Resource Limits

A process can be perfectly isolated from other users and still exhaust the machine’s resources.

A file-descriptor leak can consume every available descriptor. A runaway worker can create thousands of threads. A CPU-bound task can run indefinitely, and a large core dump can fill a filesystem.

Resource limits prevent this kind of unbounded growth by placing ceilings on what a process can consume.

On Unix-like systems, these limits are commonly called rlimits. Linux exposes them through getrlimit(), setrlimit(), and prlimit().

A resource limit is a ceiling, not a reservation. It restricts usage but does not guarantee that the resource will be available.

Setting the limit is only part of the job. Applications must also handle the resulting errors or signals safely.

This chapter explains soft and hard rlimits, examines limits for descriptors, processes, address space, stack, CPU, files, locked memory, and scheduling, and shows how to configure, budget, observe, and diagnose them safely.