Alert rules
Route each failure to the right channel and escalate only when nobody answers.
3
severity levels
2
failures before a page
0
noise from known work
Simple rules
Pick a service, a condition and a channel. There is nothing to script.
Escalation that waits
Move to the backup after five minutes, then the team lead.
Every channel
Phone, SMS, email, Slack, Teams and webhooks all work the same way.
In detail
Rules that read like sentences
If checkout fails from two regions, page whoever is on call and open a high severity incident. That is the whole rule, and anyone on the team can read it.
Escalate in steps
Decide who hears first, how long they have to answer and who comes next. Policies can differ by service, so a slow blog does not page the person looking after payments.
Maintenance and quiet hours
Maintenance windows that silence planned work
Grouping that turns twenty failures into one incident
Quiet hours for anything below high severity
All features
Uptime checks
Every service checked once a minute from six regions.
Incident timeline
One running record from the first failed check to the fix.
Status pages
A public page that updates before customers start asking.
On-call schedules
Fair rotations that everyone on the team can read.
Alert rules
Decide who hears about what, and how loudly.
Response time
Spot the slowdown before it turns into an outage.
