Server metrics
The Overview tab on a server shows what the agent is reporting: what the machine is doing now, and what it has been doing.
What is shown#
| Metric | Reading it |
|---|---|
| CPU usage | Percent in use. High is not automatically bad — a busy server is a working server. Sustained high is what matters. |
| Memory | Percent in use. On Linux, memory used for cache is not memory you have lost. |
| Disk | Percent full. The one to watch most closely. |
| Load average | Work waiting, over 1, 5 and 15 minutes. Compare to your CPU count. |
| Uptime | How long since the machine last rebooted. |
Load average, briefly#
Load average is roughly how many processes are running or waiting. Compare it to the number of CPU cores:
- Load 1.0 on a 4-core machine — quiet.
- Load 4.0 on a 4-core machine — fully busy, nothing queued.
- Load 12.0 on a 4-core machine — work is piling up.
The three numbers are 1, 5 and 15-minute averages. If the 1-minute figure is far above the 15-minute one, something started recently. If they are all high, it has been going on a while.
Disk is the one that bites#
A full disk fails in confusing ways rather than announcing itself: a database that cannot write, logs that stop, a deploy that half-completes, a certificate renewal that silently fails.
Set a disk alert at around 80%, so you hear about it while there is still room to act.
Old releases are a common culprit — every deploy keeps the previous one, which is what makes rollback possible, and that space adds up.
History#
The history chart shows these metrics over time. Its main use is answering "was it like this before?" — a number is hard to judge in isolation, and a graph that has been flat for a week tells you a spike is new.
After a server has been offline, the graph shows a gap for the period. That is missing data, not a reading of zero.
What is not here#
These are machine-level metrics. They will not tell you your application is returning errors, or that a background job has stopped. For that you want uptime checks and your own application logging.
Raise a ticket and someone will pick it up. Sign in with the same account you use for the panel.
Raise a ticket