How big should swap be, and how to add it: a swapfile walkthrough (Rocky 9)

When to use it

Processes get killed by the OOM killer, or a small-memory host needs a cushion. Swap is not “the more the better”: too small and it does nothing, too large and memory pressure turns into disk thrashing. This page gives the sizing rule first, then a copy-paste swapfile procedure.

Check the current state

cat /proc/swaps
swapon --show

Filename    Type       Size       Used   Priority
/dev/zram0  partition  24639296   22642820  32767
/dev/vdc    partition  1048572    390964    1

Output means swap is already enabled; no output means none is active. free -h works too, but minimal systems may not have procps installed.

How big

RAM	Suggested swap
≤ 2 GB	2× RAM
2–8 GB	Equal to RAM
8–64 GB	At least 4 GB
> 64 GB	Workload-dependent, at least 4 GB

The table follows Red Hat’s recommended ranges (RHEL 9 storage docs, checked 2026-10-18). It is a starting point, not a law: databases, container hosts and hosts with elastic memory needs should size from real workload data.

Create the swapfile

sudo fallocate -l 4G /swapfile
sudo chmod 600 /swapfile
ls -l /swapfile
sudo mkswap /swapfile

-rw------- 1 root root 4294967296 ... /swapfile
Setting up swapspace version 1, size = 4 GiB (4294963200 bytes)
no label, UUID=...

chmod 600 is not optional: with 644 permissions mkswap warns and swapon refuses the file — a world-readable swapfile leaks memory contents.

Enable and persist

sudo swapon /swapfile
swapon --show
echo '/swapfile none swap defaults 0 0' | sudo tee -a /etc/fstab

Only when swapon --show lists a new /swapfile entry is it active; the /etc/fstab line keeps it enabled after a reboot.

Verify

swapon --show
cat /proc/swaps
sudo swapoff /swapfile && swapon --show   # confirm it can be disabled safely
sudo swapon /swapfile

Container limitation: inside a rockylinux:9 container fallocate, chmod and mkswap all succeed, but swapon returns Operation not permitted (the container lacks the privilege). So the “enabled” step must be verified on a real system; the container only proves the file-creation steps.

Common traps

• fallocate works on XFS, but on filesystems without preallocation support (some network filesystems) create the file with dd instead;
• swappiness is not “higher is better”: database workloads usually lower it (e.g. 10) so the kernel prefers RAM;
• zram and swapfile are not either/or: small hosts can prefer zram with a swapfile as a backstop, but both consume resources and belong in capacity planning.

Practice: give a 4 GB host 4 GB of swap

Prerequisite: Rocky 9 with root or sudo and at least 5 GB free on the root filesystem.

sudo fallocate -l 4G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
echo '/swapfile none swap defaults 0 0' | sudo tee -a /etc/fstab

Checks:

1. swapon --show → a /swapfile entry of about 4G
2. cat /proc/swaps | wc -l → one line more than before
3. After sudo reboot, swapon --show → still there (fstab works)
4. sudo swapoff /swapfile && sudo swapon /swapfile → disables and re-enables cleanly

Next steps

• Growing storage (page pending): /en/storage-lvm/
• Post-install checks: First-hour checks
• Performance debugging: Performance tuning