Alerts
Alerts the team handles together — lists by status, alert detail, silences, rules and channels.
Required permission: View
An alert is created when a device disconnects, drift is detected or a Prometheus rule fires. Unlike the inbox, which goes to individuals, alerts are shared by the team. When someone acknowledges or resolves one, everyone sees the change. The number next to Alerts (경보) in the sidebar is the count of active alerts (refreshed every 30 seconds; 100 or more shows as 99+).
Every member can view; acknowledging, resolving and silencing need operate (OPERATOR, ADMIN), and managing rules and channels needs settings (ADMIN).
Alert list
| # | Area | Description |
|---|---|---|
| ① | Tabs | Active (활성) · Acknowledged (확인됨) · Resolved (해결됨) are lists by status; Silence (음소거) · Rules (룰) · Channels (채널) are settings. |
| ② | Filters | Severity, range (All time (전체 기간), 24 hours, 7 days, 30 days, by firing time), source (Alertmanager, edge sweep, drift), Rule name (룰 이름) (exact match), Search titles (제목 검색) (partial match). Filters stay in the address. |
| ③ | Alert | Severity, title, source, number of grouped occurrences (N건 그룹), firing time and who acknowledged it. Newest first, refreshed every 30 seconds. Click the title to open the detail. |
| ④ | Acknowledge (확인) | Marks that the person in charge is aware of the alert. Active → Acknowledged (operate). Shown only on active alerts. |
| ⑤ | Resolve (해결) | Handling finished. Active or Acknowledged → Resolved (operate). |
When the same alert (same fingerprint) fires again while still active, the grouped count goes up instead of adding a row. If it fires again after being acknowledged, a new alert is created.
Alert detail
| # | Area | Description |
|---|---|---|
| ① | Title | Severity, alert title, status. The Detail (상세) card below shows the source, rule and group count. |
| ② | Actions | Silence (음소거) (pause alerts from the same rule for a while; only on unresolved alerts that have a rule), Acknowledge (확인) (when active), Resolve (해결) (when unresolved). All need operate. |
| ③ | Related links | Links to the device, model or deployment the alert points to. The model link of a drift alert goes to the Drift screen; other model links go to the model registry. Alertmanager alerts also get a Prometheus graph (Prometheus 그래프) link (new tab). If the target feature is off, the link shows as plain text. Device alerts also show the device's current state at the top as Device now (장치 지금). |
| ④ | Handling history (처리 이력) | Who raised → acknowledged → resolved it, and when. For an alert that cleared by itself, the resolver is "System (auto-resolved) (시스템(자동 해소))". |
Silences, rules and channels
A silence (음소거) keeps Alertmanager alerts matching a condition (label=value) off this screen for a set time. The label defaults to alertname, and the duration is in hours (default 1, at most 720). Pressing Silence (음소거) on an alert detail opens this tab with the rule name filled in as the value. Enter the value and a reason (사유) (required) and press Silence (음소거) (operate). The server adds a condition for the current tenant, so other tenants' alerts are not silenced, and the list shows only silences this tenant created. A silence's state is active, pending or expired; lift it at once with the Lift (해제) button (a confirmation dialog appears).
The Rules (룰) tab stores alert rules (rule name, PromQL expression, severity), and the Channels (채널) tab stores notification channels (type Slack, Email, Webhook or SMS, and a target such as a webhook URL or email address). Adding, deleting and Test (테스트) need settings (ADMIN); other roles only see the lists. A channel whose Test (테스트) send succeeded is marked Verified (검증됨).
Statuses and badges
| Status | Value | Meaning |
|---|---|---|
| Active (활성) | ACTIVE | Newly fired and nobody has touched it. Counted in the sidebar number. |
| Acknowledged (확인됨) | ACKED | Someone acknowledged it (being handled). |
| Resolved (해결됨) | RESOLVED | Handling finished, or it cleared automatically because the cause went away (for example a disconnected device sent a heartbeat again). |
| Severity | Meaning |
|---|---|
| CRITICAL | Critical — act now |
| WARNING | Warning |
| INFO | Information |
Source (grey text in the list): alertmanager (Prometheus rules), edge-sweep (every minute, finds devices whose last heartbeat is older than 180 seconds (default) and raises a WARNING alert), drift (drift verdict).