Server overload happens when a machine cannot finish requests in time because demand, faults, or abuse exceed its capacity.
People see errors and slow pages, and other controls such as request caps exist to keep one client from taking the whole machine.
How it works
A popular launch, a stuck database query, or a flood of pointless requests can each push a service over the line.
The visible result is similar, so operators need metrics that separate a real crowd from a loop or a hostile burst.
This page describes the condition and the protective response.
It does not explain how to exhaust a server, and it treats abusive volume as something to absorb and shed, not as a method to copy.
Concrete examples
- A ticket sale opens and honest buyers plus refresh loops saturate the front servers together.
- A faulty report query holds connections until the site stops answering even small pages.
- Repeated calls to a cheap looking search route consume the database until checkout pages time out.
Protective measures
- Set capacity and automatic scaling for events you can predict, and load test before the public opening.
- Shed or simplify nonessential work when error rates rise, so buying and sign in stay available.
- Cap abusive clients and cache expensive reads so one pattern cannot occupy every worker.
- Page an owner when saturation lasts, and keep a short note of cause so the next event is cheaper.