Tuning Fedora 45: zram and oomd are already on, tuned is not

Before tuning anything, establish what is already done — otherwise you duplicate work or break a good default.

zram, oomd and swappiness already active; tuned needs enabling

Measured defaults

cat /proc/sys/vm/swappiness
zramctl
systemctl is-active systemd-oomd
systemctl is-active tuned

60
NAME       ALGORITHM DISKSIZE DATA  COMPR   TOTAL STREAMS MOUNTPOINT
/dev/zram0 lzo-rle       3.8G   2G 878.9M  896.9M         [SWAP]
active
inactive

Four things are already in place:

Item	F45 default	What it does
zram	enabled, `lzo-rle`, 3.8G	compressed RAM used as swap, far faster than disk swap
systemd-oomd	active	kills processes under memory pressure instead of letting the machine thrash
swappiness	60	kernel default
tuned	not running	you install and enable it yourself

tuned: install it or not

tuned is Fedora’s automatic tuning daemon; the repository carries 2.28.0 and the measured transaction is 13 packages.

sudo dnf install tuned
sudo systemctl enable --now tuned
tuned-adm active

Current active profile: virtual-guest

It detects the machine type. On a virtual machine the default profile after enabling is virtual-guest, with no manual selection.

List every profile:

tuned-adm list

Measured: 15 profiles. The ones that matter:

Profile	Use
`balanced`	general purpose, the default
`desktop`	interactive desktop work
`throughput-performance`	throughput first, power saving off
`latency-performance`	latency first
`virtual-guest`	virtual machine guest (auto-selected here)
`powersave`	power saving

Switch:

sudo tuned-adm profile throughput-performance

zram: already on, do not configure it twice

F45 creates the zram device through zram-generator, configured in /usr/lib/systemd/zram-generator.conf. It is enabled by default, so no manual setup is needed.

rpm -q zram-generator

zram-generator-1.2.1-6.fc45.x86_64

Only edit the configuration to change the size, then:

sudo systemctl daemon-reload
sudo systemctl restart [email protected]

OOM protection: systemd-oomd is running

systemctl is-active systemd-oomd
systemctl status systemd-oomd --no-pager | head -5

It kills by cgroup memory pressure, intervening earlier and more selectively than the kernel OOM killer.

An alternative is earlyoom (1.9.0), blunter but predictable:

sudo dnf install earlyoom
sudo systemctl enable --now earlyoom

Pick one. Running both means they race to kill your processes.

Observation tools

Tool	Version	Purpose
`htop`	3.5.3	live process and memory view
`btop`	1.4.7	prettier TUI including network and disks
`powertop`	2.15	power draw and wakeup analysis

sudo dnf install btop htop powertop
btop

powertop needs root for complete data:

sudo powertop

A sensible order of operations

Look at the defaults: zramctl, systemctl is-active systemd-oomd — for most machines this is already enough
Then decide on tuned: servers and VMs benefit; let it pick the profile
Only then touch kernel parameters: vm.swappiness and friends, having noted the original value

Persist a kernel parameter properly:

echo 'vm.swappiness = 10' | sudo tee /etc/sysctl.d/99-swappiness.conf
sudo sysctl --system

sysctl -w alone is temporary and disappears on reboot.

Common traps

Symptom	Cause
`tuned-adm` cannot reach DBus	the daemon is not started
swappiness change lost after reboot	only `sysctl -w` was used
zram size change had no effect	`systemd-zram-setup@zram0` was not restarted
the machine still stalls	oomd and earlyoom both running, or neither covers the actual bottleneck
no performance change after tuned	the profile is `balanced`, which barely differs from default

systemd services and timers — unit management and daemon-reload
Storage and btrfs — swap and filesystem level
Terminals and shells — where the observation tools run