Monitoring
Machines > Monitoring shows how busy every machine is: CPU, memory, and its fullest disk, with a small graph of the last 24 hours. The busiest machines come first. In a large fleet the page draws the 100 busiest; Find a machine and Only machines over a limit narrow the list, and the totals at the top always count every machine. Each machine's page has a Health card with larger graphs, for the last 24 hours or 30 days.
Nothing extra is installed. The Gardience agent already on each machine reads its own numbers every 5 minutes (or 15), and the readings stay on your Gardience server, so monitoring works without internet access.
What is read
- CPU: percent busy over one second (Windows: the processor load).
- Memory: percent in use, not counting what Linux keeps as reclaimable cache.
- Disks: percent full for the system volume and every other local volume over 1 GB. Network shares and temporary file systems are left out. The page shows the fullest one and lists the rest.
Readings are kept for 24 hours; hourly averages (and the fullest disk each hour) are kept for 30 days.
Alerts
Through the alert channels under Administration > Settings > Alerts (email, Slack or Teams, PagerDuty):
| When | Alert |
|---|---|
| A disk is 90% full or more | at once |
| CPU is at 95% or more | after three readings in a row (15 minutes), so a short spike is not an alert |
| Memory is at 95% or more | after three readings in a row |
Each alert is sent once. It is sent again only after the machine has come back under 85% and crossed the limit again.
Settings
Monitoring is on by default. Admins can turn it off, or read every 15 minutes instead of 5, from the Monitoring page; the change is written to the audit log. A person limited to some node groups sees only their machines.