Question

Your team runs a payment API on GKE that exports custom Prometheus metrics collected via Google Cloud Managed Service for Prometheus. Stakeholders want a single dashboard tile that displays the percentage of HTTP requests returning 5xx status codes over a 5-minute rolling window, computed from two counter metrics: total_requests and total_5xx_responses. What is the most appropriate way to build this tile in Cloud Monitoring?

Accepted Answer

Use a ratio-based query (e.g., MQL or PromQL) that divides the rate of total_5xx_responses by the rate of total_requests over a 5-minute alignment window and multiplies by 100.

Answer

Create two separate line charts, one for each counter metric, and let viewers visually estimate the error percentage.

Answer

Configure a log-based metric that counts 5xx entries and display its raw count on the tile.

Answer

Create an uptime check against the payment API endpoint and display its availability percentage.