Cloud Observability

Epic link: sourcegraph/customer#1151


Metrics are gathered from all resources using the included Prometheus instance. This instance scrapes and stores the metrics locally as well as forwards them to the Managed Prometheus service provided by GCP.

Only metrics queried in our monitoring generator are forwarded - this allowlist is automatically generated. If you’d like a new metric to be queryable in a centralized manner, you must create a dashboard panel for it.

These metrics are viewable through our centralised Grafana instance hosted at:

Multi-instance dashboard

We generate a dashboard that renders panels that opt-in to a multi-instance overviews dashboard.

Panels in this dashboard show the panel’s query grouped by project_id, each of which represents a Cloud instance. The template variable dropdown at the top allow you to select instances to compare, which is persisted to the URL.

To opt-in a panel to this multi-instance dashboard, see how to add monitoring.

Common operations


Traces from Cloud instances are now all available in GCP tracing. To view a trace, follow the same How to use traces procedure. After clicking “View Trace” it may take a minute or two for the trace to become available in GCP. If the trace doesn’t immediately render, wait a little bit and refresh the page.


Application logs are automatically forwarded from the Cloud instances and made available in Google Cloud Logging. To view logs for a given Cloud instance, visit and select the project for the desired instance. Then execute the correct corresponding query.

Application Logs (Sourcegraph)


HTTP Load Balancer Logs


Postgres Logs