Create and manage metric alert rules

A metric alert rule watches one host metric against one threshold.

Updated Nov 8, 2026

On this page

Summary
Prerequisites
Create a metric alert rule
Expected behavior
Troubleshooting and limitations
Related articles

Summary#

A metric alert rule watches one host metric against one threshold. You choose the comparison, how long the condition must last, which servers the rule applies to, and whether CloudStats should notify any channels. While the condition stays true, CloudStats does not keep sending the same alert. When the condition clears after an alert has fired, CloudStats sends one resolved notice.

This page covers metric threshold rules only. For agent contact, see Server availability alerts. For delivery setup, see Notification channels.

Prerequisites#

A CloudStats account
At least one Linux server with the agent installed and reporting metrics. See Install the Linux agent and Current metrics catalog
Optional: one or more notification channels if you want email, Discord, or webhook delivery. A rule can exist without channels

Create a metric alert rule#

In the CloudStats app, create a metric alert rule with these settings:

Name — a label for the rule. It appears in notifications.
Type — metric (not availability).
Metric — one host metric to watch. Exact agent names and units are listed in the current metrics catalog. The app may not list every catalog name when you create a rule.
Comparison — one of greater than (>), greater than or equal (>=), less than (<), or less than or equal (<=).
Threshold — one number. A rule has a single threshold (not a separate warning threshold and critical threshold on the same rule).
Duration — how many seconds the condition must last before CloudStats sends an alert. Use a whole number of 0 or greater. 0 means CloudStats alerts on the first matching sample.
Severity — Info, Warning, or Critical. Severity labels the rule and appears in notifications. It does not change when the rule fires.
Servers — all servers, or selected servers.
Notification channels — optional. You can attach email, Discord, and/or webhook channels, several channels, or none. If you attach no channels, the rule can still evaluate, but CloudStats does not send notifications. See Notification channels.

You can later edit or delete an existing rule in the CloudStats app. You can also turn a rule off or on if that control is shown.

Expected behavior#

If the duration is 0 and a sample matches the comparison and threshold, CloudStats sends one alert.
If the duration is greater than 0, CloudStats waits until the condition has lasted that many seconds, then sends one alert. If the condition clears before that duration elapses, CloudStats does not send an alert and does not send a resolved notice.
After an alert has fired, CloudStats does not keep resending the same alert while the condition stays true.
When the condition clears after an alert has fired, CloudStats sends one resolved notice.
After a resolved notice, a new matching condition can alert again immediately. There is no separate cooldown timer.
If the metric has labels such as mountpoint, interface, or device, CloudStats evaluates each labeled series independently. One disk, mount, or interface can alert even when others stay below the threshold.
Email, Discord, and webhook messages for a firing metric alert use an alert.firing event (and alert.resolved when it clears). See Notification channels.

Troubleshooting and limitations#

What you see	What it means
No notification, and the condition lasted less than the duration	Expected. CloudStats has not fired yet, or the condition cleared first.
No notification, and the rule has no channels	Expected. The rule can evaluate without sending anything. Attach a channel if you want delivery.
Repeated breaches while the alert is already active	Expected. CloudStats does not resend the same firing alert until the condition clears and then matches again.
One mount or interface alerts and another does not	Expected for labeled metrics. Each series is evaluated on its own.

Limitations of the current product:

Metric alerts compare a live sample to one threshold. They do not alert because a metric stopped reporting. To watch whether the agent is still contacting CloudStats, use a server availability alert.
Average, minimum, or maximum window evaluations are not available as a customer feature.
CloudStats does not provide customer ping, port, cron, website, or SSL checks, a cPanel plugin, a Docker-specific agent, or a customer Grafana product. Those are outside this article.
If notifications are missing after a rule has fired, see Troubleshoot alert and notification delivery.