An alert that fires all night gets muted by morning. Carabdheer answers that with two rules: quiet hours, and one job for each channel. This note is the routing table, so you can move the clock and watch it decide.
WhatsApp is for urgent. Telegram is for look at it today. Email carries the summary. From 23:00 to 07:00 only a site that is down and a backup that was missed get through. Everything else waits for the morning summary.
| Rule | Fires when | Goes to |
|---|---|---|
| Site down | health check fails twice | WhatsApp, now |
| Backup missed | no backup event in 26 hours | WhatsApp, now |
| Memory high | over 85% for 5 minutes | Telegram, look at it today |
| Host offline | agent silent 3 minutes | Telegram, look at it today |
| Error spike | more than 10 errors in 10 minutes | Telegram, look at it today |
| New error type | an exception heard for the first time | Telegram, look at it today |
| Failed jobs | failed after the last retry | Telegram, look at it today |
| Bounce rate | over 2% across 24 hours | Telegram, look at it today |
| Certificate expiring | fewer than 14 days left | Telegram, look at it today |
What the slider shows
Drag the time past 23:00 and most of the table goes quiet. Nothing is dropped. It moves to the 07:00 summary, where it is read once, with coffee, instead of nine times in the dark.
Turn on planned maintenance and a site going down stops being an alert at all. It opens no incident, because someone already said it would happen. The status page says why.
Still on the bench
The nine rules are fixed in code and the quiet hours are real. Which channel each rule uses by day is the part we are still tuning with the people who carry the phone. Thresholds and channels are settings on purpose: what counts as a problem is decided once, how loudly it speaks can change.


