Installation·Linux Foundation Certified System Administrator

LFCS lab: Ubuntu, AlmaLinux, and safe storage practice

Two Ubuntu hosts and one RHEL host, because LFCS is sat on either family and the candidate chooses. None of the three has a spare disk, so the Storage domain is taught on loop devices - a file, losetup, and then genuinely mkfs, mount and /etc/fstab against a real block device. That started as a constraint and turned out to be the better lesson, because a reader following along has no spare disk either.

Start Here Guide 2 of 38 Beginner

PlatformsUbuntu 26.04 LTS + AlmaLinux 10.2
LVM2.03.31 (Ubuntu) / 2.03.36 (AlmaLinux)
nftables1.1.6 (Ubuntu) / 1.1.5 (AlmaLinux)
TimeAbout 16 min

Ubuntu 26.04 with kernel 7.0.0 and systemd 259; RHEL 10.0 with kernel 6.12 and systemd 257. Loop devices behave identically on both.

Firewallufw 0.36.2 enabled but reporting inactive / firewalld active
Network confignetplan + systemd-networkd / NetworkManager 1.56.0

Two Ubuntu hosts and one AlmaLinux host, all three belonging to this path. What matters is the distribution in the OS column.
Server Name	IP Address	OS	Roles	CPU	RAM	HDD
LFCS-A01	192.168.0.70	Ubuntu 26.04 LTS	Primary host - most guides run only here	2 Core	4 GB	50 GB
LFCS-B01	192.168.0.71	Ubuntu 26.04 LTS	Second host - so networking has two ends	2 Core	4 GB	50 GB
LFCS-C01	192.168.0.72	AlmaLinux 10.2	The other distribution - dnf, firewalld and NetworkManager	2 Core	4 GB	50 GB

This guide includes

Use this before the first LFCS guide. This matters because the exam lets you choose the distribution and this lab has no spare disk - so every storage task is done on a block device you build out of a file.

meeting the three machines, two Ubuntu and one RHEL
finding that there is no spare disk, and making a real block device from a file
putting a filesystem on it, and mounting it
taking it back down in the order that actually works
checking what sudo is set up to allow, since every guide on this path uses it

Before you start

guide 1 - the domains these machines are for.

The three machines
One primary Ubuntu host, a second Ubuntu host so networking has two ends, and one RHEL host for every command that differs.
bash Example session
```
$hostnamectl | head -8  Static hostname: lfcs-a01        Icon name: computer-vm          Chassis: vm 🖴Chassis Asset Tag: 8665-3071-9889-5090-6456-3643-06       Machine ID: 0866a31c6c9e49e38f752fa07a1c8c72          Boot ID: f796628f1ad64c93a5a19e6040b4b41e   Virtualization: microsoft Operating System: Ubuntu 26.04 LTS$hostnamectl hostname; . /etc/os-release; echo "$PRETTY_NAME  kernel $(uname -r)"lfcs-a01Ubuntu 26.04 LTS  kernel 7.0.0-30-generic$hostnamectl hostname; . /etc/os-release; echo "$PRETTY_NAME  kernel $(uname -r)"lfcs-c01AlmaLinux 10.2 (Lavender Lion)  kernel 6.12.0-211.7.3.el10_2.x86_64
```
Note
hostnamectl with no argument is the whole identity of the box - hostname, machine ID, boot ID, virtualisation, OS. It is the first command to run on a machine you have just been handed, and it is Domain 2 material.

hostnamectl hostname NAME sets it persistently, which is how these two got their names: they arrived called something else, and the name appears in the prompt of every transcript on this path.

The RHEL host keeps its name because three guides on another path depend on it being that host. What matters is the OS column, not the hostname.

Expected resultlfcs-a01 on Ubuntu 26.04, lfcs-b01 on Ubuntu 26.04, and rhcsa-util01 on RHEL 10.0.

Success conditionYou know which machine each command belongs to.
No spare disk
One 50 GB disk, all of it in use - and that is the same situation a reader following along is in.
bash Example session
```
$lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINT | grep -v '^sr0\|^fd0'NAME                          SIZE TYPE FSTYPE      MOUNTPOINTsda                            50G disk|-sda1                          1M part|-sda2                          2G part xfs         /boot`-sda3                         48G part LVM2_member  |-almalinux_lfcs--c01-root 44.1G lvm  xfs         /  `-almalinux_lfcs--c01-swap  3.9G lvm  swap        [SWAP]$df -h / /boot | column -tFilesystem                            Size  Used  Avail  Use%  Mounted  on/dev/mapper/almalinux_lfcs--c01-root  44G   6.0G  39G    14%   //dev/sda2                             2.0G  454M  1.5G   23%   /boot$ls -l /dev/loop-control && losetup -a | head -3; echo "attached loop devices: $(losetup -a | wc -l)"crw-rw----. 1 root disk 10, 237 Aug 29 13:36 /dev/loop-controlattached loop devices: 0
```
Note
Ubuntu installs onto LVM by default, which is convenient here: ubuntu--vg-ubuntu--lv is a logical volume, so pvs, vgs and lvs have something real to report before you create anything.

The 1M partition is the BIOS boot partition - no filesystem, no mount point, and it is not a mistake. lsblk showing a partition with an empty FSTYPE is normal for that.

/dev/loop-control is what makes the next step possible. It is the interface for allocating loop devices, and its presence is the only prerequisite for the whole storage track.

Expected resultsda with a 1M BIOS partition, a 2G /boot, and 48G of LVM holding / - and zero attached loop devices.

Success conditionYou know why the storage track is written the way it is.
A real block device, made from a file
truncate a file, attach it with losetup, and look at what the kernel now thinks it has.

Compare the two du outputs.
bash Example session
```
$sudo mkdir -p /opt/lfcs && sudo truncate -s 512M /opt/lfcs/tour.img && ls -lh /opt/lfcs/tour.img && du -h --apparent-size /opt/lfcs/tour.img && du -h /opt/lfcs/tour.img-rw-r--r--. 1 root root 512M Aug 29 15:15 /opt/lfcs/tour.img512M	/opt/lfcs/tour.img0	/opt/lfcs/tour.img$sudo losetup --find --show /opt/lfcs/tour.img/dev/loop0$lsblk -o NAME,SIZE,TYPE /dev/loop0 && sudo blockdev --getsize64 /dev/loop0 && cat /sys/block/loop0/queue/logical_block_sizeNAME   SIZE TYPEloop0  512M loop536870912512
```
Note
The file is sparse. du --apparent-size says 512M because that is how big it claims to be; du says 0 because not a single block has been written yet. That is truncate allocating nothing - and it is why this costs no disk space until you use it. fallocate -l 512M would reserve the space for real, and dd would write it out.

losetup --find --show picks the next free device and prints it, which is the form to use in a script - hard-coding /dev/loop0 breaks the moment something else is attached.

The kernel treats it as a disk. lsblk lists it, blockdev --getsize64 reports its size, and /sys/block/loop0/queue/logical_block_size says 512. Everything in the storage track works on it.

Where it differs from a real disk: there is no partition-alignment penalty to get wrong, no SMART data, and a loop device does not survive a reboot unless something reattaches it. The storage guides say so where it matters.

Expected result512M from ls -lh and du --apparent-size, but 0 from du - then /dev/loop0, 512M, type loop, 536870912 bytes, 512-byte sectors.

Success conditionYou can create a block device on any machine.
A filesystem on it, and a mount
mkfs.ext4, blkid, mount - the same three commands you would run on /dev/sdb1.
bash Example session
```
$sudo mkfs.ext4 -q -L lfcstour /dev/loop0 && sudo blkid /dev/loop0/dev/loop0: LABEL="lfcstour" UUID="f32dc454-84fa-44dc-b3fb-3564bd6dd4be" BLOCK_SIZE="4096" TYPE="ext4"$sudo mkdir -p /mnt/tour && sudo mount /dev/loop0 /mnt/tour && df -h /mnt/tour | tail -1 && sudo touch /mnt/tour/proof && ls /mnt/tour/dev/loop0      488M   24K  452M   1% /mnt/tourlost+foundproof
```
Note
488M of 512M. The missing 24 megabytes are filesystem overhead - the superblock, the inode tables, the journal, and 5% reserved for root by default. A storage question that asks why a fresh filesystem is smaller than its device is asking about this.

The UUID is the thing to remember. Device names are not stable - today's /dev/loop0 is tomorrow's /dev/loop3, and /dev/sdb becomes /dev/sdc when a disk is added - so /etc/fstab uses UUID= or LABEL=. That is one of the most reliably examined facts in Domain 5.

lost+found exists because ext4 made it, not because anything went wrong. XFS has no equivalent.

Expected resultA LABEL, a UUID and BLOCK_SIZE="4096" from blkid; then df reporting 488M of a 512M device, with lost+found and the file just created.

Success conditionYou have a working filesystem you made yourself.
And taken back down
The cleanup, in the order it has to happen.
bash Example session
```
$sudo umount /mnt/tour && sudo losetup -d /dev/loop0 && sudo rmdir /mnt/tour && sudo rm -f /opt/lfcs/tour.img && sudo rmdir /opt/lfcs 2>/dev/null; losetup -a | wc -l; ls /opt/lfcs 2>&1 | tail -10ls: cannot access '/opt/lfcs': No such file or directory
```
Note
The order is forced. Unmount before detaching the loop device, detach before deleting the file. Skip a step and you get target is busy - which is the same error, and the same fix, as a real disk you cannot unmount.

Every guide on this path ends like this, and the cleanup is part of the captured transcript rather than something claimed afterwards. These are shared machines: the storage guides never touch sda, the networking guides work in a namespace or on a dummy interface, and the users guides delete the accounts they create.

lsof +D /mnt/tour or fuser -vm /mnt/tour is how you find what is holding a mount open when umount refuses - worth knowing before you need it.

Expected result0 attached loop devices and /opt/lfcs: No such file or directory.

Success conditionYou can leave the machine as you found it.
What sudo is set up to allow
Every guide on this path uses sudo, so it is worth seeing exactly what that means here.
bash Example session
```
$sudo -n -l | tail -4User sysadmin may run the following commands on lfcs-a01:    (ALL : ALL) ALL    (ALL) NOPASSWD: ALL$id; groupsuid=1000(sysadmin) gid=1000(sysadmin) groups=1000(sysadmin),4(adm),24(cdrom),27(sudo),30(dip),46(plugdev),100(users),101(lxd)sysadmin adm cdrom sudo dip plugdev users lxd
```
Note
NOPASSWD: ALL is a lab convenience and nothing to copy. It is why the captures have no password prompts in them; on a machine that matters, sudo should ask, and it should be scoped to the commands a role actually needs.

sudo -n -l is the non-interactive form - it lists what you may run and fails rather than prompting, which makes it safe in a script. Domain 3 returns to this properly.

Group membership is what grants it: sudo on Debian and Ubuntu, wheel on the RHEL family. That difference is the first of many, and guide 3 is next.

Expected result(ALL : ALL) ALL and (ALL) NOPASSWD: ALL, and membership of sudo among others.

Success conditionYou know why no command on this path prompts for a password.

Troubleshooting

losetup: cannot find an unused loop device.

Why: All the pre-created loop devices are attached, or the module is not loaded.

Fix:losetup -a to see what is attached and losetup -d to release. modprobe loop if /dev/loop-control is missing.
umount: target is busy.

Why: A process has a file open under the mount point, or your shell is sitting in it.

Fix:cd out, then fuser -vm /mnt/point or lsof +D /mnt/point to find the holder.
The loop file is much smaller on disk than its size suggests.

Why: truncate creates a sparse file - correct, and usually what you want.

Fix:du against du --apparent-size shows both numbers. Use fallocate to reserve the space for real.
The loop device is gone after a reboot.

Why: Nothing reattached it. Loop devices are not persistent by themselves.

Fix:For a lab, reattach with losetup. /etc/fstab can mount a file directly with the loop option if you want it back automatically.
A fresh filesystem reports less space than the device has.

Why: Metadata and the root reserve, not an error.

Fix:tune2fs -m 0 removes the reserve on ext4 if the space matters more than the safety margin.

Official sources

← All LFCS guides

LFCS lab: Ubuntu, AlmaLinux, and safe storage practice

This guide includes

Before you start

The three machines

No spare disk

A real block device, made from a file

A filesystem on it, and a mount

And taken back down

What sudo is set up to allow

Troubleshooting

Official sources