How alerts work¶
This page explains what happens between a reading arriving and an alert clearing, so you can set alerts up with confidence and know what to expect when one fires.
The parts of an alert¶
An alert watches one channel on one device and has four parts:
| Part | What it decides | Set on |
|---|---|---|
| Trigger | What counts as a problem: a value past a threshold, a digital input changing state, or no readings at all | The alert |
| Delay | How long the problem must last before anyone is told | The alert |
| Message | What the text, email or call says | The alert |
| Who to notify | A schedule group, which picks the people on call and how to reach them | The alert (which group), and the group (who and how) |
A schedule group is shared: many alerts can point at the same group, so when your on-call rota changes you change it once. See Schedule groups.
The life of an alarm¶
reading arrives
|
v
condition met? --no--> nothing happens
|yes
v
DELAY: condition must hold for the whole delay --recovers first--> nothing happens
|
v
FIRES ---> schedule group picks who is on call now
| |
| +-- everyone at once (text / email)
| +-- or one person at a time until someone acknowledges (text / voice)
v
ACKNOWLEDGED (optional) ---> everyone else is told who has it; no more notifications
|
v
value recovers past the threshold by the hysteresis amount
|
v
CLEARS ---> "cleared" message to the people on call now
|
v
DOCUMENT cause and resolution on the Alert Notes page
| Stage | What happens | Where you see it |
|---|---|---|
| Condition met | A reading breaks the rule, or the channel goes quiet for the inactivity period | Nothing yet |
| Delay | Whisker.io waits for the delay you set. Every reading in that time must still break the rule, or nothing fires | Nothing yet |
| Fires | The alert is set. The schedule group works out who is on call and starts notifying them | Status column on the Alerts tab shows Firing. Red status icons on the map, tables, tiles and dashboards |
| Notifying | Texts, emails and calls go out as the schedule and the alert's text delivery option say | Click the status to see every attempt |
| Acknowledged | Someone replied to the text, pressed 1 on the call, or clicked Ack in the portal. The sequence stops and everyone else is told | Status shows Acknowledged and who acknowledged |
| Unacknowledged | A one-at-a-time sequence ran out of people and nobody acknowledged. The group's When nobody answers setting decides what happens | Status shows Unacknowledged. You can still acknowledge it |
| Clears | The reading comes back past the threshold by the hysteresis amount, or the device reports again | Status icons return to normal. The occurrence moves to the device's Alerts History and the Alert Notes page |
Note
An alert notifies once when it fires and once when it clears. It does not send repeat reminders while it stays active. If you want someone else told when a problem goes on too long, stack a second alert with a longer delay (see Escalation), or use one person at a time text delivery so the alarm keeps moving down the list until someone acknowledges it.
Trigger types¶
| On the alert editor | Channel type | Fires when | Clears when |
|---|---|---|---|
| Value crosses a threshold | Analog (temperature, level, pressure...) | The value is Above or Below the threshold | The value comes back past the threshold by the hysteresis amount |
| Input changes state | Digital (on/off) | The input goes High (on / closed) or Low (off / open) | The input goes back |
| Device stops reporting | Any | No reading arrives on the channel for the period you set (at least 30 minutes) | The device reports again |
Tip
Give every important device both kinds of protection: a threshold alert for the dangerous reading, and a Device stops reporting alert so you find out when it goes silent. A dead sensor and a safe reading look the same on a map icon until the inactivity alert fires.
The delay¶
Under Fire you choose Immediately or Only if it stays there for a number of minutes. With a delay, the alert fires only if every reading in that time is past the threshold. If the value recovers even once, the timer starts over. This filters out momentary blips: a door opened, a defrost cycle, a pump starting.
If the device sent very few readings during the delay (fewer than about one every ten minutes), Whisker.io does not wait: it fires on the reading that is past the threshold, so a slow-reporting device never hides a real problem.
Inactivity alerts have no separate delay: the inactivity period is the delay.
Hysteresis: preventing alert flapping¶
Hysteresis is a dead band applied when an alert clears. Without it, a value hovering right at the threshold sets and clears the alert over and over ("flapping"), burying your team in texts until they start ignoring them.
With hysteresis, the alert sets at the threshold but does not clear until the value recovers past the threshold by the hysteresis amount:
Threshold: Above 40 Hysteresis: 5
value: 38 41 44 42 39 37 34 33
alert: - SET ------------------- CLEAR
(39 and 37 do NOT clear it:
not yet below 40 - 5 = 35)
Choosing a value: start with the size of the signal's normal wiggle, meaning sensor noise plus process swings (a compressor cycle, waves in a tank). Big enough that the alert doesn't re-fire on the wiggle, small enough that "cleared" still means genuinely recovered. For refrigeration, 4 to 5 °F is a common choice. A hysteresis of 0 clears the moment the value is back on the safe side.
Alerts that are off, and snoozed devices¶
- An alert switched Off never notifies anyone. It is still evaluated, so if it was firing when you turned it off, it still clears normally.
- A snoozed device's alerts never notify while the snooze is on. They too can still clear. See Snoozing alerts.
Where alerts show up¶
The Alerts tab. The Status column shows Firing, Acknowledged or Unacknowledged for any alert with an open alarm, with the time it started and the latest attempt. Click the status to see every text, email and call sent for that alarm.
Device panel. Every device's panel has an alerts status (Active, Snoozed, or Active/Scheduled to Snooze) and two tabs:
- Triggered Alerts: alerts that are set right now, with the time and value that set each one.
- Alerts History: past occurrences with set and clear times and values.
Everywhere else. An active alert turns the location red on the map, marks the device in the location tree, tiles and tables, and colors the channel's dashboard widgets red. Clicking any of these issue icons opens the list of what is wrong.