Alarm Levels and Notifiable Events

Mirox watches solar, wind and battery plants with one alarm system. This page lists every event that can notify you, the level it opens at, what triggers it, how fast it is detected and what ends it — so you can decide from which level upwards you want to be told. The tables on this page are generated from the same catalog the platform uses to send notifications, so they always describe what actually happens.

How a finding becomes a notification

Mirox detects a condition at a plant — a unit stopped by a fault, a data source that went silent, a safety alarm — and opens an event.
Mirox assigns the level. Every event type has a fixed level from Critical down to Very low. The level is defined by the platform, not by you, and it is the same on every plant.
You choose the minimum level per category. On your notification settings, each category (Production, Availability, Alarm system, …) has a minimum level. An event notifies you if its level is at or above that minimum and you have switched on at least one channel for the category.

Why the levels are the same for everyone, and what to do when one does not suit you, is explained under Why levels are fixed.

One notification per situation. A plant with forty inverters, twelve turbines or a dozen battery units does not send forty notifications when a feeder trips. Each unit's own state is recorded as an event at Normal, and the site as a whole raises one notification at High — "components down" — that lists every affected unit and is updated as units join or recover. Only safety alarms bypass this bundling.

The six levels

Level	Numeric priority	What it means
Critical	≥ 1500	People or the plant may be in physical danger now: fire, smoke, thermal runaway, a fire suppression that released, an arc in a wind turbine or a battery storage system, an intrusion alarm. Sent at once, every time — even during a service visit and even when many arrive together.
Very high	≥ 1300	A deliberate change disturbs data collection or a capability of the site is lost (a paused or deleted data source, a revoked credential). The default floor of the alarm-system row.
High	≥ 1100	Units produce nothing because they report a fault, or we have been blind to them or to the whole plant for long — one notification per site situation, not one per unit.
Normal	≥ 900	A single unit's state behind that site notification, or a device fault while the unit keeps running. Visible in the event list, not notified by default.
Low	≥ 700	Warnings and housekeeping states — a stop someone ordered.
Very low	≥ 0	Records without operational effect (comments, labels). Never notified.

Category	Admin	Moderator	Technical asset manager	Commercial asset manager	Member	External
Production	off (opt-in from Critical)	off (opt-in from Very high)	on from High	off (opt-in from High)	on from Very high	off (opt-in from Very high)
Availability	off (opt-in from Critical)	on from Very high	on from High	off (opt-in from High)	on from Very high	off (opt-in from Very high)
Sensors	off (opt-in from Critical)	off (opt-in from Very high)	on from High	off (opt-in from High)	off (opt-in from Very high)	off (opt-in from Very high)
Alarm system	off (opt-in from Very high)	off (opt-in from Very high)	on from Very high	off (opt-in from Very high)	off (opt-in from Very high)	off (opt-in from Very high)
Components	off (opt-in from High)	off (opt-in from High)	off (opt-in from High)	off (opt-in from High)	off (opt-in from Very high)	off (opt-in from Very high)
Monitoring agent	on from Critical	on from Critical	on from Critical	on from Critical	on from Critical	on from Critical
Tickets	on from Very high	on from High	on from High	on from High	on from Very high	on from Very high

Critical

People or the plant may be in physical danger now: fire, smoke, thermal runaway, an extinguishing system that released, an electric arc in a wind turbine or a battery storage system, an intrusion alarm. Critical notifications are sent at once, every time — without bundling, also during a service visit, and even when many arrive together.

Typical: a battery container reports thermal runaway; a turbine reports smoke in the nacelle; the site's intrusion alarm is raised. An inverter's own safety alarm (an arc fault or a residual-current fault it reports) is recorded as an event at Normal and is not notified by default.

Very high

A deliberate change disturbs data collection, or a capability of the site is lost — for example a paused or deleted data source, or a revoked credential. No plant-equipment event opens at this level today; it mainly serves as a threshold: it is the default minimum of the Alarm system row (so only the intrusion alarm itself notifies) and of the Member and External roles.

High

Units produce nothing because they report a fault, or we have been blind to them — or to the whole plant — for a long time. One notification per site situation, not one per unit.

Typical: "components down" on a site where two turbines stopped by a fault; a data logger that has delivered nothing for 30 minutes; the monitoring VPN to the site is down; an irradiance sensor confirmed defective; the alarm panel reports a fault.

Normal

A single unit's state behind the site notification, or a device fault while the unit keeps running. Visible in the event list and on the plant pages; most of these do not notify by default.

Typical: one inverter in outage; one turbine stopped by a fault; one battery unit not communicating; a battery converter reporting a fault while it keeps running; the communication cabinet running on its UPS.

Low

Warnings and housekeeping states.

Typical: a string under investigation before it is confirmed; a sensor that reads dark for an hour; a manufacturer warning.

Very low

Records without operational effect, such as comments or labels. Never notified.

Defaults for your role

You do not have to configure anything: your organization role sets sensible defaults, and you only change what you want to be different.

Role	Site categories (Production, Availability, Sensors, Control signals)	Alarm system	Monitoring agent
Asset Manager (Technical)	On, from High	On, from Very high	On, from Critical
Moderator	Availability on, from Very high; the others off (start at Very high when you switch them on)	Off (Very high)	On, from Critical
Admin	Off (start at Critical when you switch them on)	Off (Very high)	On, from Critical
Asset Manager (Commercial)	Off (start at High)	Off (Very high)	On, from Critical
Member	Production and Availability on, from Very high	Off (Very high)	On, from Critical
External	Off (start at Very high)	Off (Very high)	On, from Critical

Channels. Site categories notify through the App (notification center, badge and mobile push) by default. Email is off for them by default and can be switched on per category; webhooks are always your own choice.
Monitoring agent is armed for every role but starts at Critical. Agent actions are Normal or Low, so nothing arrives until you lower the row.
Why technical managers get High by default: they answer for the plant's operation. High is exactly the level of "the site needs someone": units down, a site you cannot see. Below High you would get every single unit's state, which the site notification already summarises.

Why levels are fixed

A level is a promise about consequences: Critical always means possible physical danger, High always means a site situation that costs production or leaves you blind. That promise only holds if it is the same everywhere, so the levels are defined by the platform — not per customer and not per plant.

Automation needs one vocabulary. A technical manager who looks after many plants automates the response: a webhook that triggers a call-out, an escalation step that alerts the technician on call. Those steps can only be written once for every situation if every event is tagged the same way on every plant. If levels differed per customer or per plant, a colleague's "High" would mean nothing to you, and the same rule would fire on different things for different people.
The thresholds are already chosen per data source. The agent at each plant knows how its data arrive and picks reasonable timing for each source: a wind turbine's 10-minute rows from the manufacturer's database are judged differently from a battery's minute-by-minute data (see Timing). What matters is that a real problem is detected and acted on — not a threshold tuned plant by plant.
If a level does not suit you, tell us. In your notification settings, Suggest a different level on a category or an event sends your suggestion to our support, optionally with the latest example from your plants. We review it and, if it holds, adjust the level for everyone.

What is yours to choose is how far down the list you want to be told: the minimum level per category.

Metric-based alerts (planned)

Event alarms keep their fixed levels. For anything about a measurement — "tell me when this value crosses that threshold" — metric-based alerts are planned: thresholds on any metric of the same unified metric collection the Metric Export uses, with full flexibility over what to watch and when to alert.

Timing: when we know, when we are blind

How fast an event opens depends on what we know:

When the data says a unit is in fault or stopped, we are not guessing — only a short confirmation stands between the reading and the event.
When no data arrives, silence is not proof of a fault: a network blip, a logger restart or a software rollout look exactly the same for minutes. So we wait longer before claiming an outage — long enough to clear the gaps a healthy data path shows, short enough that a real loss is still reported the same morning.

Every threshold below is a named constant in the platform's code; the generated tables on this page read the same values.

The data says fault or stop

Equipment	Unit event	Site notification ("components down")	Why
Wind turbine stopped by a fault or a protection	On the first 10-minute row that reports it	After 2 consecutive fault rows — about 20 minutes (about 23 with data-path lag)	One 10-minute row is already ten minutes of evidence. A single interval can be a stop the turbine resets by itself, so the site notification waits for the second.
Battery unit whose status word says fault	After 5 consecutive one-minute samples	After 15 minutes of fault	Battery controllers retry a trip within minutes; a real fault stays (one converter fault we observed stood for 17 hours).
Inverter / combiner box producing nothing in good light	After 3 judged 5-minute windows — about 20 minutes	30 minutes later — about 50 minutes after the first zero	Solar outages are inferred (zero output compared with the light), not read from the device, so they need more confirmation.
Safety alarm (wind turbine, battery storage)	At once, no debounce	— (safety alarms notify on their own)	Danger is never debounced. The latency is the data path's own: a turbine's 10-minute data (typically about 10 minutes), the battery check every minute (within about 2 minutes).

What is silent	Unit event	Site notification	Why these numbers
One turbine, while the other turbines of the site deliver	After 30 minutes (three missed rows)	After 3 hours; after 12 hours if the turbine's last state was a service visit or a commanded stop	A healthy turbine's newest row is 1–11 minutes old, rarely 19. Service visits that switch a turbine's communication off with it last 0.5–4.5 hours.
The whole turbine data source (the SCADA database)	—	Source availability when its newest row is older than 30 minutes, measured against the database's own clock	Over 38 park-months there was not a single whole-site gap longer than 30 minutes; both real outages observed (56 and 73 minutes) are caught. Measuring against the database clock keeps a software rollout on our side from triggering it.
One battery unit, while its data source keeps delivering	After 15 minutes	After 30 minutes	Measured over 30 days on six storage sites: units report every 60–68 seconds; gaps of a single unit while the others delivered stayed at or below 2.5 minutes (99.9th percentile), the largest jitter seen was 5.7 minutes, and there was no false gap of 5 minutes or more. The real losses (27, 162 and 178 minutes) are all caught. The rule is "the larger of 15 minutes and three times the 99.9th-percentile gap" — here 15 minutes.
One inverter / combiner box, while its siblings report	After 1 hour of daylight	After 12 hours	Inverters sleep at night and wake at different times; the clock only runs in daylight.
A data source (logger, SCADA database, battery controller)	—	Source availability after 30 minutes (or three of its own reporting intervals, if it reports less often); checked every 5 minutes, so 30–35 minutes	Long enough to ride out a restart or a short network interruption.
The VPN to the site	—	About 8–11 minutes: three failed one-minute checks after the handshake went stale, plus a 5-minute hold	The hold drops the short nightly 1–3 minute blips, so they notify nobody.
The site network (gateway, devices) while the VPN is up	—	Three failed one-minute checks plus the 5-minute hold	As above.

VPN alarms apply only to sites with a configured VPN (a direct VPN or a VPN peer); a configured VPN that is down raises VPN offline.

Silence clocks stop when a bigger cause is known

A unit's silence clock only runs while the rest of its data source still delivers. When the whole source or the link to the site is down, one event speaks for all of it — Source availability or VPN offline — instead of one per unit.

Data cadence (10-minute turbine rows)

Turbine data reaches Mirox through the turbine manufacturer's SCADA database as 10-minute rows: the row for 10:00–10:10 exists shortly after 10:10. With the database's own delay (about one minute) and our poll, the newest row is normally 1–11 minutes old. This is why nothing about a turbine can be known faster than about ten minutes — and why the wind thresholds are counted in rows. Battery units report every minute and solar components are judged in 5-minute windows, so their thresholds are counted in minutes.

Resolution notices

When the condition behind a notification ends — the unit produces again, the data source delivers again, the panel is reset — Mirox closes the event and sends a resolution notice on the same channels, so you know you can stand down. The original notification is marked resolved in your notification center.

An event that closes again during its hold (for example a 2-minute VPN blip) sends neither a notification nor a resolution notice.
One-off notices — a ticket comment, an action on the monitoring agent, an inverter replacement to confirm — have nothing to resolve and send none.

The Resolution notice column of the category tables below shows which events send one.

Categories

Each row on your notification settings is one category. The tables list every event that can notify in that category.

Production

Units that stop producing, safety alarms of wind turbines and battery storage, and implausible production. This is the category for every plant type: a solar inverter, a turbine and a battery unit that stop are reported the same way — as members of the one site notification.

Event	Level	Triggers when	Timing	Applies to	Resolution notice
Battery safety alarm `component_battery_safety_alarm`	Critical	A battery container or converter reports a danger: a fire alarm, smoke or gas detection, thermal runaway, or a released fire suppression. Checked every minute and sent at once.	Device signal: at once	Battery storage	yes
Turbine safety alarm `component_turbine_safety_alarm`	Critical	A wind turbine reports a danger: fire or smoke, or an arc in the converter or switchgear. Sent with the first data row that carries it (the SCADA data arrive every 10 minutes), also during a service visit.	Device signal: at once	Wind turbines	yes
Overproduction `overproduction`	High	The plant reports more than 125 % of what a clear sky allows, or more than 100 % for 30 minutes — almost always a frozen or mis-scaled meter, not real production.	Data says fault: at once	Solar (inverters, combiner boxes, strings)	yes
Component outage summary `summary_component_outage`	High	Units of the site are down: producing nothing, stopped by their own fault, or silent while the rest of the site reports. ONE notification per site situation, however many units join or leave. Solar: about 50 minutes after the first zero, 12 hours of silence. Wind turbines: two fault rows (about 20 minutes), 3 hours of silence. Battery storage: 15 minutes of fault, 30 minutes of silence.	Data says fault — Solar (inverters, combiner boxes, strings): data says fault 50 min, no data 12 h; Wind turbines: data says fault 20 min, no data 3 h; Battery storage: data says fault 15 min, no data 30 min	Solar (inverters, combiner boxes, strings), Wind turbines, Battery storage	yes
Combiner box production outage `component_gak_outage`	Normal	A combiner box produces nothing for three consecutive 5-minute checks although there is enough light — about 20 minutes. Joins the site notice after 30 more minutes.	Data says fault: 20 min	Solar (inverters, combiner boxes, strings)	yes
Inverter production outage `component_inverter_outage`	Normal	An inverter produces nothing for three consecutive 5-minute checks although there is enough light — about 20 minutes. One unit's state: by default you are told through the site notice ("components down") it joins after 30 more minutes.	Data says fault: 20 min	Solar (inverters, combiner boxes, strings)	yes
String Outage `component_string_outage`	Normal	A string's outage is confirmed: zero under an overcast sky with enough irradiance on the modules (about 20 minutes), or zero through the next day's high sun.	Investigated: 20 min	Solar (inverters, combiner boxes, strings)	yes
String Investigation `component_string_investigation`	Low	A string reads zero for about 20 minutes. The first, unconfirmed stage: shading, snow and soiling look the same at first, so this stays low until it is confirmed.	Investigated: 20 min	Solar (inverters, combiner boxes, strings)	yes

Availability

Data sources, the VPN, the site network and network devices that stop reporting.

Event	Level	Triggers when	Timing	Applies to	Resolution notice
Network Device Mass Outage `availability_network_devices_mass_outage`	High	More than 30 % of a site's network devices have each been unreachable for at least 15 minutes.	No data: 18 min	Network and VPN	yes
Park Network Offline `availability_park_network_offline`	High	The site's own network (its gateway and devices) failed three consecutive one-minute checks while the VPN itself is up. Held 5 minutes like the VPN notice.	No data: 3 min + 5 min hold	Network and VPN	yes
VPN Offline `availability_vpn_offline`	High	The monitoring VPN tunnel to the site has had no handshake for three consecutive one-minute checks. A further 5-minute hold drops short blips, so a real outage is reported about 8 to 11 minutes after it starts.	No data: 6 min + 5 min hold	Network and VPN	yes
Source Availability `source_availability`	High	A data source of the site (a data logger, a SCADA mirror, a battery controller) has delivered no data for 30 minutes — or for three of its own reporting intervals, if it reports less often.	No data: 30 min	Solar (inverters, combiner boxes, strings), Wind turbines, Battery storage, Irradiance sensors, Alarm system	yes
Network Device Offline `availability_network_device_offline`	Normal	A monitored network device (switch, router, logger) failed three consecutive polls — about 5 minutes. Not sent while the whole site is offline, and not for devices whose data source is already reported.	No data: 5 min 30 s	Network and VPN	yes

Control signals

Commands from outside the plant: grid-operator and direct-marketer curtailments. No event notifies in this category today. Curtailments are recorded as Detected curtailment episodes and settled (see Events); a stop the grid operator or the marketer orders is status, never a fault, so it does not count as a unit down either.

Sensors

Irradiance sensors that stop measuring plausibly — first an unconfirmed investigation, then a confirmed defect.

Event	Level	Triggers when	Timing	Applies to	Resolution notice
Sensor Defect `component_sensor_defect`	High	A sensor defect is confirmed: three hours of daylight readings contradicted by its siblings or by the plant's own production.	Investigated: 3 h	Irradiance sensors	yes
Sensor Investigation `component_sensor_investigation`	Low	An irradiance sensor reads dark for an hour of daylight while the plant produces. The first, unconfirmed stage.	Investigated: 1 h	Irradiance sensors	yes

Alarm system

The site's intrusion alarm panel and the supervision contacts of its communication cabinet — see Alarm System. The row has its own group on the settings page and starts at Very high, so by default only the intrusion alarm itself (Critical) notifies.

Event	Level	Triggers when	Timing	Applies to	Resolution notice
Alarm raised `alarm_triggered`	Critical	The site's intrusion alarm panel raises an alarm. Mirrored as the panel reports it — within about half a minute, without any delay of our own.	Device signal: 25 s	Alarm system	yes
Main switch off `alarm_cabinet_main_switch_off`	High	The main switch of the communication cabinet is off.	Device signal: 25 s	Site	yes
UPS alarm `alarm_cabinet_ups_alarm`	High	The uninterruptible power supply of the communication cabinet reports a fault — monitoring may be lost at the next mains outage.	Device signal: 25 s	Site	yes
Alarm panel fault `alarm_fault`	High	The alarm panel reports a fault of its own — intrusion detection may be impaired.	Device signal: 25 s	Alarm system	yes
Heating breaker tripped `alarm_cabinet_heating_breaker_tripped`	Normal	The circuit breaker of the cabinet heating has tripped.	Device signal: 25 s	Site	yes
UPS on battery `alarm_cabinet_ups_on_battery`	Normal	Mains power to the communication cabinet is lost and the UPS carries it.	Device signal: 25 s	Site	yes

Monitoring agent

Administrative actions on the site's monitoring agent: assignment, pause, restart, upgrade. Armed for every role but floored at Critical, so nothing arrives until you lower the row.

Event	Level	Triggers when	Timing	Applies to	Resolution notice
Agent assigned `agent_assigned`	Normal	An administrator assigned a monitoring agent to the site.	One-off	Monitoring agent	no
Agent operator changed `agent_operator_changed`	Normal	The site's monitoring agent moved to another operator.	One-off	Monitoring agent	no
Agent paused `agent_paused`	Normal	The site's monitoring agent was paused — no data is collected until it resumes.	One-off	Monitoring agent	no
Agent moved to another node (rebalance) `agent_rebalanced_local`	Normal	The site's monitoring agent was rebalanced within its cluster.	One-off	Monitoring agent	no
Agent moved to another region (rebalance) `agent_rebalanced_regional`	Normal	The site's monitoring agent was rebalanced to another region.	One-off	Monitoring agent	no
Agent restarted `agent_restarted`	Normal	The site's monitoring agent was restarted.	One-off	Monitoring agent	no
Agent moved to another operator `agent_shifted`	Normal	The site's monitoring agent moved to another operator together with the site's VPN service.	One-off	Monitoring agent	no
Agent unassigned `agent_unassigned`	Normal	An administrator removed the site's monitoring agent.	One-off	Monitoring agent	no
Agent upgraded `agent_upgraded`	Normal	The site's monitoring agent was upgraded to a new version.	One-off	Monitoring agent	no
Agent deployment spec updated `agent_deploy_spec_updated`	Low	The deployment settings of the site's monitoring agent were changed.	One-off	Monitoring agent	no
Legacy agent config removed `agent_legacy_config_deleted`	Low	The manual configuration override of the site's monitoring agent was removed.	One-off	Monitoring agent	no
Legacy agent config set `agent_legacy_config_set`	Low	A manual configuration override was set on the site's monitoring agent.	One-off	Monitoring agent	no
Agent resumed `agent_resumed`	Low	The site's monitoring agent was resumed.	One-off	Monitoring agent	no

Components

Changes to the site's equipment that need your confirmation — for example an inverter that looks replaced.

Event	Level	Triggers when	Timing	Applies to	Resolution notice
Inverter replacement detected `comp_solar_inverter_replacement_detected`	High	An inverter has been silent for 48 hours while a new unit reports on the same logger — most likely a swap. One notice asks you to confirm it so the new unit is analysed.	One-off: 2 d	Solar (inverters, combiner boxes, strings)	no

Tickets

Tickets you create, are assigned to, comment on or are mentioned in. Mentions and assignments always reach you, whatever your settings.

Event	Level	Triggers when	Timing	Applies to	Resolution notice
Ticket assigned `ticket_assigned`	High	A ticket was assigned to you.	One-off	Site	no
Mentioned in a ticket `ticket_mention`	High	Someone mentioned you in a ticket. Always delivered, whatever your settings.	One-off	Site	no
Ticket unassigned `ticket_unassigned`	High	A ticket was taken off you.	One-off	Site	no
Ticket closed `ticket_closed`	Low	A ticket you take part in was closed.	One-off	Site	no
Ticket reopened `ticket_reopened`	Low	A ticket you take part in was reopened.	One-off	Site	no
Ticket updated `ticket_updated`	Low	The status, assignee or category of a ticket you take part in changed.	One-off	Site	no
Ticket comment added `ticket_comment_created`	Very low	Someone commented on a ticket you take part in.	One-off	Site	no

Organization

Members joining or leaving, invitations, role changes and cooperations with other organizations. These notices concern your organization, not a plant, so they carry no detection timing; your role decides which of them are on by default.

Wind turbines

Mirox judges each turbine from its own 10-minute data and notifies on the judged state, not on raw manufacturer codes:

Stopped by a fault (or a protection): the turbine reports a fault stop. The unit event opens on the first such row; the site notification "components down" follows after two consecutive rows.
Not communicating: the turbine's rows stop while the other turbines of the site still deliver — after 30 minutes for the unit, 3 hours for the site notification (12 hours after a service visit or a commanded stop).
Safety alarm (Critical): fire or smoke, or an arc in the converter or the switchgear. Sent with the first row that carries it, also during a service visit.

Stops that are not faults. A turbine stands still for many reasons that are perfectly in order. These are recorded as status — they open no event, never notify and only pause the clock that watches for silence:

a remote pause from the turbine's control system;
environmental curtailment: bat protection, shadow flicker, ice, noise reduction;
a stop ordered by the direct marketer or the grid operator;
a pause pressed on the turbine's keyboard or a service stop by a technician;
automatic routines such as yaw untwisting.

Manufacturer codes are detail. The turbine's own alarm code and text travel with the event, so you can see why it stopped — but a code alone never notifies. Plain manufacturer warnings stay visible on the turbine's alarm list on the wind plant pages and do not notify.

Battery storage

Battery units — the power converters (PCS) and the battery containers — are judged the same way, from the status word they report every minute:

Stopped by a fault: the status word says fault for 5 consecutive minutes; the site notification follows at 15 minutes.
Not communicating: one unit's data stops while its data source still delivers — 15 minutes for the unit, 30 minutes for the site notification.
Device fault while running (Normal, event only): the unit reports a fault of its own but keeps operating — a service item, not a stop.
Safety alarm (Critical): a fire alarm, smoke or gas detection, thermal runaway, a released fire suppression or an arc. Checked every minute and sent at once. A fault of the fire-detection equipment itself (a sensor or panel error) is a device fault, not a safety alarm.

Idle, standby and stops that someone ordered are status, never faults.

Storage on solar and wind plants. These checks run wherever battery hardware is present — not only on plants of type Battery. A solar plant with a storage system gets the same battery alarms, and a hybrid site still raises one site notification that lists inverters and battery units together.

Battery events appear under the Battery category in the event list, but they notify through the Production row of your settings — there is no separate battery row to configure. See Battery Storage.

Examples by level

Real, anonymised cases per level, newest first.

Level	Event	Example
Critical	Battery safety alarm `component_battery_safety_alarm`	Typical case: a rack's gas detector trips stage 1 in a storage container; the notice goes out within two minutes.
Critical	Turbine safety alarm `component_turbine_safety_alarm`	A turbine reported arc light in its converter during a service visit; the safety notice would have gone out with the first data row, about 10 minutes later.
Critical	Alarm raised `alarm_triggered`	Typical case: a site's intrusion alarm trips at night; the notification goes out about 25 seconds later and clears when the panel is reset.
High	Overproduction `overproduction`	Typical case: a meter keeps reporting its midday value after its clock stopped; the notice clears when the reading updates.
High	Component outage summary `summary_component_outage`	An inverter outage at sunrise rolled up into the site notice at 07:46; it closed at 10:46 when the last unit recovered.
High	Network Device Mass Outage `availability_network_devices_mass_outage`	Sixteen of nineteen devices on a hybrid PV and storage site went dark one evening; the notice stood for 15 hours and cleared the next morning.
High	Park Network Offline `availability_park_network_offline`	A site's network was lost at 09:23 and restored at 12:36. Most such interruptions last about 4 minutes and are absorbed by the hold.
High	VPN Offline `availability_vpn_offline`	A site's VPN dropped at 22:59 and the notification followed a few minutes later; it reconnected at 10:01 the next morning. Several 2-minute blips that month notified nobody.
High	Source Availability `source_availability`	A logger delivered its last reading at 11:28; the notification went out at 11:58 and the all-clear at 14:08 when data returned.
High	Sensor Defect `component_sensor_defect`	An irradiance sensor confirmed defective at 13:30 kept the event open until 14:40 the next day.
High	Main switch off `alarm_cabinet_main_switch_off`	On the same storage site the cabinet main switch was off for 21 hours after a service visit.
High	UPS alarm `alarm_cabinet_ups_alarm`	A cabinet UPS raised an alarm at 07:40, reported 30 seconds later; it stayed open for days until the battery was replaced.
High	Alarm panel fault `alarm_fault`	Typical case: the panel loses a detector line and reports a fault until a technician repairs it.
High	Inverter replacement detected `comp_solar_inverter_replacement_detected`	An inverter silent for more than two days was matched to a newly reporting unit on the same logger; the operator confirmed the swap the same day.
High	Ticket assigned `ticket_assigned`	A ticket was opened 19 minutes after a battery over-temperature fault and assigned to the site's technician.
High	Mentioned in a ticket `ticket_mention`	A colleague asked you to check a battery fault.
High	Ticket unassigned `ticket_unassigned`	A ticket was reassigned to a colleague.
Normal	Combiner box production outage `component_gak_outage`	A combiner box read zero from 12:50; the event opened at 13:10 and closed at 16:10.
Normal	Inverter production outage `component_inverter_outage`	An inverter dropped to zero at 11:20 in good light; the event opened at 11:40 and closed at 14:20.
Normal	String Outage `component_string_outage`	A string confirmed out at 10:15 closed at 14:15 when it produced again.
Normal	Network Device Offline `availability_network_device_offline`	A switch stopped answering at 12:35; the event opened about 5 minutes later and closed at 15:26.
Normal	Heating breaker tripped `alarm_cabinet_heating_breaker_tripped`	A storage site's cabinet heating breaker tripped at 19:45 together with the main switch; both were restored at 17:04 the next day.
Normal	UPS on battery `alarm_cabinet_ups_on_battery`	Typical case: a short grid interruption at the site puts the cabinet on battery for a few minutes.
Normal	Agent assigned `agent_assigned`	A new site got its monitoring agent assigned during onboarding.
Normal	Agent operator changed `agent_operator_changed`	A site's agent was moved to another cluster operator.
Normal	Agent paused `agent_paused`	An agent was paused during a site's network rebuild.
Normal	Agent moved to another node (rebalance) `agent_rebalanced_local`	An agent was rebalanced to another node at 03:12.
Normal	Agent moved to another region (rebalance) `agent_rebalanced_regional`	An agent moved to the second data centre during maintenance.
Normal	Agent restarted `agent_restarted`	An agent was restarted to apply a configuration change.
Normal	Agent moved to another operator `agent_shifted`	A site's VPN service was reassigned to another operator, and its agent moved with it.
Normal	Agent unassigned `agent_unassigned`	A decommissioned site's agent was removed.
Normal	Agent upgraded `agent_upgraded`	An agent was upgraded during the weekly roll-out.
Low	String Investigation `component_string_investigation`	A string went dark at 07:10; the investigation opened at 07:30 and closed at 10:05 when it produced again.
Low	Sensor Investigation `component_sensor_investigation`	A pyranometer went dark at 06:15; the investigation opened at 07:20 and closed at 09:20 when readings returned.
Low	Agent deployment spec updated `agent_deploy_spec_updated`	An agent's resources were raised after a site grew.
Low	Legacy agent config removed `agent_legacy_config_deleted`	The legacy override was removed after the migration.
Low	Legacy agent config set `agent_legacy_config_set`	A legacy configuration was applied during a migration.
Low	Agent resumed `agent_resumed`	The agent resumed after the network rebuild.
Low	Ticket closed `ticket_closed`	A ticket about an inverter fault was closed after the repair.
Low	Ticket reopened `ticket_reopened`	A ticket was reopened when the fault came back.
Low	Ticket updated `ticket_updated`	A ticket moved to "in progress" when a technician took it.
Very low	Ticket comment added `ticket_comment_created`	A technician posted photos from the site visit.

Notifications — choose channels and a minimum level per category
Events — every event family the platform records, notified or not
Monitor page — how events drive a site's state
Inverter Events — the solar component events in detail
Wind Plants · Battery Storage · Alarm System