How do I avoid false alarms and choose what happens when a target is unreachable?
Raise the thresholds on the monitor form. A threshold is the number of consecutive bad results needed before a status update is published. Set Execution failures and Publish as to control what a timeout or error means.
Steps
- Open the monitor from Monitoring and select it to edit.
- Under Thresholds, set:
- Degraded: consecutive Degraded results before publishing Degraded Performance.
- Critical: consecutive Critical results before publishing Critical Outage.
- Execution failures: consecutive timeouts and errors before publishing the status chosen in Publish as.
- Save the monitor.
Questions
What is an execution failure?
The check itself failed: the server timed out, refused the connection, answered with a 4xx code, or the answer did not match any of your values.
Which values should I use?
A threshold of 2 or 3 removes most one-off blips. Higher values delay real alerts: the delay is the threshold times the interval.
Is a status update published on every failing check?
No. It is published once, when the streak reaches the threshold. A later Operational result publishes an "operational" update.