Notifications and service checks
Server notifications use rules. Each rule monitors one category and can send messages to multiple recipients through email or Slack.
Monitored events
| Category | What it monitors |
|---|---|
| CPU usage | Average percentage usage across measured server cores. |
| RAM usage | Memory usage based on the operating system's total and available memory. |
| System disk usage | System filesystem usage; this is not a total across every attached disk. |
| VM availability change | Changes in the server's running state. |
| SSL certificate expiry | Certificate check results. |
| Service availability change | Results of configured service checks. |
| Restart without a confirmed request | A boot change not matched to a confirmed request or recorded maintenance. |
Create a rule
- Open notifications in server navigation and select Add rule.
- Choose an event and check Active.
- For CPU, RAM or disk, configure threshold, duration and recovery margin.
- Enable recovery notifications and reminders as needed.
- Add recipients, selecting a channel and entering its details for each.
- Save and verify the category, channels and active state in the list.
A new form starts with RAM, a 90% threshold, a five-minute duration and a five-percentage-point recovery margin. Recovery messages and reminders are initially disabled.
Managing rules requires server developer-tool permissions. Active rules are automatically evaluated when the operator has enabled the new automatic notification system and relevant data collection is running. Saving a rule alone does not confirm a message was sent.
Thresholds, duration and recovery
For CPU, RAM and system disk, configure:
- Usage at least: an integer from 1–100%. Equality with the threshold counts as a breach.
- Breach duration: immediately, 1, 5, 10 or 15 minutes. Immediately means at a valid observation, not continuous monitoring of every instant.
- Recovery margin: a nonnegative integer smaller than the threshold, in percentage points. An incident ends only when usage falls below threshold minus margin.
For example, a 90% threshold, five-minute duration and five-point margin trigger a notification after sufficiently long measured usage of at least 90%. Recovery occurs below 85%; exactly 85% is insufficient. A recovery message is sent only if enabled.
Missing or invalid data is not treated as zero usage or recovery. A sampling interruption can restart the breach-duration countdown. The configured duration therefore does not guarantee delivery at an exact second.
Reminders and state events
Disable reminders or choose 30 minutes, 1 hour, 6 hours or 24 hours. Metric reminders require a current value at or above the threshold; a value between the recovery boundary and threshold does not itself trigger a reminder.
VM availability, services and SSL do not use percentage thresholds or configurable breach durations. They react to check results. An initially observed stopped VM establishes a baseline rather than a new outage. Repeated new VM outages can have delivery suppressed within a 30-minute period.
For SSL, an ongoing problem uses a daily reminder when no interval is selected. Selecting no reminders for SSL therefore does not remove this default daily behavior.
Email and Slack
A rule requires 1–20 recipients. The same email address or Slack webhook cannot appear twice in one rule.
- Email: enter a valid address. It also serves as the recipient label in the list.
- Slack: enter a channel label and Slack Incoming Webhook matching
https://hooks.slack.com/services/…. The label is descriptive; the webhook determines where messages arrive.
Saved webhook URLs are not shown while editing. Leave the field empty to keep the existing webhook, or enter a new URL to replace it. Treat it as a secret and keep it out of public support reports.
Edit, disable and delete
Open the edit icon for a rule. To disable it temporarily, clear Active and save. Changing category, enabled state or metric conditions resets its evaluation state; observations are assessed against the new configuration.
Deleting requires confirmation. Previously recorded event history remains. A pending message may be cancelled if its rule or recipient no longer matches the recorded event. A message already handed to the transport cannot be recalled.
Restart without a confirmed request
The first valid measurement establishes a baseline. A notification requires a change in the operating system boot ID; an uptime drop or unavailable observation alone does not prove a restart.
A change matching a confirmed application request or recorded maintenance window is recorded as suppressed and sends no message. Unconfirmed or failed requests do not suppress the notification.
The event does not determine the cause: a restart outside the application may have been planned. Messages distinguish detection time from any estimated boot start. Restart rules have no recovery messages or repeated reminders.
Notification history
History below the rules lists events and recipients. Filter by:
- period: all, the last 48 hours, 7 days or 30 days;
- event category;
- type: test, threshold breach, problem, recovery, reminder or suppressed event;
- delivery result.
A delivery-result filter selects events where at least one recipient has the matching result. The recipient list can also include other channels with different outcomes. History covers the new notification system, not older emails from legacy subscriptions.
| Delivery result | Meaning |
|---|---|
| Pending | The message has not yet been handed to the transport. |
| Sending | A delivery attempt is in progress. |
| Sent | The transport accepted the message; this does not confirm it was read. |
| Failed | Delivery was rejected or configuration was invalid. |
| Delivery unknown | Acceptance cannot be confirmed; the message may have been sent and is not retried automatically. |
| Suppressed | Delivery was cancelled, for example after changing or removing a rule or recipient. |
The server overview also shows the five newest events from the last 48 hours and links to history for that period.
Service checks and troubleshooting
In service checks, configure a name (3–255 characters) and port (1–65535). Then create an active Service availability change rule with recipients. Saving a check alone does not configure who receives notifications.
If a message is missing, check the active rule, category, conditions, recipients and history. Check spam for email and the channel and webhook validity for Slack. An unknown result can still mean delivery occurred. Missing events can indicate insufficient valid measurements or disabled automatic evaluation; contact support if the problem persists.
An open port alone does not confirm a correct application response. Test actual requests and compare monitoring and logs.
In-app security alerts
Critical vulnerabilities with known fixes in managed services appear in the global security alerts overview. The application reports progress and the outcome of an update you start; no email or Slack rule is required. See security updates for details.