Skip to main content

Metabase Production - Quick Fix Guide

One-Command Fix for Production​

# Run the automated fix script (auto-detects barto-prod)
./scripts/fix-metabase-cloudsql.sh production

Manual Production Fix​

If you prefer to fix it manually:

# 1. Verify Cloud SQL instances exist
gcloud sql instances describe metabase-db-production --project=barto-prod
gcloud sql instances describe flowpos-db-production --project=barto-prod

# 2. Fix the service configuration
gcloud run services update flowpos-metabase-production \
  --region=us-central1 \
  --project=barto-prod \
  --clear-vpc-connector \
  --add-cloudsql-instances=barto-prod:us-central1:metabase-db-production,barto-prod:us-central1:flowpos-db-production

# 3. Verify the fix
gcloud run services describe flowpos-metabase-production \
  --region=us-central1 \
  --project=barto-prod \
  --format="value(spec.template.metadata.annotations['run.googleapis.com/cloudsql-instances'])"

Production Configuration​

  • Service Name: flowpos-metabase-production
  • Project: barto-prod
  • Region: us-central1
  • Cloud SQL Instances:
    • barto-prod:us-central1:metabase-db-production (Metabase metadata database)
    • barto-prod:us-central1:flowpos-db-production (Application database)

What Changed​

The Dockerfile now includes a wait script that:

  1. Waits for Cloud SQL Proxy to be ready before starting Metabase
  2. Tests the connection to ensure the proxy is accessible
  3. Prevents Metabase from crashing on startup due to connection failures

This fix applies to both staging and production - the same Dockerfile is used for both environments.

Next Steps​

  1. The fix is already in the Dockerfile - no additional changes needed
  2. The next deployment (via GitHub Actions or manual) will automatically use the new image
  3. Monitor the logs to see the wait script in action: 
    gcloud logging read \
      "resource.type=cloud_run_revision AND resource.labels.service_name=flowpos-metabase-production" \
      --limit=50 \
      --project=barto-prod \
      --format="table(timestamp,severity,textPayload)"

Troubleshooting​

If issues persist, see the full troubleshooting guide: cloudsql-troubleshooting.md