Server
The Server sensor watches the machine a private agent is installed on: CPU, memory and disk usage. It raises an incident when a resource stays above its threshold for a given time — a disk filling up, memory exhausted, a CPU pinned at 100%.
How it works#
You do not create server sensors: one is created with each private agent, Core or Advanced, named after the agent. As soon as the agent runs, it reports the CPU usage, the memory and each disk of its machine about once a minute.
Server sensors do not count toward your plan's sensor limit. They cannot be paused, duplicated or deleted on their own: they follow their agent.
Thresholds#
Edit the sensor to set Server thresholds — "Raise an incident when a resource stays above its threshold for the given duration":
| Resource | Threshold |
|---|---|
| CPU | Above a percentage, For a number of minutes. |
| Memory | Above a percentage or a number of MB (In percent), For a number of minutes. |
| Each disk reported by the agent | Above a percentage or a number of MB, For a number of minutes. |
The duration avoids incidents for short peaks: a CPU at 95% for 30 seconds during a deployment is normal; for 15 minutes, it is not.
The incident says which resource crossed its limit ("CPU is over the limit") and is resolved when it goes back under it.
Reading the page#
The sensor's page shows Server health (CPU and memory over time) and Disk space for each disk, over the period you choose. See Results and history.
Example#
par1-prod-dc server sensor: CPU above 90% for 10 minutes, memory above 85% for 10 minutes, the /var disk above 80% for 5 minutes — alerting the infrastructure Slack channel.
Good to know#
Onagre focuses on your applications and services; the server sensor gives a simple health view of the machines that run your agents, not full infrastructure monitoring.