Composite Monitor Use Cases

Docs > Alertes > Guides sur les monitors > Composite Monitor Use Cases

Cette page n'est pas encore disponible en français, sa traduction est en cours.
Si vous avez des questions ou des retours sur notre projet de traduction actuel, n'hésitez pas à nous contacter.

Overview

This guide lists non-exhaustive use cases for composite monitors. These examples illustrate how composite monitors can be configured to address various use cases in monitoring environments:

Error rates
Monitor frequent metrics
Step monitor
Renotifying on recovery
Delay on notification

Error rates

Alert when the error rate exceeds a threshold only when hits are above a certain number.

Create two monitors:

Monitor A: Alert when trace.requests.request.errors / trace.requests.request.hits > X
Monitor B: Alert when trace.requests.request.hits > Y

Composite Monitor C: Alert when both Monitor A and Monitor B are in alert (A && B).

Monitor A	Monitor B	Composite Monitor C
Alert Error rate above threshold	Alert Hits above threshold	Alert
Alert Error rate above threshold	OK Hits below threshold	OK Only 1 condition met, no alert
OK Error rate below threshold	Alert Hits above threshold	OK Only 1 condition met, no alert

For more state combinations, see Composite Monitor.

Monitor frequent metrics

Monitor latency for services, ignoring occasional spikes due to low traffic. For example, a period of time during the night where services report very few values.

Create two monitors:

Monitor A: Alert when latency > X
Monitor B: Alert when sum:latency{*}.rollup(count) > Y over the last 1 hour

Composite Monitor C: Alert when both conditions are met.

Monitor A	Monitor B	Composite Monitor C
Alert Latency above threshold	Alert More than Y metrics	Alert
Alert Latency above threshold	OK Less than Y metrics	OK Not enough metrics
OK Latency below threshold	Alert More than Y metrics	OK Latency below threshold

Step monitor

Trigger an alert in the absence of a paired metric. For example, log metrics for sent/received, down/up, or create/resolve. You can adjust the evaluation windows for the monitors if paired metrics are expected to be N minutes apart.

Monitor A: Alert when action:create is above 0
Monitor B: Alert when action:resolve is above 0

Composite: Alert if a && !b.

Monitor A	Monitor B	Composite Monitor C
Alert Action create above 0	Alert Action resolve above 0	OK
Alert Action create above 0	OK	Alert Action resolve not present
OK	Alert Action resolve above 0	OK

Renotifying on recovery

Renotify on recovery using two monitors with a timeshift.

Monitor A: Current metric status
Monitor B: Past metric status using timeshift

Composite Monitor: Logic is !a && b.

Monitor A	Monitor B	Composite Monitor C
Alert Real-time metric	Alert Past metric	OK
Alert Real-time metric	OK Metric not triggered	OK
OK Metric not triggered	Alert Past metric	Alert

Delay on notification

Alert after errors persist for a set duration. For example, a set of errors triggered for at least 15 minutes.

Monitor A: Real-time metric
Monitor B (timeshifted): Metric shifted by X minutes

Composite Monitor: Alert if a && b.

Monitor A	Monitor B (timeshifted)	Composite Monitor C
Alert Real-time metric	Alert Past metric	Alert
Alert Real-time metric	OK Metric not triggered	OK
OK Metric not triggered	Alert Past metric	OK