Alert catalog
This article refers to Platform v3.4.0. The current Platform version is v3.4.0.
Overview
Every notification rule is based on one alert type from the platform catalog. You cannot create rules over metrics or events outside it. This article lists the alert types and explains which thresholds each one supports.
Available alert types
| Alert | Value Threshold | Time Threshold |
|---|---|---|
| Node CPU Usage | 85 % | 15 min |
| Node RAM Usage | 85 % | 15 min |
| Node Disk Usage | 85 % | 15 min |
| Node GPU Usage | 85 % | 15 min |
| Node GPU RAM Usage | 85 % | 15 min |
| Workload Unhealthy | None | 5 min |
| Node Offline | None | 15 min |
| Workload Log Overload | None | 15 min |
| Disk Safety Action Executed | None | None |
| Node Not Activated | None | None |
| Docker Daemon Error | None | None |
| Docker Network Inconsistent | None | None |
The values are suggestions. You can adjust them within the minimum and maximum limits the platform defines. The minimum Value Threshold is 30 %.
Telemetry alerts
The five usage alerts (CPU, RAM, disk, GPU, and GPU RAM) are evaluated per node. They use both thresholds: the condition becomes active when the usage exceeds the Value Threshold, and Barbara notifies you once it has held for the Time Threshold.
Workload and node status alerts
Workload Unhealthy, Node Offline, and Workload Log Overload use only a Time Threshold: Barbara notifies you when the condition has persisted for that time.
Event alerts
Disk Safety Action Executed, Node Not Activated, Docker Daemon Error, and Docker Network Inconsistent have no thresholds. The rule fires as soon as the event occurs.
Disk Safety Action Executed reports that the node's disk usage reached a critical level and that Barbara automatically executed a safety action to free up space and protect system stability. When you receive it, review the disk usage on that node.
Summary
The catalog combines telemetry alerts that use a value and a time threshold, status alerts that use only a time threshold, and event alerts without thresholds. Pick the alert type that matches what you need to watch when you create a rule.
To create one, see Notification rules.