restoring or refreshing the state file to sync with existing resources, using the `terraform import` command if the state file is irretrievable, and ensuring required APIs are enabled before deployment. It is also recommended to use a remote backend for state file storage and to preview changes with `terraform plan` before applying them. If problems persist, users may consider manually deleting resources for a clean slate, rebuilding the state through imports, or seeking assistance from Snowplow's professional services. Treating infrastructure as code and maintaining the state file with care are crucial for successful deployments.