See everything that needs you, and what happened before

DedNets watches several kinds of things for you: your hosts, the services they export, your watchdogs, uptime monitors and signals. Whenever any of them is in trouble it becomes an alert, and every alert lands in one place, ranked worst first. (The Alerts page and alert history since v0.15.0.)

An alert is one of:

Source	Becomes an alert when	Severity
Host	the machine is offline	critical
Service	its health probe fails (unhealthy)	critical
Watchdog	it is down (late past its grace) / late	critical / warning
Uptime monitor	its target is down / its probing host is offline	critical / warning
Signal	it is open	its own severity (info signals are left out)

The “Needs you now” panel

On the Overview, the right-hand panel shows what needs you right now:

• a bar split by severity, with the counts (“3 critical · 6 warning”);
• the five worst alerts. Each row shows what it is, where it lives and, for a signal reported again and again, how many times (signal · disk-full-web1 · ×14). Click a row to open it in place and jump to the page that owns it;
• Last 24 h: one bar per hour, as tall as the alerts opened in it and coloured by the worst of them, with how many opened and resolved;
• +N more and View all alerts when there are more than five.

The Alerts page

Open Alerts in the sidebar, under Overview. Its count is the number of open alerts, and it turns red while any of them is critical.

• Open alerts - every open alert, not just the top five, grouped by severity. Click one to see its details and open it on its own page.
• Figures for the selected range: alerts opened, resolved, the median time from opened to resolved, and the noisiest source.
• Last 30 days - one square per day, coloured by the worst alert opened that day (grey for a quiet day). Hover a square, or tap it on a phone, to see the day, how many alerts opened and the worst one.
• History - every change of state, newest first and grouped by day: Opened, Escalated (a signal reported again at a higher severity), Resolved, Paused and Resumed (a watchdog you paused or resumed), and Deferred (since v0.16.0): a signal opened inside one of its quiet windows, so its notification was held back.
  • Choose the range (24 h, 7 days, 30 days), an event type, a Source, or search by name and detail.
  • Click an event to see its details, the recent history of that same host, service, watchdog, monitor or signal, and an Open in … link to its page. All N events narrows the history to just that one; Show every alert clears it.
  • Load older events goes further back.

An event marked Notification sent means DedNets sent a notification for it (Telegram or your webhooks, by your notification settings). It does not mean the message was delivered: DedNets does not know whether your phone or webhook received it.

A signal held back by a quiet window shows Opened without “Notification sent” at first; if it is still open when the window ends and the notification goes out, that same event turns to Notification sent. While its notification is held back, the signal is left out of Needs you now and the open alerts.

A host is recorded as offline only after it has been unreachable for about two minutes, so a quick reconnect - a restart, a network blip - does not fill the history.

Reading alert history from the CLI, API and MCP

(Since v0.16.0.) The same history is available outside the Console, so you can graph it, alert on it, or check how often a quiet window is doing its job. All of it needs a token with the read scope.

dednetsctl alerts list                                   # the last 50 events
dednetsctl alerts list -since 7d -kind deferred -source signal
dednetsctl alerts list -signal disk-iops-web1              # one signal's whole story
dednetsctl alerts list -since 30d -all -json | jq .kind  # every event, one JSON per line
dednetsctl alerts summary -window 7d                     # totals, median time to resolve

-kind takes opened, escalated, resolved, paused, resumed and deferred; -source takes host, service, watchdog, uptime and signal. -json prints one event per line. -all follows every page.

Over the API: GET /api/v1/alerts/events (filters since in unix seconds, kind, source, entity with one source, q, limit, and before with the previous page’s next_before) and GET /api/v1/alerts/summary?window=24h|7d|30d&tz=<zone>. An AI agent connected to DedNets has alert_history and alert_summary for the same thing.

How far back the history goes

The history keeps about 90 days by default (the operator can change this with CONSOLE_ALERT_HISTORY_DAYS). Older events age out. When the history first appeared, the console filled it in from what watchdogs, uptime monitors and signals had already recorded; services and hosts start from that release, because nothing was kept for them before.