Reading your usage meters
What CPU, memory and disk are telling you — and when a high number is fine.
Where the numbers come from
Your console page shows live CPU, memory and disk for your server, updated every ten seconds, with the last hour drawn behind the CPU and memory meters. Your dashboard shows CPU and memory for every server you own.
The figures are measured by the hypervisor your server runs on, not by anything installed inside it, so nothing you do to your server can break the reporting.
Memory almost always looks high, and that is normal
This is the number that worries people, so it is worth explaining. Linux treats unused memory as wasted memory. Anything not currently needed by a program gets used to cache recently read files, because a cached file is enormously faster to read a second time. That cache is counted as used.
The moment a program actually needs that memory, the cache is handed over instantly. So a server sitting at 90% is usually a server making good use of what it has, not one about to run out.
Some apps also reserve memory up front regardless of what they are doing. A Minecraft server is the clearest example: it claims its whole allowance at startup and holds it whether one person is playing or nobody is.
Worry about memory when the server becomes slow or a program is killed unexpectedly — not because the bar is long.
CPU is about the shape, not the number
A brief spike to 100% means your server did something, which is what you pay it for. The hour of history behind the meter is the useful part: a line that goes up and comes back down is healthy, and a line that went up an hour ago and has stayed there means something is stuck or genuinely undersized.
The percentage is of everything your plan includes. A two-core server using both cores completely reads 100%, not 200%.
Disk is the one to actually watch
Unlike memory, a full disk breaks things, and it breaks them in confusing ways — databases stop accepting writes, uploads fail, logs vanish. This is the meter worth glancing at occasionally.
Disk shows real usage of your server's main filesystem, read from inside it. If it is missing, your server is either off or its guest agent is not responding, and only the allocated size is shown.
Above about 85%, do something. Old Docker images are usually the easiest win: docker system prune -a removes images nothing is using.
When to move up a plan
Sustained CPU near 100% for hours at a time, or a disk you keep having to clear, are the two signals worth acting on. Contact us and we will size up your server — your data stays where it is.