Manage swap on Linux: file, partition, zram and swappiness

Swap is a space Linux uses as an extension of RAM: when RAM fills up, the least used pages are moved to it. This guide shows how to add swap through a file, a partition or compressed zram, make it persistent at boot, and tune the swappiness. For beginner to intermediate administrators, on Debian, Ubuntu or RHEL. All commands were tested on an AlmaLinux 10 VM.

What you will learn

• Create a secure swap file with mkswap and swapon
• Choose between disk swap, a partition and zram (compressed in RAM)
• Make swap persistent in /etc/fstab by UUID
• Tune vm.swappiness and size the swap

Prerequisites

• A Linux server with a sudo account
• Free space on a file system (for a swap file)
• Knowing how to identify a disk with lsblk

Skill targeted (RHCSA / LFCS)

Creating and enabling swap (file or partition) and making it persistent is part of the storage domain of the RHCSA and LFCS certifications. mkswap, swapon and vm.swappiness are on the syllabus.

What swap is for

Swap (exchange memory) acts as a safety valve: when RAM is saturated, the kernel moves inactive memory pages to it to free up room. Without swap, a memory spike triggers the OOM killer (Out Of Memory), which abruptly kills a process.

Swap is not a patch for a lack of RAM: a system that swaps permanently is slow, because even an NVMe SSD remains orders of magnitude slower than RAM. It plays two roles: absorbing occasional spikes and enabling hibernation (which writes all of RAM into swap).

Check the current state with free -h and swapon --show:

free -h
swapon --show   # empty if no swap is active

Create a swap file

The swap file is the method recommended today: no repartitioning, easy to grow or remove. It works on every distribution.

1. Create the file (here 512 MB) and secure it immediately.

   sudo dd if=/dev/zero of=/swapfile bs=1M count=512 status=none
   sudo chmod 0600 /swapfile

   The chmod 0600 is mandatory: a swap file readable by everyone would expose the contents of memory. Do not count on mkswap to stop you: on a 0644 file, it prints insecure permissions 0644, fix with: chmod 0600 then formats anyway, with a zero exit code (measured with util-linux 2.40 and 2.41).

2. Format as swap space.

   sudo mkswap /swapfile

   Setting up swapspace version 1, size = 512 MiB (536866816 bytes)
   no label, UUID=c1d4b2fe-5513-404b-9133-18b22e25fdae

3. Enable the swap and check.

   sudo swapon /swapfile
   swapon --show

   NAME      TYPE SIZE USED PRIO
   /swapfile file 512M   0B   -2

The swapon --show line confirms that the 512 MB swap file is active. To disable it: sudo swapoff /swapfile.

dd rather than fallocate for a swap file

The swapon(8) manual is clear: a swap file created with fallocate is not reliable on ext4 (sparse file, swapfile has holes error). mkswap itself flags a sparse file: contains holes or other unsupported extents. This swap file can be rejected by kernel on swap activation!. The portable method remains dd with /dev/zero, shown above. Special case of Btrfs: use sudo btrfs filesystem mkswapfile --size 512M /swapfile, because a Btrfs swap file must be NOCOW and uncompressed.

Use a swap partition

A dedicated swap partition remains common, especially for hibernation. The logic is identical: mkswap then swapon. You can assign a priority when several swap areas coexist.

sudo mkswap /dev/sdb1
sudo swapon --priority 10 /dev/sdb1
swapon --show

NAME       TYPE      SIZE USED PRIO
/dev/sdb1  partition 256M   0B   10

The PRIO column shows the priority: the kernel fills the swaps with the highest priority first (0 to 32767). At equal priority, it uses them in parallel (RAID 0-style striping), useful to spread over several disks.

Compressed swap: zram and zswap

Rather than writing to a slow disk, Linux can compress pages in memory. Two mechanisms coexist, not to be confused.

Mechanism	Principle	Needs a disk?
zram	a virtual disk compressed in RAM (/dev/zram0) serves as swap	No
zswap	a compressed cache in RAM placed in front of a real disk swap	Yes

Swipe to see more

zram suits machines with little RAM or without a dedicated disk: it swaps in RAM, without any disk I/O. Fedora enables it by default; on Debian, Ubuntu and RHEL you have to install it (zram-generator package, configured through /etc/systemd/zram-generator.conf). Here is the manual creation, tested on AlmaLinux:

sudo modprobe zram
ZDEV=$(sudo zramctl --find --size 256M --algorithm zstd)
sudo mkswap "$ZDEV"
sudo swapon --priority 100 "$ZDEV"
zramctl "$ZDEV"

NAME       ALGORITHM DISKSIZE DATA COMPR TOTAL STREAMS MOUNTPOINT
/dev/zram0 zstd          256M   4K   58B   20K       2 [SWAP]

The COMPR column shows the data really stored after compression: zstd offers the best ratio. zswap, for its part, is only useful if you already have a disk swap: it speeds it up by keeping hot pages compressed in RAM (echo 1 | sudo tee /sys/module/zswap/parameters/enabled).

Make swap persistent

Swap enabled with swapon disappears at reboot. For it to come back, add it to /etc/fstab, by UUID rather than by device name (which can change).

sudo blkid -s UUID -o value /dev/sdb1

/etc/fstab

UUID=e28f4374-0322-4875-922c-bfe3f274d461 none swap sw,pri=10 0 0

For a swap file, give its path:

/etc/fstab

/swapfile none swap sw 0 0

Test without rebooting with swapon -a, which enables all the swap entries of fstab:

sudo swapon -a
swapon --show

An error in fstab blocks the boot

An incorrect fstab line can prevent booting. Back up (sudo cp -a /etc/fstab /etc/fstab.bak) and validate with sudo swapon -a (no error shown) before rebooting. zram, for its part, does not go into fstab: it is managed by zram-generator.

Tune the swappiness

The vm.swappiness parameter controls how aggressive swapping is. Contrary to a widespread belief, its range goes from 0 to 200 (the maximum was raised from 100 to 200 in kernel 5.8). The kernel default is 60, but that is not necessarily what your machine shows.

sysctl -n vm.swappiness

On a freshly installed AlmaLinux 10, the answer is 30, without any file in /etc/sysctl.d/, /usr/lib/sysctl.d/ or /etc/sysctl.conf declaring anything: the value comes from the active tuned profile, as the box below explains. Check, do not assume: someone who "fixes" an imaginary 60 towards a low value may change nothing at all.

On a server with enough RAM, a low value (10) keeps data in memory and only uses swap as a last resort. Conversely, with zram or an NVMe, swapping costs little: a value above 100 becomes relevant. Make the setting durable in /etc/sysctl.d/:

echo "vm.swappiness = 10" | sudo tee /etc/sysctl.d/99-swappiness.conf
sudo sysctl -p /etc/sysctl.d/99-swappiness.conf
# vm.swappiness = 10

RHCSA trap: tuned imposes its swappiness

On RHEL, the active tuned profile sets the kernel value. The throughput-performance profile forces vm.swappiness = 10, virtual-guest sets it to 30. Check with tuned-adm active before concluding that a value "does not hold".

A detail that matters for the exam: between your file and the profile, your file wins. /etc/tuned/tuned-main.conf carries reapply_sysctl = 1 by default, so tuned rereads /etc/sysctl.d/ after applying its profile and logs Overriding sysctl parameter 'vm.swappiness' from '30' to '60'. The effective order is therefore: /etc/sysctl.d/ > tuned profile > kernel default. A side effect to know: tuned-adm verify then fails, reporting 'vm.swappiness' = '60', expected '30', which looks like a failure while it is the intended behaviour.

The durable setting therefore comes down to a single file in /etc/sysctl.d/, stronger than the tuned profile. The reverse path, removing that file to roll back, holds a surprise however.

Deleting the file does not restore the original value

Rolling back is not symmetrical, and it is a classic trap. Removing your file from /etc/sysctl.d/ then rerunning sysctl --system restores nothing: the value set lives in memory until the reboot.

Measured on AlmaLinux 10: starting value 30, a file set to 77, file deleted, sysctl --system, and sysctl -n vm.swappiness still answers 77.

To really roll back, set the value explicitly again:

sudo sysctl -w vm.swappiness=30

A nuance exists all the same: if another file declares the same parameter, that one takes over on its own. A grep -rn vm.swappiness /etc/sysctl.d/ /usr/lib/sysctl.d/ settles it before concluding.

Size the swap

How much swap should you plan? It depends on the RAM and on the need for hibernation. Here is the Red Hat recommendation (RHEL 8/9/10), the reference for the RHCSA:

Installed RAM	Recommended swap	With hibernation
≤ 2 GB	2 × RAM	3 × RAM
2 to 8 GB	= RAM	2 × RAM
8 to 64 GB	at least 4 GB	1.5 × RAM
> 64 GB	at least 4 GB	not recommended

Swipe to see more

The old rule "2 × RAM" now only applies to very small configurations. On a well-equipped modern server, 4 to 8 GB are enough: swap absorbs spikes, it does not replace RAM.

Security and points of attention

• Strict permissions: a swap file is always in 0600 (root only), because it contains potentially sensitive memory pages.
• Encryption: swap can write secrets (keys, passwords) in clear text. On a sensitive system, encrypt it, either on a LUKS volume or with an ephemeral key through /etc/crypttab (regenerated at each boot). The ephemeral key is incompatible with hibernation; see the Encrypt a disk with LUKS guide.
• Containers and Kubernetes: kubelet long refused to start with active swap (failSwapOn: true). Swap support (the NodeSwap feature) became stable in Kubernetes v1.34 (2025), but must still be enabled explicitly and requires cgroup v2; the default behaviour is NoSwap. Before adding swap on a node, check the cluster's policy.

When swap is no longer enough: the OOM

If RAM and swap are saturated, the kernel's OOM killer kills a process, often too late (the system is already crawling). Daemons act upstream: systemd-oomd (enabled by default on Fedora and Ubuntu Desktop) and earlyoom watch memory pressure and kill a chosen process before the freeze. Useful above all with zram or a large swap, where thrashing can freeze the machine.

Troubleshooting

Swap failures almost all come down to three causes: permissions too open, a sparse file and an /etc/fstab entry missing or wrong. A fourth one is about tuning: a misunderstood swappiness, often imposed by tuned.

Symptom	Probable cause	Solution
insecure permissions 0644 warning from mkswap or swapon	Swap file too open, enabled anyway	sudo chmod 0600 /swapfile
swapfile has holes	File created with fallocate/truncate	Recreate with dd (or mkswapfile on Btrfs)
Swap missing after reboot	Entry missing in /etc/fstab	Add the line, test swapon -a
The system swaps too much, latency	vm.swappiness too high, or tuned profile	Lower to 10, check tuned-adm active
Boot blocked after editing fstab	Wrong swap line	Rescue mode, fix /etc/fstab

Swipe to see more

Put it into practice

A swap declared in fstab but never tested at reboot is a bet, not a configuration. In this VM lab, you create a secure swap file with mkswap and swapon, write it into /etc/fstab and adjust the swappiness: the validation inspects the machine's state after a reboot to confirm that the swap space did come back.

Guided lab Practice

Learn how to do it

Add and manage swap space

$ dsoxlab run l2-swap-management

Environment: VM·Duration: ~25 min

How it is checked: the state of the system is checked after a reboot, not the command you typed.

First time? What DSOXLab is and how to install it

DSOXLab is the command line that prepares a disposable environment (an isolated shell or a VM), runs the lab scenario, then checks the real state of the system and computes a score. Nothing is installed or broken on your workstation.

It needs Python 3.11 or later, uv or pipx, and git. The installation guide details these prerequisites; on Windows, or on a Mac without Python, use the appliance instead.

This lab starts virtual machines: it also needs ansible-playbook, a hypervisor (KVM or Incus), Terraform and hardware virtualization (/dev/kvm), which inside a VirtualBox VM means nested virtualization. The first time, dsoxlab start l2-swap-management checks all of it before creating the machines. Prepare your machine for vm labs.

Install the CLI and add the catalogue, once:

uv tool install dsoxlab   # or: pipx install dsoxlab
dsoxlab catalog add https://github.com/stephrobert/linux-dsoxlab-training

Before you run the lab, check:

• DSOXLab installed
• catalogue available
• compatible environment (isolated shell or VM)

To check, nothing is probed from this page: run dsoxlab doctor. Read the diagnosis

Run this lab, then validate your work:

dsoxlab run l2-swap-management
dsoxlab check l2-swap-management

See every lab: dsoxlab list-labs. Track your progress: dsoxlab progress. The full dsoxlab guide covers scoring and the pitfalls.

Key points

• Swap absorbs memory spikes and enables hibernation; it does not make up for a lack of RAM.
• The swap file (dd + chmod 0600 + mkswap + swapon) is the most flexible method; fallocate is not reliable on ext4.
• zram swaps in compressed RAM, without a disk; zswap speeds up an existing disk swap.
• Make swap persistent in /etc/fstab by UUID; zram, for its part, goes through zram-generator.
• vm.swappiness goes from 0 to 200 (60 by default on the kernel side, 30 on a fresh AlmaLinux 10); 10 on a server, above 100 with zram. Beware of the tuned profile on RHEL.
• A swap file is always in 0600; on Kubernetes nodes, swap (NodeSwap, stable in v1.34) must still be enabled explicitly.

Knowledge check

Check that the essentials of this guide are acquired. The questions only cover what has just been explained here.

Knowledge check

Check what you have learned with this interactive quiz

6 questions

6 min

70% to pass

How it works

• The timer starts when you click Start
• Multiple choice, true/false and short answer questions
• You can move between questions
• Detailed results are shown at the end

Start the quiz

Starts the quiz and the timer

×

See in the glossary →