On a VM with the Stackdriver Monitoring agent, you need an alert if a Unix process hasn't run for at least 5 minutes and you can't modify the app to emit metrics or logs. Which alert condition do you configure?

The correct answer is Metric absence. Since the Stackdriver Monitoring agent is installed, it automatically collects process health metrics, including whether a specific Unix process is running. A "metric absence" alert condition can then be configured to trigger if the metric indicating the process's presence is not reported for a specified duration (e.g., 5 minutes). Uptime check is incorrect because it monitors the availability of a public endpoint, not an internal process on a VM. Process health is a type of metric, not an alert condition itself. Metric threshold is incorrect because it triggers when a metric's value crosses a certain level, not when the metric stops being reported altogether.