Search⌘ K
AI Features

Resource Exhaustion Triage and Reversible Mitigation

Resource exhaustion can significantly slow down services, manifesting through full disks, depleted memory, or saturated CPUs. Identifying the specific resource constraint is crucial, as each has distinct indicators and requires tailored, reversible mitigations. For disk issues, moving or archiving logs is preferable to deletion, while memory issues may necessitate service restarts. CPU saturation can be addressed by reducing request concurrency. Effective triage involves gathering evidence to confirm the limiting resource and ensuring that any mitigation leads to measurable improvements without compromising system integrity.

Interviewers also ask about slow service, and they watch whether you test before you tune. The same symptom can come from three different resource limits, and a change that fits one does nothing for the others.

How exhaustion shows up

A resource is exhausted when demand for it reaches its capacity: the disk is full, memory is used up, or the CPU has no idle time. Exhaustion is a common reason a service slows down or restarts.

Each resource leaves its own signature:

  • Disk: A filesystem runs out of available space or inodes. We can locate space usage with df -h, check for inode exhaustion using df -i when errors persist despite available space, and estimate allocated usage with du beneath a directory. When df and du disagree, that is a cue to ...