Monitoring

Three things get monitored

Uptime from outside, application behaviour from inside, and infrastructure underneath. Products lead with one and add the others.

Start with the failure you actually had last quarter, then buy for that.

Alerts are the product

A monitor that notices nothing useful is noise, and noise gets muted. Look at routing rules, escalation, on-call schedules, and how easily an alert is silenced without being forgotten.

Retention and cardinality

Metrics pricing depends on how much detail you keep and for how long. High-cardinality data is where costs run away, and it is exactly the data that helps during an incident.

What to check

Trigger a real alert at three in the morning during the trial. How it reaches you is the whole purchase.