Alert Rules
A rule answers three questions: what it watches, when it fires, and where — on which plants. Open Alert Manager ▸ Rules and click New rule, or click a rule to change it. Open in Mirox
The editor groups the fields in five tabs:
- Condition — name, notes, level, whether the rule is active, the metric, when it fires and what happens without a value.
- Schedule — the weekdays and hours the rule is judged.
- Plants — which plants it covers.
- Notification — who and what is notified, the alert text and the receivers — your direct marketer among them.
- Preview — the replay on a plant and the history on your plants.
A dot on a tab marks a missing field or a problem there. Save stores the rule; Save as copy creates name (copy) — switched off until you enable it — which is the quickest way to make a variant of a rule.
Changes take effect on the plants at once. A rule can be switched off with its toggle in the list — it then stops being checked and its open alerts resolve.
The Rules table shows for every rule its name (with the badge Receivers only when nobody in the organization is notified), its description, the metric and the plants it covers, the Condition, the Level, its Receivers — a chip per receiver with the icon of its kind — what is Firing / pending right now, when it was last Changed (hover for the version and who changed it) and whether it is Active.
The Metric
Under What it watches you choose the value the rule observes:
- Platform series — the values Mirox records for every plant, such as the AC power of the inverters, the grid feed-in or the irradiance. Each is offered with its unit.
- Your metrics — the metrics your organization defined. A rule uses a metric exactly as a report would; if the metric is changed, the rule follows.
Among the platform series is the availability family — the share of 15-minute slots in which a unit was available, per unit (inverter, turbine, battery converter) and as the plant's share, plus Units unavailable as a plain count. An inverter counts available while it feeds in during daylight; a turbine or a battery converter while its operating state is at most stopped by command, around the clock — so a fault, a protection trip, a unit the logger cannot reach and a unit without any value count unavailable.
The field opens the same metric list as Data Export: search by name, description or id, grouped like there, and every row shows its unit, the steps it is recorded at, the plant types it exists on and — behind ⓘ — what it measures. The metric you chose is explained right under the field.
Then choose how the metric is judged:
- Plant total — one value for the whole plant, so one alert per plant.
- Each component — one value per inverter, combiner box, string, turbine or other component the metric reports on, so one alert per component that meets the condition. This is only offered for metrics that report per component.
Components you have hidden on the plant are left out: no alert opens for them, and an open alert closes when its component is hidden.
When It Fires
- Value — Latest value compares the most recent value. Over the last … min first folds the metric over that window — in the same way the metric folds in a report (average, sum, share, …) — and compares the result. The window is whole quarters of an hour.
- is — above, at or above, below, at or below, equal to, not equal to.
- Threshold — the number to compare with, in the unit shown.
- For at least (min) — the condition must hold without interruption for this long before the alert opens. Until then the alert is pending. Durations are multiples of 5 minutes.
- Clears only at (optional) — a second threshold for resolving. Without it the alert resolves as soon as the condition no longer holds. With it — for example fires below 10, clears only above 50 — a value hovering around the threshold does not open and close the alert again and again.
- Keep firing for (min) — keeps the alert open for this long after the condition cleared; a short dip back to normal does not resolve it.
Several times within a window
Switch on Fire when it happens several times to alert on something that keeps coming back rather than something that lasts. The rule then counts how often the condition started within the window — for example 3 times within 60 minutes — and fires when that number is reached. It resolves once the count in the window falls below it again.
When There Is No Value
If the metric has no value, the rule does what you choose — whether the plant sends no data or the value does not exist for this plant (for example the compensation of a plant without an agreed compensation):
- No alert — no value is never an alert; an open alert resolves.
- Alert — missing data is itself the alert, for example to notice a meter that stopped reporting. Like a condition, it opens only once the data has been missing for the rule's For at least duration, so one late value does not alert.
- Keep last state — an open alert stays open, a quiet rule stays quiet, until values arrive again.
A plant whose whole agent stops reporting is covered separately: it raises Alert rules not checked — see Engine status.
Only at Certain Times
On the Schedule tab, switch on Only during a schedule to limit the rule to weekdays and to a time of day, in the plant's local time; switched off, the rule runs always — every day, around the clock. With the switch on but no days and no hours chosen, the rule still runs at any time — a dot on the tab says so. A range such as 22:00 – 06:00 spans midnight. Outside the schedule the rule is not judged and an open alert resolves.
Choose Sunrise / sunset instead of a clock time for anything that depends on daylight. The rule is then active from sunrise plus the first offset to sunset plus the second, computed every day at the plant's location; positive minutes mean after, negative minutes before. +60 / −60 runs from one hour after sunrise to one hour before sunset, so the window follows the seasons without being adjusted. The weekdays still apply. On a day the sun does not rise at the plant the rule is not judged.
Level
Every rule sets the level of its alerts, from Very low to Critical — the same six levels the platform uses, described in Alarm Levels. The level decides who is notified, because everyone chooses in their own settings from which level on they want to hear about Alert Manager alerts.
Which Plants
On the Plants tab: All plants, the plants of chosen portfolios, or chosen plants. A plant can be targeted by at most 50 rules; an organization can have up to 500 rules.
If a rule cannot be checked on a plant — for example because the metric needs a series that plant does not record — the editor lists the plant under Not evaluated on some plants with the reason.
Except these plants leaves single plants out of all plants or of a portfolio — the rule never watches them, whatever else selects them.
Notification and Alert Text
Everyone in your organization who has access to the plant is notified according to their own notification settings (group Alert Manager). Under Receivers you add automated targets on top — a webhook, an email list or your direct marketer, each shown with its kind and the template it sends with — see Receivers.
Switch off Notify users for an alert that is meant for systems only: the rule then sends to its receivers, and nobody in your organization is notified. The alerts still appear on the Alerts tab, and the rule carries the badge Receivers only in the list. With no receiver chosen, nobody is told when the rule fires.
Alert text is the one line shown with every alert. Placeholders are filled in for each alert:
| Placeholder | Becomes |
|---|---|
{component} | the component, e.g. Inverter 3 (empty for a plant total) |
{value} | the value that was judged, with its unit |
{threshold} | the threshold, with its unit |
{park} | the plant name |
{rule} | the rule name |
Example: {component} delivers only {value} (limit {threshold}).
On a rule that watches the turbines of a wind plant one by one, the tab also offers Report outages: with it on, every firing turbine alert is an outage — sent to every receiver of the rule as start, renewal and end, with the outage reason and the remaining power you choose here, as many hours ahead as the plant's direct marketer wants (2 without one). The switch is forced on while the direct marketer's receiver is attached. The tab names the plants that are linked to another marketer or to none, and those whose agent still needs an update for outage reports; a rule judged on the plant total is never reported. A direct marketer receives no alert text, only the outage — see Direct marketer.
Replay on a Plant
On the Preview tab, Replay on a plant shows what the rule would have done over the last 1, 2, 3 or 7 days on a plant you pick: the watched values, the threshold, and every period in which an alert would have been open. It is decided by the same engine the plant runs, so what you see is what the plant will do. Use it before saving, to find a threshold that fires when it should and stays quiet otherwise.
If the plant delivered no values for the metric in that period, the replay says so — pick another plant or a longer period.
History on your plants goes further: it replays the rule on every plant it targets over the last day, 30 days or year — on the same 5-minute grid the plants evaluate, exact also for a whole year — and shows per plant how often it would have alerted, for how long in total, and every alert on a timeline; hover one for its start, end, duration and extreme value. Use it to find false alarms before you save.
A history over many components — every string of a large plant — or over a whole year is shown first with what it involves: the components and the values per plant. It then runs plant by plant in short steps with a progress bar; deselect plants or pick a shorter period to get the answer sooner, or stop it. A rule that watches more than 500 components is not evaluated — the same limit the plants apply. A rule on a handful of values, such as the feed-in, answers right away.