Эксплуатация6 мин чтения
Reading a server load graph without guessing
What the memory line really means when it is always near the top, why CPU spikes are usually fine, and the shapes that are worth acting on.
Эта статья пока на английском. Мы её переводим.
The most common misreading in hosting is a memory graph near the top of its range. For a Java server that is not a warning, it is the design: the runtime takes the heap it was given and uses it. A flat line at ninety per cent can be perfectly healthy, and a sawtooth that never returns to its floor can be a leak.
Shapes worth knowing
- Sawtooth returning to the same floor: normal garbage collection.
- Sawtooth with a rising floor: something is not being released. That is a leak.
- Flat at the ceiling with pauses: the heap is too small for the workload.
- CPU spikes to 100% for seconds: usually a save or a chunk load, and usually fine.
- CPU pinned at 100% continuously: the simulation cannot keep up, and no memory will help.
Compare against yourself
There is no correct number, only a normal one for your server. Look at the graph on a good evening so that you know what good looks like; a bad evening is then obvious rather than debatable.