Coordinator¶

The Coordinator is the entry point for the Skewer system, deployed as a stateless Cloud Run service. It serves as the bridge between the user (CLI) and the serverless execution pipeline.

Responsibilities¶

1. Job Validation & Submission¶

The Coordinator receives gRPC requests from the CLI and performs:

Scene Validation - Checks that referenced assets exist in GCS.
Workflow Triggering - Initiates a skewer-render-pipeline execution in Cloud Workflows.
Execution Tracking - Maps user job IDs to internal GCP execution resource names.

2. Pipeline Configuration¶

The Coordinator translates high-level job parameters into the precise environment variables and Batch profiles needed by the pipeline:

Resource Selection - Determines if the job requires Spot or Standard instances based on user flags.
Bucket Mapping - Passes the correct GCS bucket URIs for data and caching to the workflow.

3. Monitoring & Progress¶

Status Aggregation - Queries the Cloud Workflows and Cloud Batch APIs to provide a unified status back to the user.
Metadata Management - Stores job metadata and completion status for future reference.

Architecture¶

User CLI ─▶ Coordinator (Cloud Run) ─▶ Cloud Workflows ─▶ Cloud Batch

Key Components¶

Component	Platform	Description
gRPC Server	Cloud Run	Handles `SubmitJob` and `GetJobStatus` RPCs.
Workflow Client	Google SDK	Triggers and monitors the managed orchestration DAG.
IAM Integration	Google Cloud	Uses Service Account impersonation to securely interact with Batch and Storage.

Deployment¶

The Coordinator is built via Cloud Build and deployed using Terraform.

# Example manual deploy (handled by CI/CD)
gcloud run deploy skewer-coordinator \
  --image gcr.io/PROJECT/skewer-coordinator \
  --env-vars DATA_BUCKET=my-bucket,WORKFLOW_NAME=render-pipeline

Configuration¶

The Coordinator's behavior is primarily controlled via Environment Variables in the Cloud Run service:

Variable	Description
`DATA_BUCKET`	The GCS bucket where assets and renders are stored.
`CACHE_BUCKET`	The GCS bucket used for content-hash layer caching.
`WORKFLOW_NAME`	The full resource ID of the Cloud Workflow to trigger.
`GCP_PROJECT`	The project ID for API calls.