Support / Configuring alerts
Configuring alerts
Get told the moment one of your own links leaves its baseline —
in Slack or by webhook. Email is coming soon. Alerting watches the paths you measure, on the
same scale the dashboard already colours.
What alerting watches
Alerting runs on your links only — the site-to-site paths and targets
measured from your own probes. It never watches the global fleet, and it is entirely
separate from any status page you follow.
Each link is judged against its own history, not a fixed number. We learn
what normal looks like for every link, and raise an alert when it moves far enough
off that norm, or stops answering altogether.
It is your circuit compared to itself, in milliseconds. A
164 ms subsea path that always runs at 164 ms is healthy; a cross-town link
that normally runs at 6 ms and jumps to 60 ms is not. Alerting reacts to the
change from normal, never to the raw number.
The scale an alert fires on
Sensitivity is expressed as milliseconds added over the link's trended norm —
the same bands you see on the dashboard:
| Band | Meaning |
| +25 ms | Watch — a measurable rise over normal. |
| +50 ms | Warn — clearly off baseline. |
| +150 ms | Critical — a large, sustained departure. |
| Unreachable | The link stopped answering entirely. |
You pick the band an alert fires at, and a separate packet-loss threshold.
Set it looser for a noisy consumer link, tighter for a circuit you expect to be flat.
Turning it on
- Open your Customer dashboard and sign in.
- On the Alerts panel, click Configure — or open the
Alerting section in the settings rail.
- Choose which links to watch. You can watch all of them or a subset.
- Set the sensitivity band and loss threshold from the table above.
- Add a channel (below) and, if you want, quiet hours.
- Save. Alerting starts on the next measurement cycle.
Channels — where alerts go
Slack
Create an incoming webhook in your Slack workspace and paste its URL into the
channel field. Each alert arrives as a single message naming the link, its reading
against baseline, and when it started. This is the fastest channel to set up — no
account configuration on our side.
Webhook
Give us any HTTPS endpoint and we POST a JSON payload for each alert and
each resolution. Use it to page your own tooling, open a ticket, or fan out to a system
we don't integrate with directly.
Email
Recipients who are members of your account confirm automatically; anyone else
confirms by clicking a link we email them once, so an alert is never sent to an address
that didn't opt in. Email delivery is not switched on yet — the email option stays
greyed out until it is. Use Slack or a webhook in the meantime, or ask
[email protected] to switch it on.
Quiet hours
Set quiet hours in your own timezone. Alerts that occur inside them are
held, not dropped — they are delivered when the quiet window ends, so an
overnight problem is waiting for you in the morning rather than lost.
Why it won't flood you
Alerting is built so a single event is a single message, and so a reading has to earn
its way to your screen:
- A reroute that settles at a worse-but-steady level won't keep re-alerting,
and can't quietly "recover" on its own.
- A flapping link is summarised as one notice, not a message per bounce.
- Too little data stays quiet. A link we haven't measured enough yet, or a
destination that has never answered, produces no alert — an unknown is
reported as unknown, never as a fault.
- Accept as new normal. When a change is expected — you moved a circuit,
a carrier re-homed you — mark it and the baseline resets. Otherwise the link
re-learns its own normal over time and the alert clears.
Muting
Planned maintenance on a site? Mute that link for the window. Muting suppresses
delivery without deleting the policy or losing history — the link is still
measured, you just aren't paged about it.
What an alert tells you
Every alert names the link, its current reading against its own norm
(milliseconds over baseline, and loss if that crossed), and when it started. When
the link returns to normal you get a matching resolved message, so an open alert
always has a close.
Next: back to all guides ·
How our probes measure