A box is slow or swapping. Work through it step by step: measure, find the offender, kill it if needed, then check whether a service — not a one-off process — is the real cause via systemctl and journalctl.
Use this when a Linux host feels slow, swap is climbing, or an alert says memory is high. It walks from "how bad is it" to "who's doing it" to "stop it" to "was it a service, and what does its log say."
kill -15 <PID> (SIGTERM) — ask the process to exit cleanly; always try this first.
kill -9 skips the process's own cleanup — open files, temp locks, in-flight DB writes. Give SIGTERM a few seconds to work before escalating, or you can trade a memory problem for a corruption problem.
drop_caches only frees reclaimable page cache, and Linux refills it immediately from disk reads — it does nothing for memory a leaking process is actually holding. Use it to confirm a diagnosis, not as the fix.