Linux Out of Memory (OOM)

When the kernel runs out of RAM and swap, the OOM killer terminates a process to keep the system alive. Symptoms include random service deaths, SSH lag, and
Killed in logs without a clean application stack trace.

Start with htop and free and
swap if you need short-term headroom while you fix the leak.

What OOM Looks Like

Apps disappear, load spikes, database connections drop. The kernel log names the victim process.

Check Memory Now

Use available in free -h; watch swap I/O.

bashsnapshot

free -h
swapon --show

Find the Memory Hog

Sort processes by RSS before and during incidents.

bashtop memory

ps aux --sort=-%mem | head -15

Read OOM Killer Lines

Kernel messages include the score and chosen PID.

bashkernel log

sudo dmesg -T | grep -i 'out of memory\|Killed process' | tail -20
sudo journalctl -k --since today | grep -i oom

Limits with systemd and ulimit

Cap runaway services with MemoryMax= so one unit cannot take the whole box.

iniunit drop-in

[Service]
MemoryMax=2G

Short-Term Relief vs Real Fix

Restarting frees RAM; adding swap buys time; adding RAM or fixing the leak solves it.

Prevent Next Time

Alert on memory percentage; graph trends in Prometheus/Grafana if you have it.

WarningDisabling the OOM killer is almost never appropriate on production servers.

Quick Reference

Grep Killed process in dmesg / journalctl -k
ps aux --sort=-%mem for suspects
MemoryMax= on heavy systemd services

Linux Out of Memory (OOM): Find the Killer, Fix and Prevent Recurrence