Alerts & Thresholds
Server Pulse continuously evaluates server metrics against your configured thresholds and sends notifications when metrics exceed acceptable levels or when your server stops reporting.
How alerting works
Every time a metric report arrives from your server, Pulsewise evaluates each metric against your configured thresholds. A single report over the threshold does not alert on its own, because a short-lived process can briefly spike a metric like CPU.
- 2 consecutive reports over the threshold send a threshold breach alert to your notification channels.
- 3 consecutive reports over the threshold start a phone call to anyone on a Phone Call channel set up for that alert type.
- When all metrics return to normal after an alert, a recovery notification is sent. A brief spike that never triggered an alert does not send a recovery.
- A phone call stops as soon as a report shows that metric back under its threshold.
If no reports arrive for 5 minutes, the server is marked as stale and you're notified immediately.
Server status
Each server has a status based on its most recent metrics and threshold configuration:
| Status | Meaning |
|---|---|
| Pending | Server Pulse is set up but no reports have been received yet |
| Healthy | All metrics are within thresholds |
| Warning | 1–2 metrics exceed their thresholds |
| Critical | 3 or more metrics exceed their thresholds |
| Stale | No metrics received for 5 minutes |
Default thresholds
Server Pulse ships with sensible defaults for each metric. You can customize these per server based on your workload and hardware.
| Metric | Default threshold | Description |
|---|---|---|
| CPU usage | 90% | Percentage of CPU utilization |
| Memory usage | 85% | Percentage of RAM used |
| Swap usage | 80% | Percentage of swap space used |
| Disk usage | 90% | Highest usage percentage across all mounted disks (virtual filesystems like tmpfs and devfs are excluded) |
| Load average | 2.0 per core | 1-minute load average divided by CPU core count |
| Network receive | Disabled | Bytes received per second (must be enabled manually) |
| Network send | Disabled | Bytes sent per second (must be enabled manually) |
Customizing thresholds
Thresholds can be adjusted for each server individually:
- Open the Server Pulse dashboard for your site.
- Click the Alert Thresholds button.
- Adjust any threshold value, or set it to empty/null to disable alerts for that metric.
- Valid ranges: 1–100 for percentages, 0.1–100 for load average per core.
- Changes take effect immediately on the next report.
Alert cooldown
After an alert fires for a specific metric, a 30-minute cooldown starts for that metric. During cooldown, no duplicate alerts are sent for the same metric — even if the metric remains above the threshold.
This prevents notification spam during sustained high usage. Once the cooldown expires, if the metric is still breached, a new alert is sent. The cooldown does not delay phone calls - a metric that reaches 3 consecutive breaches still starts a call.
Notification types
Server Pulse sends several notification types based on metric thresholds and server health:
| Notification | Severity | When it fires |
|---|---|---|
| High CPU | Warning | CPU usage exceeds threshold on 2 consecutive reports |
| High Memory | Warning | Memory usage exceeds threshold on 2 consecutive reports |
| High Swap | Warning | Swap usage exceeds threshold on 2 consecutive reports |
| High Disk | Warning | Highest disk mount exceeds threshold on 2 consecutive reports |
| High Load | Warning | Load average per core exceeds threshold on 2 consecutive reports |
| Server Recovered | Info | All metrics return below thresholds after an alert was sent |
| Server Stale | Critical | No metrics received for 5 minutes |
Notification channels
Alerts are sent through your site's configured notification channels. Supported channels include Email, Slack, Discord, and Telegram.
If no email notification channel is configured, alert emails fall back to the organization owner's email address.
Configure channels on your site's notification settings.
Stale detection
Pulsewise checks for stale servers every minute. A server is marked stale if no report has been received in the last 5 minutes.
Common causes include:
- The collector service stopped
- The server is offline
- Network connectivity issues
The stale notification includes the last time a report was received, displayed in your timezone. When reports resume, the server status automatically returns to healthy, warning, or critical based on current metrics.