Being told when it matters.
WHAT AN ALERT SHOULD MEAN
Someone must do something now.
WHAT HAPPENS WHEN ALERTS DO NOT MEAN THAT
They are ignored, including the important ones.
WHAT TO ALERT ON
Service unavailable Error rate sharply raised Disk nearly full Backup failed Certificate expiring soon Queue not being processed
WHAT NOT TO ALERT ON
Brief spikes Anything self-correcting Conditions nobody can act on
WHERE TO SEND ALERTS
Somewhere a person will see, by a route independent of the failing system.
THAT POINT
An alert about mail failing, delivered by mail, does not arrive.
WHAT TO CONFIGURE
Escalation, if the first person does not respond Quiet handling of known issues
WHAT TO AVOID
Alerts to a shared address nobody owns Alerts firing repeatedly for the same condition
WHAT TO REVIEW
Every alert that fired, monthly.
WHAT TO DO ABOUT ONE THAT WAS NOT ACTIONABLE
Change or remove it.
WHAT TO TEST
That alerts actually arrive.
