FOCUS dataset

FOCUS is a Google Cloud feature that exports the project’s cloud cost data to a BigQuery table. GENI Workflows reads that table to attribute cost to submissions, tasks and the rest.

Google’s setup guide: Set up FOCUS billing export to BigQuery.

What you need afterwards

Two values, both passed when creating the environment:

Value	Example
Table identifier, project.dataset.table	useful-loop-496215-i8.gcp_billing_immutable_01F464_AC8B27_889258_us.gcp_billing_export_focus_01F464_AC8B27_889258
Table location	us

All three parts are required

The table identifier is project.dataset.table. The dataset alone (…immutable_01F464_AC8B27_889258_us) looks complete but is not — a two-part value is rejected, and the export cannot be read without the table name. The section below reads all three parts straight from BigQuery.

Note

The FOCUS table is not required to create the environment. Without it, submissions run normally but carry no cost data. You can enable it later with geni environment update.

Read both values from the CLI

bq ships with the gcloud CLI, so no extra install is needed. Sign in to the project that holds the export:

gcloud auth login
export PROJECT_ID=<project-id>

1. Find the dataset. The FOCUS export lives in a dataset named gcp_billing_immutable_…, suffixed with the billing account ID and the region:

bq ls --project_id=$PROJECT_ID --datasets --max_results=100

      datasetId
 -------------------------------------------------
  gcp_billing_immutable_01F464_AC8B27_889258_us

2. Find the table inside it. The FOCUS table is the one named gcp_billing_export_focus_…. A dataset often also holds gcp_billing_export_v1_… and gcp_billing_export_resource_v1_… — those are the older non-FOCUS exports and will not work:

export DATASET_ID=gcp_billing_immutable_01F464_AC8B27_889258_us
bq ls --project_id=$PROJECT_ID --max_results=100 "$PROJECT_ID:$DATASET_ID"

                 tableId                  Type
 ---------------------------------------- -------
  gcp_billing_export_focus_01F464_AC8B27_889258   TABLE

3. Read the location. This is the --focus-location value:

bq show --project_id=$PROJECT_ID --format=prettyjson "$PROJECT_ID:$DATASET_ID" \
  | grep '"location"'

Print both values ready to paste

With PROJECT_ID set, this finds the dataset and table for you and prints the two arguments:

DATASET_ID=$(bq ls --project_id=$PROJECT_ID --datasets --max_results=100 \
  | grep -o 'gcp_billing_immutable_[A-Za-z0-9_]*' | head -1)

TABLE_ID=$(bq ls --project_id=$PROJECT_ID --max_results=100 "$PROJECT_ID:$DATASET_ID" \
  | grep -o 'gcp_billing_export_focus_[A-Za-z0-9_]*' | head -1)

LOCATION=$(bq show --project_id=$PROJECT_ID --format=prettyjson "$PROJECT_ID:$DATASET_ID" \
  | grep '"location"' | cut -d'"' -f4)

echo "--focus-dataset $PROJECT_ID.$DATASET_ID.$TABLE_ID"
echo "--focus-location $LOCATION"

If DATASET_ID or TABLE_ID comes back empty, the export is not set up yet — follow Google’s setup guide linked above and wait for the first table to appear. A newly created export takes a few hours to publish its first rows.

No BigQuery access?

These commands read only dataset and table names, so a user with roles/bigquery.metadataViewer on the project is enough. If billing is managed by another team, send them this page and ask for the three parts.

Enable it on an existing environment

Pass the environment ID (from geni environment list) followed by both values:

geni environment update <environment-id> \
  --focus-dataset <project.dataset.table> \
  --focus-location <focus-region>

For example:

geni environment update 8f0de10b-8cc7-48dc-ad16-74a295d3bb46 \
  --focus-dataset useful-loop-496215-i8.gcp_billing_immutable_01F464_AC8B27_889258_us.gcp_billing_export_focus_01F464_AC8B27_889258 \
  --focus-location us

The environment moves to UPDATING while the change is applied. --focus-location requires --focus-dataset; running the command again with another table replaces the previous one.

Cost data appears after the next daily cost sync, so allow a day before checking geni costs submission. If it stays empty, the usual cause is a --focus-dataset that was missing the table name.