Monitoring

Knowing a server is running out of disk before the customer calls you.

The agent measures CPU, memory, disk, uptime and services on every device. When something crosses a threshold, an alert opens. If you have also set up email, the notice reaches you.

Where it lives

Side menu, Monitoring. Inside:

• Monitoring panel, with the state of your whole fleet
• Network, for anything that cannot run an agent
• Screen Wall, the wall of live screens
• Agent deployments

From the monitoring panel you reach the alert inbox, the history, threshold settings and silence rules. Those four have no menu entry of their own.

Each device also has its Monitoring and Alerts tabs inside its record.

The two switches

This is the first thing to understand, because it is the number one cause of "monitoring never tells me anything".

Every device has two separate switches, on the Monitoring tab of its record:

Monitoring for this device. Starts on. This is what makes the device get measured. Turn it off and metrics stop being stored, thresholds stop being evaluated, and every open alert on that device is closed.

Email alerts for this device. Starts off. This is the only thing that decides whether an email goes out.

So with factory settings, a new device does all of this:

• Gets measured every 5 minutes
• Charts the last 24 hours
• Evaluates every threshold
• Opens alerts, with severity
• Shows them in the inbox and the menu counter
• Auto-resolves them when the problem clears
• Stores them in the history and the PDF

And sends not a single email. Not when the alert opens, not when it escalates from warning to critical.

There is no other channel: no in-app bell, no push notification, no messaging. The only notice that leaves KairosLink is the email. With that switch off, nobody finds out about anything unless someone opens the console and looks.

Getting it working

Two things, and you need both:

One. Turn on Email alerts in the record of every device that matters to you.

Two. Set who receives them. That lives in the customer's notification routing, Monitoring category. With no recipients, the switch being on is not enough and nothing goes out either.

Today this is done device by device, there is no bulk action. If you have a large fleet, start with the servers.

What gets measured, and how often

The check runs every 5 minutes by default. That interval is configurable between 1 and 60 minutes.

A device is marked offline after 70 seconds with no signal, and that detection is separate from the metrics check: it is much faster.

Thresholds

Factory values:

Two more things start switched off and are worth setting up if they matter to you: the uptime alert (useful on servers that need periodic reboots) and the free-gigabytes disk alert, which on a 4 TB drive is far more useful than a percentage.

Where to change them

For the whole organization: monitoring panel, monitoring policy block. There you set CPU, memory, disk and interval.

For one customer, group or device: monitoring settings, specific adjustments tab. That gives you all thirteen fields, including the ones not available at organization level.

Which one wins

Resolution is field by field, not policy by policy. Each value is looked up in this order and the first level that defines it wins:

Device, then group, then customer, then organization, then the factory value.

That means you can set disk to 95% on one file server, and that device will keep using your general CPU and memory thresholds. You do not have to redefine everything.

In each device record, the effective thresholds block shows you where each value came from, in brackets. If you are unsure why a device alerts when it should not, look there first.

If a device belongs to several groups that define the same field, the lowest value wins.

A device is in red

1. Open the monitoring panel. Critical devices come first.
2. Click the device to open its monitoring record.
3. Look at the last 24 hours chart. A one-off spike and a steadily rising curve are different problems.
4. Check the effective thresholds block: the threshold may simply be wrong for that device.
5. If you need the reading right now rather than the last cycle's, request a manual check.

One warning about manual checks: if the device is offline, the request is accepted anyway and nobody tells you afterwards that it never arrived. It waits and is discarded after 15 minutes. If you requested a check and nothing happened, first look at whether the device is online.

Silencing during a migration

When you know a server is going to sit at 100% all night because you are migrating data, there is no point in it alerting.

You have two tools:

A maintenance window, if it is planned with a time range. No alerts open during the window.

A silence rule, if you want something surgical: silence only the disk alerts on one device and leave everything else live.

What you should not do is turn off the monitoring switch. That closes every open alert on the device and stops storing metrics, so afterwards you will not have the chart of what happened during the migration, which is usually exactly what you want to look at.

Network monitoring

This covers what cannot run an agent: routers, switches, printers, NAS boxes, access points.

It works by designating one of the customer's devices as a probe. That device, which does have the agent, is the one that scans the network from inside and reports what it found and what responds.

You need one device with a running agent on that network. If the probe goes down, discovery stops, and network devices keep their last known values with nothing warning you they are stale. If you see the network all green, confirm the probe is online.

The Screen Wall

A wall of live device screens, meant to sit on an office monitor.

You build it by choosing which devices go on it. Privacy is configured separately: you decide whether the person using the device sees a notice that their screen is on the wall, and how.

One practical detail: if you click the remote control button from the Screen Wall and access is blocked by rules, you will see no message at all. The button simply does nothing. If you suspect that is the cause, try from the Devices list, where the button greys out and the reason is visible.

When the console cannot see

This section matters because it is where a green dashboard can mislead you.

An offline device is clearly distinguished. The card says there is no connection and when the last contact was, and it counts in the offline tally.

A device that stopped reporting metrics is not. If the agent is alive but the check hangs, the card keeps showing the bars and numbers from the last good check. A device whose monitoring stopped working six hours ago looks exactly like a healthy one.

The only signal is the small line reading "Checked ago". If that text says longer than the policy interval, the reading is stale.

That is the habit worth forming: on the panel, alongside the colours, read the dates.

And there is a case that vanishes from the count. A device whose check returned something the console could not interpret is left in an unknown state, and it counts in none of the four summary cards. Not ok, not warning, not critical, not offline. If the card numbers do not add up to your fleet total, that is where the missing ones are.

If you turned monitoring off at organization level, device cards keep showing their last data and nothing in the device record says it is off. The only place it shows is the monitoring enabled checkbox, on the panel.

Common errors

"Monitoring never tells me anything." The email alerts switch is off on that device, or there are no recipients set on the customer. Check both.

I requested a manual check and nothing arrived. The device is offline. The request waits and is discarded after 15 minutes, silently.

A device alerts on disk and has plenty of space. It is a large drive and the threshold is a percentage. Set a free-gigabytes threshold for that device or customer.

I changed the threshold and the device still alerts the same. A more specific adjustment is winning. Look at the effective thresholds block in the device record: it tells you where each value comes from.

The card numbers do not add up to my device total. Devices in an unknown state are in none of the four.

Frequently asked questions

Why do I get no emails if monitoring is working? Because they are two separate switches. The email one starts off on every device.

Is there any other way to be notified? No. Email is the only outbound channel for monitoring alerts.

Can I turn email on for every device at once? Not today. It is done device by device.

How often does it measure? Every 5 minutes by default, configurable between 1 and 60.

How long does it take to detect a device going down? Between 70 seconds and two minutes. Much faster than the metrics check.

Can I set different thresholds per customer? Yes, and per group and per device too. The most specific wins, field by field.

Can I set thresholds per disk volume? Yes, at customer, group or device level. Not at organization level.

What if a device is in several groups with different thresholds? The lowest value wins.

How do I silence a device without losing its metrics? With a maintenance window or a silence rule. Turning monitoring off closes alerts and stops storing data.

Does network monitoring need anything installed on the router? No. It needs a device with an agent on that network to act as a probe.

What happens if the probe goes down? Discovery stops and devices keep their last values, with no warning that they are stale.

Can I open a ticket from an alert? Yes, from the alert itself, manually.