> ## Documentation Index > Fetch the complete documentation index at: https://docs.ilyama.golain.io/llms.txt > Use this file to discover all available pages before exploring further. # Alarms in the Console > Work the alarm board, write and test alarm definitions, and set up silences and inhibitions from the Golain Console. Alarms in the Console are project-scoped. Open a project, then use **Alarms** in the sidebar: | Sidebar item | What it shows | | - | - | | **Board** | What is firing right now. Acknowledge, shelve, and unshelve alarms here. | | **Definitions** | The rules that open alarms. Create, edit, test, activate, deactivate, and delete them. | | **Silences** | Silences and inhibitions that hold notifications back. | For the concepts behind these screens (kinds, lifecycle, priority), see [Alarms overview](/alarms/overview.md). For a guided walkthrough, see [Your first alarm](/alarms/first-alarm.md). ## Board The board header reads **Alarms** and links to **Definitions** and **Silences**. ### Summary Above the list, the board counts the alarms on the current page: * **open**, **unacknowledged**, and **shelved** totals. The **unacknowledged** count turns red when it is above zero. * A small donut and legend splitting open alarms by priority: **P1**, **P2**, **P3**, **P4**. > [!NOTE] > > The summary counts the rows loaded on the current page, not the whole project. ### Search and filters Use **Search alarms, assets or labels** to filter the loaded rows by definition name, key, labels, or asset. The filter pills toggle on and off. Click an active pill again to clear it. | Pill | Shows | | - | - | | **Unacked** | Alarms nobody has acknowledged yet | | **Acked** | Alarms someone has acknowledged | | **Latched** | Alarms whose condition cleared but stay open until acknowledged | | **Shelved** | Alarms currently shelved | | **History** | Closed alarms. The summary is hidden and the **Since** column becomes **Opened**. | ### Columns | Column | What it shows | | - | - | | **Pri** | Priority chip: `P1` (**Critical**), `P2` (**High**), `P3` (**Medium**), `P4` (**Low**) | | **Alarm** | The instance title (built from its labels), the definition name and key, and up to three label chips. Click the title to open the instance. | | **Condition · Ack** | Condition badge and acknowledgement badge, with a one-line hint underneath | | **Trigger** | The value that fired the alarm. For an offline-device alarm: `offline > {window}` and how long it has been open. | | **Since** | How long the alarm has been open, plus a status line such as **Unacked · subscribers notified** | | *(actions)* | **Shelve** or **Unshelve**, and **Acknowledge** when the alarm needs it | Rows are sorted **P1 first, then oldest unacknowledged**. ### State badges Each row shows two independent things: whether the condition is true, and whether someone has acknowledged it. | Badge | Meaning | | - | - | | **Pending** | The condition is true but has not held long enough to fire | | **Firing** | The condition held; the alarm is active | | **Latched** | The condition cleared, but the alarm stays open until acknowledged | | **Resolved** | The condition cleared | | **Unacked** | Nobody has acknowledged it (outlined in red). Not shown while **Pending**. | | **Acked** | Acknowledged, followed by the person's name when known | | **Shelved** | Shelved. Replaces the ack badge until the shelf expires. | The hint under the badges explains the state, for example **holds until acknowledged** for a latched alarm or **until {time}** for a shelved one. ### Actions | Button | When it appears | What it does | | - | - | - | | **Acknowledge** | The alarm is firing or latched and unacknowledged | Marks it acknowledged | | **Shelve** | The alarm is open and not shelved | Opens the shelve dialog | | **Unshelve** | The alarm is shelved | Removes the shelf immediately | Closed alarms show no actions. The board has no bulk actions; act on one row at a time. ### Refresh The board refreshes itself every 10 seconds while the tab is in focus. The footer shows **Live · last transition … ago** and when the board last **checked**. Use **Previous** and **Next** to page through 50 alarms at a time. ## Instance detail Click an alarm title on the board to open its detail page. Use **Alarms** at the top to go back. ### Header * Priority chip with its label (for example **P1 Critical**), and a link to the definition. * The instance title. * The state badges and a headline such as **Firing for 14 min · unacknowledged · subscribers notified 2×**, or **Shelved until …**, or **Closed … ago**. * **Shelve** / **Unshelve** and **Acknowledge**, following the same rules as the board. ### Condition | Field | What it shows | | - | - | | **Trigger value** (or **Offline for** for offline-device alarms) | The value that fired, and when the alarm was **last evaluated** | | **Labels** | Every label on the instance. Device and asset IDs are shown by name when the Console can resolve them. | | **Fires when** | The definition's condition and, for a query-based rule, its SQL. Offline-device rules show a plain sentence instead. | Below **Fires when**, a timing line summarizes the rule, for example `held 1 min before firing · keeps firing 2 min after · latching`. ### Guidance | Row | What it shows | | - | - | | **Consequence** | What happens if nobody acts, with the safety class chip if set. **Not written down.** if empty. | | **Response procedure** | The response steps, one per line | | **Work orders** | Linked work-order IDs, or **None linked.** When the definition creates work orders automatically, a link to the ticketing board appears. See [Ticketing](/ticketing/overview.md). | ### Transition history **Transition history** lists the instance's milestones, newest first. Each entry shows the time, how long ago, a title, and who acted where known. | Entry | When it appears | | - | - | | **Opened · pending** or **Opened · firing** | When the instance opened, with its trigger value | | **Pending → firing** | The condition held for the configured duration | | **Acknowledged** | Someone acknowledged it | | **Shelved until …** | Someone shelved it, with the reason they gave | | **Condition cleared · still firing** | The condition cleared, but the alarm is waiting out its keep-firing time | | **Resolved** or **Condition cleared · latched** | The condition cleared. A latched alarm **Stays open until acknowledged.** | | **Closed** | The instance closed | > [!NOTE] > > The history is rebuilt from the instance's own timestamps, so each kind of milestone appears once. Repeated reminders show as a count in the headline, not as separate entries. ### Shelve dialog **Shelve** opens **Shelve this alarm**. Shelving keeps the alarm open but stops reminders and re-evaluation until the shelf expires. It comes back on its own. | Field | Details | | - | - | | **Shelve for** | Presets **30 min**, **1 h**, **4 h**, **End of shift**, or **Custom** (minutes). **End of shift** runs to the next 06:00, 14:00, or 22:00 local time. | | **Expires** | The time the shelf ends | | **This alarm allows up to …** | The definition's **Max shelve** limit (24 h by default). Presets longer than the limit are hidden, and a longer custom value blocks the shelve button. | | **Reason** | Optional. Recorded on the instance with your name and visible to everyone on the board. | Finish with **Shelve for {duration}**, or with **Acknowledge and shelve** if the alarm is still unacknowledged. > [!WARNING] > > Shelving does not acknowledge. An unacknowledged alarm stays counted as unacknowledged on the board. Use **Acknowledge and shelve** to do both. ## Definitions list **Definitions** opens **Alarm definitions**. The header has **Guided setup**, **Board**, and **New definition**. ### Search and filters * **Search by key** filters definitions on the server. * **Active** (on by default) and **Inactive** pills filter by status. Click the active pill again to show both. ### Columns | Column | What it shows | | - | - | | **Definition** | Priority chip, name (links to the editor), key, and an **INACTIVE** badge when deactivated | | **Kind · cadence** | `event`, `threshold`, or `staleness`, a `latching` chip if set, and the cadence, for example `every 30s · for 60s`, `offline 20m · checked every minute`, or `on device event` | | **Scope** | **Project**, **Asset type**, **Single asset**, or **Every device** (offline-device rules), plus `one alarm per {columns}` or `one alarm per rule` | | **Subscribers** | How many notification subscriptions listen to this alarm and on which channels, or **None** / `nobody is told` | | **Status** | **Evaluating** (scheduled rules), **Listening** (event rules), or **Deactivated** | | **Open** | Open instances, summarized as `N unacked`, `N firing`, `N pending`, or `N shelved` | The list has no inline toggle or delete. Open a definition to activate, deactivate, or delete it (see [Activate, deactivate, and delete](#activate-deactivate-and-delete)). ## Create a definition **1. Pick a starting point** Click **New definition**. The **New alarm** dialog offers four starting points: | Option | Kind | What it sets up | | - | - | - | | **A device goes offline** | `staleness` | One alarm per MQTT device the broker has seen disconnected for longer than a window (default 20 min). Resolves when the device reconnects. | | **Telemetry stops arriving** | `threshold` | One alarm per device when a dataspace receives no rows from it for longer than a window. The Console generates the SQL for you. | | **A value crosses a limit** | `threshold` | Query a dataspace on a schedule and fire when a row breaks a condition | | **Start blank** | `event` | Write the SQL and CEL yourself | Click **Continue** (or double-click an option). Everything is editable afterwards. **2. Fill in the editor** Complete the sections described below. The right-hand rail previews how the rule behaves as you type. **3. Test it** Click **Test** to run the query and check notifications without saving. See [Test a definition](#test-a-definition). **4. Save** Click **Save and start watching**. The definition starts evaluating as soon as you save. > [!TIP] > > Notifications are not set in this dialog. After you save, subscribe a channel to the alarm under Integrations. See [Notifications in the editor](#notifications-in-the-editor). ## Editor fields The editor heading is **New alarm** or **Edit definition**, with the kind as a chip. The form shows only the fields the chosen kind needs. ### Identity The card is titled **What to call it** when creating and **Identity** when editing. | Label | API field | Meaning | | - | - | - | | **Name** | `name` | Display name. Required. | | **Key** | `key` | Unique in the project. Lowercase letters, digits, dots, dashes, and underscores. Generated from the name until you edit it. **Locked** after creation. | | **Description** *(optional)* | `description` | Free text | | **Priority** | `priority` | **Critical** (1), **High** (2), **Medium** (3, default), **Low** (4). Drives sort order on the board. | | **Scope** | `asset_type_id` / `asset_id` | **Whole project**, **Asset type** (enter an **Asset type id**), or **Single asset** (enter an **Asset id**). Not shown for offline-device rules. | ### What to watch Pick the kind with the **event**, **threshold**, and **staleness** tabs at the top of the card. You can change the kind only while creating. **Threshold** | Label | API field | Meaning | | - | - | - | | **Dataspace** | `source_dataspace_id` | The dataspace the query reads. Required. Shows the read view name once chosen. | | **Query** | `query_spec.sql` | SQL against the read view. Return one row per labelset, and bound `event_ts` with a relative window of 24 h or less (for example `WHERE event_ts >= now() - INTERVAL 1 HOUR`). | | **Fires when** | `condition_cel` | CEL over each row, as `event.`. Example: `event.temp_c > 235.0`. | | **Resolves when** *(optional)* | `resolve_cel` | Defaults to **Fires when** going false. Set it lower to add a deadband. | | **One alarm per** | `labels_by` | Columns whose values identify an instance. Leave empty for one alarm per rule. | With **Telemetry stops arriving**, the card shows a simplified form instead: **Dataspace**, **One alarm per** (the device column), **Fire when quiet for** (minutes), and **When rows return**. The generated SQL appears under **Generated query**. Click **Edit the SQL by hand** to switch to the full threshold form. See [Kinds and conditions](/alarms/kinds-and-conditions.md) for query and CEL rules. **Staleness** | Label | API field | Meaning | | - | - | - | | **Fire when offline for** | `staleness_window_seconds` | Entered in minutes, sent in seconds. Each MQTT device the broker has seen disconnected for longer than this opens its own alarm. A device that has never connected counts as offline. | | **One alarm per** | — | Always `device_id` | | **When the device reconnects** | `latching` | **Resolve on its own** or **Stay open until acknowledged** | > [!NOTE] > > LoRaWAN and HTTP devices never connect to the broker. Use **Telemetry stops arriving** for those. **Event** | Label | API field | Meaning | | - | - | - | | **Fires when** | `condition_cel` | CEL over the changed dataspace row, as `event.`. `asset` and `now` are also in scope. Example: `event.estop == true`. | | **Clears when** *(optional)* | `clear_cel` | Defaults to **Fires when** going false | | **On delay** | `on_delay_seconds` | Held this long before firing (seconds) | | **Off delay** | `off_delay_seconds` | Clear this long before resolving (seconds) | | **Deadband** *(optional)* | `deadband` | Value must fall this far below the limit to clear | | **After it clears** | `latching` | **Resolve on its own** or **Stay open until acknowledged** | ### Timing The **Timing** card is hidden in the simplified **Telemetry stops arriving** form. | Label | API field | Kinds | Meaning | | - | - | - | - | | **Evaluate every** | `eval_interval_seconds` | threshold | How often the query runs (seconds, default 60) | | **Must hold for** | `for_seconds` | threshold | Pending until the condition has held this long | | **Keep firing for** | `keep_firing_for_seconds` | threshold | After the condition clears, before it resolves | | **Re-alarm after** | `re_alarm_delay_seconds` | all | A resolved alarm won't re-open sooner than this (default 300 s). Stops flapping. | | **Re-notify every** | `repeat_interval_seconds` | all | Remind subscribers at this interval while unacknowledged. `0` sends once. | | **When a labelset stops producing rows** | `on_no_data` | threshold | **Keep firing** (`keep_firing`), **Resolve** (`resolve`), or **Fire** (`alert`) | | **After it clears** | `latching` | threshold | **Resolve on its own** or **Stay open until acknowledged** | For offline-device rules, the card explains that devices are checked every minute against the broker's connection state. ### Operator guidance Titled **What the operator should do** when creating and **Operator guidance** when editing. All fields are optional and shown on every instance. | Label | API field | Meaning | | - | - | - | | **Consequence** | `consequence` | What happens if nobody acts | | **Response** | `response_procedure` | One step per line | | **Safety class** | `safety_classification` | **None**, `process`, `safety`, `environmental`, or `quality` | | **Max shelve** | `shelve_max_hours` | How long one shelf may last, in hours (default 24) | ### Work order on activation Turn the **On** / **Off** switch on to raise a work order the moment an instance fires. | Label | API field | Meaning | | - | - | - | | *(switch)* | `auto_create_wo` | Create a work order when an instance fires | | **Workflow** | `wo_workflow_def_id` | The ticket workflow the work order opens in | | **Dedup key** *(optional)* | `wo_dedup_key_cel` | CEL. The same key inside the cooldown reuses the open work order. See [Automatic work orders](/alarms/notifications.md#automatic-work-orders) for the variables it can read. | See [Notifications](/alarms/notifications.md) for how work orders relate to alarm events. ## Preview rail The right-hand rail updates as you edit. ### If this were on right now Shown for threshold rules once you pick a dataspace and bound the query. The Console runs the draft SQL and shows how many rows come back, with the first few listed: * In the simplified **Telemetry stops arriving** form, the count reads **devices would fire**, with how long each has been quiet. * For a hand-written threshold rule, it reads **rows to evaluate**. The condition itself is not run here, so the count is the rows the condition will be checked against. Until the query is ready, the card reads **Pick a dataspace and bound the query to see what it returns today.** ### How it behaves A short step-by-step sketch of what the rule does to one spike, built from your timing settings. For example: pending for the hold time, then fires, then reminders (or **Subscribers are told once; nobody is reminded.**), then resolve or latch, then the re-alarm delay. ## Test a definition Click **Test** in the editor header. The test dialog, titled **Test “{name}”**, runs checks without saving anything and without opening an alarm. Close it with **Close**. | Section | Shown for | What you see | | - | - | - | | **Query** | Rules with SQL | Runs the SQL and shows up to 20 rows. Badges: **not bounded**, **running…**, **failed**, or the row count. Use **Run** / **Run again** to repeat. | | **Condition** | Threshold rules | See the note below | | **Devices that would fire** | Offline-device rules | MQTT devices offline longer than the window, as `{n} of {total} MQTT devices`, each marked **never connected** or **offline {duration}** | | **Condition** | Event rules | Explains that event rules fire from device events and can't be dry-run. Send a matching event from a device and watch the board. | | **Notifications** | All | Subscriptions that listen to this alarm, with **Send test** per subscription | > [!WARNING] > > The condition dry-run is not available on the platform yet. For threshold rules, the **Condition** section shows a **not available** badge and the message "This platform cannot dry-run the condition yet". Use the **Query** section to see the rows the condition will be checked against, then save the rule and watch the board. ### Test notifications The **Notifications** section works only for a saved definition. Before you save, it reads **Save the rule first — subscriptions target a saved definition.** For a saved definition: 1. Find the subscription in the list. Each row shows the channel (for example `Slack`, `email`, `SMS`), the destination, whether it covers **every alarm** or **this alarm**, and how many events it subscribes to. 2. Click **Send test**. The Console sends a test message through the channel and shows **delivered**, or the provider's error. 3. Click **Send again** to repeat. Disabled subscriptions can't be tested. A test is a generic message, not an alarm payload: it proves the route, not the template. The platform waits up to 10 s for the provider's answer. Use **Add a subscription** to create one under Integrations. ## Notifications in the editor The editor's **Notifications** card is read-only. Who gets told, and where, is set up under Integrations; the definition only decides when it fires. * While creating, it reads **Nobody is subscribed yet. It will still show on the alarm board.** * When editing, it lists the subscriptions that listen to this alarm, for example `2 subscriptions listen to this alarm — Slack #ops, email oncall@example.com.` A subscription counts if it is enabled for alarm events and either targets this definition or targets every alarm. * **Open Notifications** (shown when editing) opens the notification settings under Integrations for this definition. See [Notifications](/alarms/notifications.md) for subscriptions, re-notify, and escalation. ## Activate, deactivate, and delete Open a definition from the list. The editor header has: | Button | What it does | | - | - | | **Save definition** | Saves changes. They apply on the next evaluation tick; open instances keep their state. | | **Deactivate** / **Activate** | Stops or resumes evaluation. Activating also clears an automatic deactivation caused by slow queries. | | **Delete** | Asks **Delete this definition?** Confirm with **Delete definition**. The rule stops evaluating. | | **Test** | Opens the test dialog | ## Silences **Silences** opens **Silences and inhibitions**. Both hold notifications back; the alarms stay open and still show on the board. The header has **Board**, **New inhibition**, and **New silence**. ### Silences list Each silence shows its matchers, **Window**, **Reason** (with creator and `approved by …` or `not yet approved`), and **State**: | State | Meaning | | - | - | | **Silencing** | Enabled and inside its window | | **Scheduled** | Enabled; the window has not started | | **Expired** | The window has ended | | **Disabled** | Turned off | Each row has **Edit**, **Disable** / **Enable**, and **Delete**. Deleting asks for confirmation; notifications it was holding back resume on the next evaluation. ### Create a silence **1. Open the form** Click **New silence**. **2. Add matchers** Under **Matchers**, enter a label name, an operator, and a value. Click **Add matcher** for more. All matchers must match. **3. Set the window** Set **Starts** and **Ends**. The defaults are now and four hours from now. The window must end after it starts. **4. Give a reason** Enter a **Reason** (required). Optionally fill **Approved by**. **5. Save** Click **Create silence**. ### Matchers | Operator | Matches when the label value… | | - | - | | `=` | equals the value | | `!=` | does not equal the value | | `=~` | matches the regular expression (fully anchored) | | `!~` | does not match the regular expression (fully anchored) | Matchers compare against the instance's **labels** — the `labels_by` columns of a threshold rule, or `device_id` for an offline-device rule. The label name field also suggests names such as `alarm`, `priority`, and `kind`, but instances do not carry those labels, so a matcher on them never matches. Open an instance and check its **Labels** before you write a matcher. A missing label counts as empty: `site=""` matches alarms with no site, and `site!=pune-2` matches them too. ### Inhibitions An inhibition mutes target alarms while a source alarm fires. Each inhibition shows **While** (source matchers and reason), **Mute** (target matchers), **When these match** (equal labels, or **any**), and **Armed** or **Disabled**. Rows have the same **Edit**, **Disable** / **Enable**, and **Delete** actions. To create one, click **New inhibition** and fill in: | Field | Details | | - | - | | **While a source fires** | Matchers for the source alarm. Required. | | **Mute targets** | Matchers for the alarms to mute. Required. The form pre-fills `kind=staleness`; replace it with labels your instances actually carry. | | **When these labels match** | Comma-separated labels that must be equal on source and target, for example `gateway_id, site`. Empty means any source mutes any target. | | **Reason** *(optional)* | Free text | Click **Create inhibition**. See [Silences and shelving](/alarms/silences-and-shelving.md) for how silences, inhibitions, and shelving differ. ## Next steps > **[Kinds and conditions](/alarms/kinds-and-conditions.md)** > > Write SQL queries and CEL conditions for each kind. > **[Notifications](/alarms/notifications.md)** > > Subscribe channels to alarm events. > **[API reference](/alarms/api-reference.md)** > > Do everything on this page over HTTP. > **[Troubleshooting](/alarms/troubleshooting.md)** > > Fix alarms that don't fire, resolve, or notify. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.