Observability, backups, and CDC

This page covers dashboard surfaces that often look empty or disabled even when the system is behaving correctly.

Use it to distinguish empty-but-valid states from operational failures.

Metrics are empty

Metrics require an active profile on a supported provider. The current route supports AWS and GCP. If the database is newly provisioned, wait for the collector interval and refresh the selected time window.

For AWS, confirm CloudWatch Agent is present on the replica host and that the regional private gateway is healthy. For GCP, confirm Telegraf is forwarding DogStatsD metrics into Cloud Monitoring and the Analytics Engine dataset.

Fix: choose a supported time window, wait for one collector interval, then verify provider telemetry before changing application code.

Logs show no rows

An empty log stream for the selected window is valid. Try switching between 15 minutes, 1 hour, 6 hours, and 24 hours. Live ranges poll for newer logs, while historical ranges remain fixed until the range changes.

Fix: widen the time window first. Treat a friendly loader failure as different from a valid empty stream.

Backup button is disabled

Manual backups require:

• an active database profile
• GCP or AWS backup infrastructure
• no active conflicting backup, restore, import, provision, or decommission workflow

Restore button is disabled

Restore eligibility is intentionally strict. The selected backup must be completed, have complete artifacts, include restore metadata, match the current region and cluster shape, and match TigerBeetle version and storage characteristics.

Webhook history is empty

The Webhooks page shows managed CDC delivery history. If Managed AMQP is not active, configure CDC first from the Cluster Change Data Capture tab. If Managed AMQP is active, check destination status, HTTPS URL validation, authentication secrets, and delivery filters.