Argon

Please note: this is not an officially supported Google product.

This middleware automates the import of both Campaign Manager 360 (CM360) and Display & Video 360 (DV360) Offline Reporting files into BigQuery. This allows you to maintain a robust long-term view of your reporting data. It can be deployed to Cloud Functions, Cloud Run, or any other Cloud Provider that supports Docker or Serverless Functions. You can trigger jobs by issuing POST calls with configured JSON, which allows for use with tools like Cloud Scheduler. Argon always checks schemas, uploads all values as string type, and appends a File ID column to track ingestions.

Setup

Google Cloud Project

• Use a Google Cloud project, where you are the Owner.

• Create a BigQuery dataset - tables will be created automatically per report, and appended to for every new report file.

• Create a new IAM Service Account (Eg. [email protected]), and grant these roles:

  • BigQuery Admin (bigquery.admin)
  • Cloud Scheduler Job Runner (cloudscheduler.jobRunner)
  • For Google Cloud Functions:
    • Cloud Functions Invoker (cloudfunctions.invoker)
  • For Google Cloud Run:
    • Cloud Run Invoker (run.invoker)

• Add your own account as a principal for the new Service Account, and grant these roles:

  • Service Account User (iam.serviceAccountUser)
  • Service Account Token Creator (iam.serviceAccountTokenCreator)

• Enable the necessary APIs in API Explorer, or via gcloud services enable :

  • DV: DoubleClick Bid Manager API (doubleclickbidmanager.googleapis.com)
  • CM: DCM/DFA Reporting And Trafficking API (dfareporting.googleapis.com)
  • Cloud Build API (cloudbuild.googleapis.com)
  • For Google Cloud Functions:
    • Cloud Functions API (cloudfunctions.googleapis.com)
  • For Google Cloud Run:
    • Cloud Run Admin API (run.googleapis.com)
    • Artifact Registry API (artifactregistry.googleapis.com)

• Deploy argon to:

  • Google Cloud Functions, using the latest zip release.

  • Google Cloud Run, using the latest prebuilt Docker image - ghcr.io/google/argon:latest.

  • Any Cloud Provider that supports Docker - ghcr.io/google/argon:latest or Serverless Functions.

• Note down your deployed Endpoint's URL.

Google Marketing Platform

Accounts

• Ensure that the CM Account has the following Permissions:

  • Properties > Enable account for API access
  • Reporting > View all generated reports

• Create a CM/ DV User Profile with the Service Account's email address with the respective role:

  • DV: Reporting only
  • CM: Advanced Agency Admin, with permissions:
    • View all generated files
    • View all saved reports

Report

Warning: Argon does not support pre-existing reports, as they can cause hard-to-debug issues. Kindly create a new report as detailed below, and do not change the Dimension/Metrics/Events selections once Argon has started ingesting files. Always create a new Report, if you want to change the report template. All columns are string type, and date-like fields will be transformed to suit BQ date parsing. Argon will also append an additional column (file_id), to keep track of ingested files. If you change the schema in Bigquery, Argon's schema check will fail.

• Choose the necessary report template in "Offline Reporting".

• Choose the "CSV" File type.

• Select the required Dimensions, Metrics, and Rich Media Events.

• Add the service account's email address to the "Share with > +add people", and use the "Link" option.

• If you want historical data to be backfilled initially, select the appropriate backfill Date Range with "Custom".

• If this range is significant, break it up into much smaller chunks, otherwise ingestion timeouts will result in partial uploads.

• Save and run the report, for each chunk, if necessary.

• Now, edit the report again, and select a Date Range of "Yesterday".

• Activate the Schedule for repeats "Daily" every "1 day" and choose a far-off in the future "Expiry" date.

• Save (and do not run) the report.

Google Cloud Scheduler

• Create a Scheduler Job with:

  • Frequency: 0 */12 * * * (repeating every 12 hours)

  • Target type: HTTP

  • URL: Cloud Function URL

  • HTTP Method: POST

  • Auth header: Add OIDC token

  • Service account: Previously created Service Account

  • Audience: Deployed Argon URL

  • Body:

    {
      "product": "[PRODUCT]", // required: CM or DV
      "reportId": [REPORT_ID],
      "profileId": [PROFILE_ID], // required: for CM reports
      "datasetName": "[DATASET_NAME]",
      "projectId": "[BIGQUERY_PROJECT]", // default: current cloud project
      "single": [SINGLE_FILE_MODE], // default: true
      "ignore": [IGNORE_FILE_IDS], // default: []
      "email": "[EMAIL_ADDRESS]" // default: no impersonation
    }

• Notes:

  • Use projectId if the output BigQuery dataset lives outside the currently deployed cloud project.

  • Set single to false, to process more than one file per run. Beware with files that are multiple GBs large, the Cloud Function will timeout after 540s. This will result in partial ingestion or corrupted data.

  • Set ignore to a list of Report File IDs, to skip wrongly generated or unnecessary report files.

  • Set email to a Service Account email address, to impersonate it for local development or testing purposes.

• Save the job and run once to ingest any initially generated historical data files. Alternatively, you can run Argon on your local machine to ingest larger files [how-to].

• If it fails, check the logs for error messages and ensure all the above steps have been appropriately followed, with the correct permissions.

• Moving forward, Cloud Scheduler will trigger Argon for regular ingestion.

• Argon will always attempt to ingest the oldest file that is not present in the BigQuery table and not ignored in your config body.

• Warning: All failed file ingestions will be logged. You will need to manually drop rows with the corresponding File IDs, to force Argon to try and re-ingest them on future runs. Or use ignore to skip them.

Commands

Install the following on your local machine:

• git
• gcloud SDK
• NodeJS & NPM (v16+)

Alternatively, you can use the Google Cloud Shell which comes with all of these tools pre-installed.

# Clone the source code
git clone https://github.com/google/argon.git
cd argon

# Install dependencies
npm install

# Authenticate with GCP
gcloud auth login

# Build from source, outputs to ./dist/
npm run build

Deploy to Google Cloud Platform

Using local source:

# Deploy to Cloud Functions
gcloud functions deploy argon \
  --trigger-http \
  --source ./dist/ \
  --runtime nodejs16 \
  --memory 512M \
  --timeout 540s \
  --service-account "[email protected]"

# Deploy to Cloud Run
gcloud run deploy argon \
    --source ./dist/ \
    --memory 512M \
    --timeout 3600s \
    --service-account "[email protected]"

Using pre-built Docker image:

# Choose your GCP image destination
GCP_CONTAINER_URL="gcr.io/PROJECT-ID/argon:latest"
# OR
GCP_CONTAINER_URL="LOCATION-docker.pkg.dev/PROJECT-ID/argon/argon:latest"

# Pull pre-built image from GitHub Container Registry
docker pull ghcr.io/google/argon:latest

# Tag image locally
docker tag ghcr.io/google/argon:latest $GCP_CONTAINER_URL

# Push image to GCP
docker push $GCP_CONTAINER_URL

# Deploy to Cloud Run
gcloud run deploy argon \
    --image $GCP_CONTAINER_URL
    --memory 512M \
    --timeout 3600s \
    --service-account "[email protected]"

Ingest large report files

# Run a local server, default PORT=8080
npm run watch

# Send a POST from a separate shell terminal
# or using any REST API client
# config.json contains your desired Argon config
curl \
  -H "Content-Type: application/json" \
  --data @config.json \
  localhost:8080

Local Development

# Lint your changes against the guidelines
npm run lint

# Apply formatting rules to your changes
npm run format

Docker

Argon can be containerised using Pack.

# Build & Run a Docker image, from your local source

pack build argon \
  --path ./dist/ \
  --builder gcr.io/buildpacks/builder:v1 \
  --env GOOGLE_FUNCTION_SIGNATURE_TYPE=http \
  --env GOOGLE_FUNCTION_TARGET=argon

# Run a local server, default PORT=8080
docker run --rm -p 8080:8080 argon