A quiet chart can hide a broken connection
A sensor reading of 21 degrees may look reassuring long after its transmitter has stopped. The same problem affects server agents and scripts: an old value remains visible, while the next observation never arrives. Heartbeat monitoring answers a separate question from a threshold alarm: is this source still reporting when expected?
MetriBeacon supports missing-contact and missing-value rules. Choose the signal according to what you need to detect. A monitor can remain in contact while one of its measurement series stops updating. Treating these as the same event makes an alert harder to diagnose.
Heartbeat versus measurement freshness
| Observation | Useful interpretation | Suitable check |
|---|---|---|
| The monitor reports regularly | The reporting path is active | Missing heartbeat |
| A temperature series receives new readings | That specific measurement is current | Missing value |
| Temperature exceeds a threshold | A received value is outside your range | Numeric rule |
A heartbeat does not prove that every measurement is correct. A threshold rule does not prove that its last input is fresh. For a useful first setup, monitor contact for the device and freshness for the one series you rely on most.
Choose a timeout from the reporting interval
Suppose a script normally sends an observation every minute. A timeout of exactly one minute can trigger around normal delivery jitter. A five-minute timeout may suit a non-critical sensor, but it also delays detection. Pick a margin based on the expected interval, normal delays and the time you can tolerate without new evidence.
Use a different expectation for a daily report. A daily script should not be treated like a continuously connected device. For scheduled work, a job monitor can describe the expected start and completion more clearly; see the cron job monitoring guide.
Set up a missing-data rule
- Open the monitor and confirm that normal observations arrive.
- Identify whether you need monitor contact or freshness for a specific series.
- Create the corresponding missing-heartbeat or missing-value rule.
- Set the waiting period longer than the normal reporting interval, with a deliberate margin.
- Select a confirmed notification destination and save the rule.
- Test the complete path using a source you can safely pause.
The missing-heartbeat documentation explains the rule types and timing. A newly created rule should establish its own observation period; do not assume it will reconstruct every earlier interruption.
Test failure and recovery deliberately
Pause reporting from a test script rather than disconnecting a production system unexpectedly. Record the last successful observation, wait for the configured interval and check the incident together with its delivery status. Restart reporting and verify recovery. This distinguishes an evaluation problem from a message that could not be delivered.
Check the source and the monitoring path separately. Missing data can result from a stopped process, lost connectivity, an invalid key or an account intake limit. The alert tells you evidence is missing; it does not identify which of these causes applies.
Keep gaps visible in your dashboard
Do not send zero merely to fill a hole. Zero degrees, zero free disk space and zero errors are real measurements with different meanings. An unavailable measurement should remain unavailable until you can collect a valid value again.
Read the current value, unit and last-update time together. When comparing charts, use the same period and consider coverage. The server monitoring guide shows how freshness complements CPU, memory and disk readings.
Frequently asked questions
Can one heartbeat cover all my sensors?
It can show contact with their shared transmitter. It cannot prove that each sensor is still producing readings. Add a missing-value rule where an individual measurement matters.
Should I alert after one missed interval?
Only if your reporting timing and response needs support it. A small margin reduces messages caused by ordinary scheduling and network delays. Test your choice against real delivery times.
Does no new value mean the value is zero?
No. It means there is no new observation. Preserve that distinction in your collector and when reading a chart.
Start with one reporting path
Create a free account, connect one source and verify its normal rhythm before adding missing-data rules. If you also operate a public website, the Website/API template below adds an independent availability check; it does not replace a heartbeat for your own measurements.
