Alert problems generally fall into one of two opposite categories, and the diagnosis for each is different.
Not receiving alerts you should
- Confirm the alert condition itself is actually configured correctly on the check in question — see Alert Conditions. A threshold set too high, or a check-attempt count set very generously, can mean a real problem simply never crosses the line that would trigger a notification.
- Confirm the contact group and time period assigned to the alert are correct — an alert routed to the wrong contact group, or restricted to a time period that excludes when the problem actually occurred, will fire without anyone being told.
- Check whether a maintenance window is active and unintentionally still covering the host — a maintenance window left open past its intended end date will silently suppress real alerts along with the planned ones it was meant to cover.
- Confirm SMTP is actually working — see Email Notifications, and send a test notification if you haven't confirmed delivery recently. A notification that iPM believes it sent, but that never arrives, is usually a mail delivery issue rather than an iPM configuration problem.
- Check whether the alert was acknowledged — an acknowledged alert intentionally stops repeat notifications until it resolves or the acknowledgement expires, which can look like "alerts stopped working" if you weren't the one who acknowledged it.
Receiving far too many alerts
- Revisit the alert condition — a threshold with no duration requirement, or a check-attempt count of one, reacts to every brief blip. See Alert Conditions for how duration and check attempts reduce this.
- Check for flapping — a check that's rapidly switching between states will generate a notification on every transition unless flapping detection is active for it.
- Review the re-notification interval — see Email Notifications — if it's set very short, a single ongoing problem can look like a flood of separate alerts rather than one issue with periodic reminders.
If alerts are firing correctly but reaching the wrong audience as an incident continues, that's usually an Alert Escalation configuration issue rather than a problem with the alert itself.