Sentinel and network score
Two tools working on the same diagnosis: the sentinel warns about the failures the signal does not betray while you work on something else, and the score sums up the whole network's state in one grade.
The service sentinel
A link can look perfect — excellent signal, 100 % quality — and the customer still be without service: a loose LAN cable does not touch the signal. The app is centered on signal, and that is exactly why the sentinel exists: it watches the failures that do not show on the traffic light.
What it watches
| Notice | What it detects |
|---|---|
| Access point down | An AP that was online stops responding — you get one notice per node, saying how many stations it drags down, instead of one alert per client. |
| Session down | A station that was online stops responding. |
| LAN cable unplugged | The radio's port lost its cable: the link looks fine, the customer cannot browse. |
| LAN at half duplex | The port negotiated badly: service goes out degraded without the signal showing it. |
| Misaligned antenna | A strong imbalance between the two polarities: aim or polarization. |
| Radio saturated | The radio's CPU or memory at the limit: it drops packets with the signal intact. |
| Watched device not answering | A device from Others with the watch enabled stopped responding — meant for a node's router. |
Notices stack by severity: first what leaves a whole cell without service, then what affects one client, then what degrades it, and last what anticipates a problem.
What it does NOT watch, on purpose
Signal, noise, link quality, capacity and traffic stay out: they vary with rain, wind and whatever the customer is downloading. If they were in, the panel would fill up every day and within a month nobody would open it. The sentinel only admits conditions that fail in binary and that a healthy network does not produce.
How it behaves
- It confirms before warning: a condition must hold for a few seconds — a port that blinks during a reboot does not deserve a house call.
- One alert per device and condition while it lasts; once resolved it is marked and retires itself minutes later.
- Cause, not consequence: if an AP falls, the AP is alerted and its stations are silenced — a fallen node must not bury the panel under fifty notices.
- Absence of data is never a failure: if a firmware does not report the port, that axis is simply not evaluated for that device.
- Freezing the tables does not stop it: the sentinel watches the background monitoring, not what is painted.
Three channels for the same notice
The sentinel always watches everything; what you choose is what leaves the app and through where:
- The panel — every alert, always. Turning an event off in Preferences does not hide it from here.
- Windows notifications — they reach the desktop even with the app minimized, and clicking one takes you to the device. Which events arrive is chosen in Preferences → Alerts.
- The phone — WhatsApp or Telegram, with its own event selection (see below). Alerts and Messaging are independent settings: you may want everything on the desktop and only the serious ones on the phone.
Phone notices (WhatsApp or Telegram)
The same alerts can reach your phone, so you learn about an outage without being in front of the screen. Three ways, all through CallMeBot: WhatsApp, Telegram to a user or Telegram to a group.
It is enabled in Preferences → Messaging:
- Turn on "Enviar avisos por mensajería" and pick the way. Each one asks for its own — WhatsApp: phone and key; Telegram to a user: just your
@username; Telegram to a group: just the key (the key is the group). The "Cómo obtener estos datos" link takes you to CallMeBot's instructions: the authorization is granted from your phone, and that step the app cannot do for you. - Send the test message — it is not optional: CallMeBot validates nothing until a real send, so without the test you would find out about a bad key the day a link falls and no notice arrives.
- Under "Qué se manda a este destino", choose the events.
What arrives: one message per alert type — not a digest — with the project and, for each device, its name, IP and the AP it hangs from. The time is that of the first alert, not the send: what matters is when the customer lost service.
Each type has its wait and its tolerance. The wait is not a delay, it is the grouper: a power cut takes several APs down seconds apart, and that must be a single message — and it also discards passing failures.
| Priority | Events | Wait | Not repeated before |
|---|---|---|---|
| 1 | AP down · LAN cable unplugged · Others not answering | 20 s | 1 h |
| 2 | Session down | 1 min | 4 h |
| 3 | Radio saturated on an AP | 5 min | 6 h |
| 4 | LAN at half duplex · Radio saturated | 15 min | 12 h |
| 5 | Misaligned antenna · A watched device's IP taken by another | 30 min | 24 h |
The limits, upfront
- It watches while the app is open — the sentinel's same limit, and with notices reaching your phone it is easier to believe it is a background service. It is not.
- CallMeBot is a free third-party service: it can lag or fail. Do not use it as the only alert mechanism for a critical service.
- Notices leave through your internet — if your own link goes down, so do the notices. An irony worth knowing in advance.
A Telegram group is an audience
The whole group reads the notice — with your customers' names and IPs inside. Choose the group way knowing who is in it.
It watches while the app runs
The sentinel is not a background service: it watches while SignalScope is open. Whatever happens with the app closed it will not see — the app itself says so at the bottom of the panel.
The network score
A 0-to-100 grade for the whole network, meant to be shared: it downloads as an image ready to send to a partner or a customer.
The grade does not look at signal alone: it runs the full diagnosis on every station — the same one as the table, so they can never contradict each other — and averages.
| Grade | Verdict |
|---|---|
| ≥ 90 | Excellent network |
| ≥ 75 | Stable network |
| ≥ 50 | Needs attention |
| < 50 | Critical network |
Two rules that make it defensible in front of whoever receives it:
- The grade always travels with its coverage: "Graded over N of M stations". A grade that does not say how much was measured is not auditable.
- No measurements, no grade — it shows "Not enough data", not a 0 %. And absent stations do not penalize: a customer who turns their antenna off at night is not a failure of your network; they are reported in their own block.
The card groups into three blocks — the network, link health and availability — and comes out identical on any machine, in light or dark theme.
