Start here: your Linux survival kit

Never used Linux or a command line before? Good, this page was written for you. Spend a little time here and the rest of the labs will feel much easier.

The story for all six labs. A small company has hired you as its only IT technician. It cannot afford real servers, so you will build its whole "data centre" inside a single ordinary computer using virtual machines. By the last lab you will create servers by typing one command, the way real cloud engineers do.

How this page helps you

  • Words marked like this have a plain-language explanation. Click one and it opens in the Help panel. You can also search the Help panel for any word.
  • Under most commands there is a link called What does this mean? It explains the command piece by piece. Open it every time until the command feels familiar.
  • Tick each step when it works. Your progress is saved in this browser.
  • Each lab has three levels. Guided gives you every command. Your turn gives you a goal and hints you can reveal one at a time. Challenge gives you only the goal. The help gets smaller as you get stronger.

Three computers, three places to type

You have one real computer and two pretend ones running inside it. Every command box on this page has a label at the top telling you where to type it. Typing a command in the wrong place is the most common beginner mistake, so always read the label first.

  • Linux serverThe black text screen of your Linux VM, called srv-linux on this page. It is a guest.
  • Windows clientInside your Windows VM (win-client), in a program called PowerShell. Also a guest.
  • Host computerThe real, physical computer you are sitting at, the host. It may be your own laptop or a PC in the computer lab. Use its terminal: PowerShell on Windows, Terminal on macOS or Linux.

If your VMs have different names, use your names wherever you see srv-linux and win-client.

Mouse stuck inside a VM window? Press the Right Ctrl key once (on a Mac, the left ⌘ key). VirtualBox calls this the host key. It gives the mouse and keyboard back to your real computer.

Your own laptop, or a computer-lab PC?

Both work. The instructions are the same, and this page calls either one the host. If you use a computer-lab PC, four things are different:

On a lab PCWhat to do
Your VMs live on that PC's diskThey do not follow you to another computer. Sit at the same PC every session and write down its number. If that is not possible, ask your instructor about keeping your VMs on an external drive (Lab 3 shows how to move a VM as a single file).
You may not be an administratorInstalling software on the host, or changing its BIOS settings, may be blocked. When a step needs that (installing Vagrant in Lab 5, for example), ask your instructor instead of looking for a way around it. Inside your VMs you are always the administrator.
Other people use the same PCYou may see VMs in VirtualBox that are not yours. Never start, change or delete them. Shut your own VMs down before you leave, so the next student has the memory.
Your progress ticks are saved in this browserThey stay on that PC and that browser only. Keeping your own notes of where you stopped is a good habit anyway.

Eight things every Linux beginner should know

ThingWhat it means for you
The promptA line such as student@srv-linux:~$ means "I am ready, type a command". If you do not see it, the computer is still busy.
sudoPut sudo in front of a command to run it as the administrator. It asks for your own password. Nothing appears while you type the password, not even dots. That is normal. Type it and press Enter.
Capital letters matterDocuments and documents are two different names. Copy commands exactly.
Tab finishes words for youType the first letters of a command or file name and press Tab. Linux completes it. Fewer typing mistakes.
↑ repeatsThe up arrow brings back your previous commands, so you can fix a typo without retyping everything.
Ctrl+C stopsIf a command never ends (like ping on Linux), this stops it and gives you the prompt back.
No news is good newsMany Linux commands print nothing when they succeed. Silence means it worked.
Read the errorErrors look scary but usually say exactly what is wrong. command not found = typo or program not installed. Permission denied = you forgot sudo. No such file or directory = wrong name or wrong folder.

Editing a file with nano

Several labs ask you to edit a settings file. You will use nano, the simplest text editor on Linux. There is no mouse. Move with the arrow keys and type normally. To save and leave:

  1. Press Ctrl+O (the letter O, for "write Out"), then Enter to confirm the file name. The file is saved.
  2. Press Ctrl+X to exit.

GuidedWarm-up: your first commands

  1. Start srv-linux in VirtualBox and log in with the username and password you chose during installation. When you see the prompt, you are in.

  2. Ask the server a few questions about itself. Type each line, press Enter, and read the answer before typing the next one.

    whoami
    hostname
    pwd
    ls
    What does this mean?
    whoami
    "Who am I?" Prints the username you are logged in as.
    hostname
    Prints the name of this computer.
    pwd
    "Print working directory". Shows which folder you are standing in. /home/yourname is your personal folder.
    ls
    "List". Shows the files in the current folder. On a fresh server it may show nothing, because the folder is empty.
  3. Now a command that needs administrator rights. It refreshes the list of software the server is able to install. Remember: the password stays invisible while you type.

    sudo apt update
    What does this mean?
    sudo
    Run the rest of the line as the administrator.
    apt
    The program that installs software on Ubuntu and Debian. Think of it as the app store, without pictures.
    update
    Download the latest catalogue of available software. It does not install or change anything yet.

    If this fails with network errors, your VM has no internet. Check that its network adapter is set to NAT (VM Settings > Network), then try again.

  4. Type your Linux username into the box at the bottom of the left menu on this page. From now on the commands here will show your real username instead of youruser.

  5. Shut the server down properly. Never just close the window of a running server, in the same way you would not pull the plug on a real one.

    sudo poweroff

Your turn

Start the server again. Find out how much free disk space it has, shown in units a human can read (GB, not bytes).

The command is two letters long and stands for "disk free".
Many Linux commands accept -h to mean "human-readable".
df -h. Look at the line that ends with /. That is your main disk. The Avail column is the free space.

Checkpoint

  • I can log in to srv-linux and I know what the prompt looks like.
  • I ran a sudo command and I was not surprised by the invisible password.
  • I can explain the difference between the host and a guest.
  • I know how to save and exit in nano.

Lab 1: Connect the machines

A server nobody can reach is useless. In this lab you discover the four ways VirtualBox can connect a VM to the world, and you build the network that every later lab depends on.

Your job today. The company wants its server to download updates from the internet, but it must also talk privately to the office PC. Find out which kind of virtual network can do what, then set it up.

Before you type anything: five ideas

Click each word and read its explanation. You will need all five today: IP address, ping, network adapter, DHCP, firewall.

Copy this table into your notes and fill it in as you go. Write your prediction first, then test.

ModeVM's IP addressVM reaches internet?VMs reach each other?Host reaches VM?
NAT
Bridged
Host-only
Internal network

Your testing toolkit

You will repeat these same tests in every part. On the server:

ip -br addr
ping -c 3 8.8.8.8
ping -c 3 <client-ip>
What does this mean?
ip -br addr
Show this machine's network adapters and their IP addresses. -br means "brief": one short line per adapter. Ignore the line called lo; that is the machine talking to itself.
ping -c 3 8.8.8.8
Send 3 test messages (-c 3 = count 3) to 8.8.8.8, a Google server that always answers. Replies mean "I can reach the internet".
<client-ip>
A placeholder. Do not type the < > signs. Replace the whole thing with the Windows VM's address, for example ping -c 3 192.168.56.101.
lo        UNKNOWN   127.0.0.1/8
enp0s3    UP        10.0.2.15/24

In that example the adapter is called enp0s3 and its address is 10.0.2.15. The /24 is explained under subnet.

On the Windows client, open the Start menu, type PowerShell, and open it:

ipconfig
ping 8.8.8.8
ping <server-ip>
What does this mean?
ipconfig
The Windows way to show IP addresses. Look for the line IPv4 Address.
ping
Same idea as on Linux. Windows stops by itself after 4 messages, so no -c is needed.

GuidedPart A: NAT, the default

  1. For each VM, open Settings > Network > Adapter 1 in VirtualBox and check that Attached to says NAT. Start both VMs.

  2. Find each VM's IP address with your toolkit. You will probably see 10.0.2.15 on both. That is not an error: under NAT each VM lives behind its own private pretend router, so they never meet.

  3. Run the three tests (internet, other VM, and from your host try ping 10.0.2.15). Write the results in your table.

GuidedPart B: Bridged

  1. Change Adapter 1 on both VMs to Bridged Adapter. In the Name list choose the network card the host is really using: usually Wi-Fi on a laptop, or Ethernet (the cable) on a computer-lab PC. You can do this while the VM is running.

  2. Ask for a new address. The easy way is to reboot each VM. On Windows you can run ipconfig /renew instead.

  3. Look at the addresses now. They should resemble the host's own address, as if the VMs were two extra computers plugged into the same network. Repeat the three tests.

    Many campus networks refuse bridged VMs. If a VM gets no address, write "blocked by the network" in your table and move on. Think about why an administrator would want that.

GuidedPart C: Host-only

  1. In the main VirtualBox window open File > Tools > Network Manager and choose the Host-only Networks tab. If the list is empty, click Create. Note the address range (normally 192.168.56.x) and make sure its DHCP server is enabled.

  2. Change Adapter 1 on both VMs to Host-only Adapter. Reboot the VMs and repeat the tests.

  3. Pinging the Windows client will probably fail even though the network is fine. Windows Firewall silently drops incoming pings. Allow them: open the Start menu, type PowerShell, right-click it, choose Run as administrator, then:

    New-NetFirewallRule -DisplayName "Allow ping" -Protocol ICMPv4 -IcmpType 8 -Direction Inbound -Action Allow
    What does this mean?
    New-NetFirewallRule
    Add a new rule to Windows Firewall.
    -DisplayName "Allow ping"
    The label you will see in the firewall's list of rules.
    -Protocol ICMPv4 -IcmpType 8
    The technical name for a ping request. ICMP is the "are you there?" language of networks; type 8 is the question itself.
    -Direction Inbound -Action Allow
    Let these messages come in instead of dropping them.
    Remember this moment. "The network is fine but a firewall drops the traffic" is the single most common problem in real cloud systems.

GuidedPart D: Internal network

  1. Change Adapter 1 on both VMs to Internal Network and type the same name on both: labnet. Reboot both.

  2. Check the addresses. Nobody hands out addresses on this network (there is no DHCP), so the server has none, and Windows invents a 169.254.x.x address for itself.

  3. Give the server an address by hand. First find your adapter's name, then use it in the second command.

    ip -br link
    sudo ip addr add 10.10.10.1/24 dev enp0s3
    What does this mean?
    ip -br link
    List the network adapters and their names, without addresses. Yours is usually enp0s3. If it is different, use your name in the next command.
    ip addr add 10.10.10.1/24
    Add the address 10.10.10.1 to an adapter. /24 means "my neighbours are everyone whose address starts with 10.10.10."
    dev enp0s3
    "On the device called enp0s3". This says which adapter receives the address.
    (temporary)
    An address added this way disappears when the machine reboots. Good for experiments.
  4. Give the client an address too, in an administrator PowerShell. If it complains about the adapter name, run Get-NetAdapter to see the real name and use that.

    New-NetIPAddress -InterfaceAlias "Ethernet" -IPAddress 10.10.10.2 -PrefixLength 24
    What does this mean?
    New-NetIPAddress
    Give a Windows network adapter a fixed address.
    -InterfaceAlias "Ethernet"
    Which adapter. "Ethernet" is the usual name of the first one.
    -PrefixLength 24
    The Windows way of writing /24.
  5. Repeat the three tests and complete your table. Then put the client back to automatic addressing:

    Remove-NetIPAddress -InterfaceAlias "Ethernet" -IPAddress 10.10.10.2 -Confirm:$false
    Set-NetIPInterface -InterfaceAlias "Ethernet" -Dhcp Enabled

Check your table

Only look after you have filled in your own. If a result differs, find out why before moving on.

ModeInternetVM to VMHost to VM
NATYesNoNo (only with port forwarding, Lab 2)
BridgedYesYesYes
Host-onlyNoYesYes
Internal networkNoYesNo

GuidedPart E: The network you keep

No single mode gives the company everything. Real servers solve this by having two network adapters: one road to the internet, and one private road between your own machines.

  1. Shut down both VMs. On each one, set Adapter 1 back to NAT. Then open the Adapter 2 tab, tick Enable Network Adapter, and attach it to Host-only Adapter. Start both VMs.

  2. Windows sets up its second adapter by itself. Ubuntu Server does not. Check: the new adapter (usually enp0s8) shows as DOWN with no address.

    ip -br addr
  3. A server should never change its address, or clients could not find it. So give it a static address, permanently this time. Ubuntu keeps network settings in netplan files. Create a new one:

    sudo nano /etc/netplan/60-hostonly.yaml
    What does this mean?
    sudo nano
    Open the nano text editor as administrator, because system settings files are protected.
    /etc/netplan/
    /etc is the folder where Linux keeps almost all its settings. netplan is the sub-folder for network settings.
    60-hostonly.yaml
    The new file's name. The name is your choice; it must end in .yaml.
  4. Type exactly this into the file. The file is written in YAML, where indentation is part of the meaning: use the space bar, two spaces per level, never the Tab key.

    network:
      version: 2
      ethernets:
        enp0s8:
          dhcp4: false
          addresses: [192.168.56.10/24]
    What does this mean?
    ethernets:
    "Here come settings for wired network adapters."
    enp0s8:
    The adapter these settings belong to. Use your own adapter's name if it differs.
    dhcp4: false
    Do not ask anyone for an automatic address.
    addresses: [...]
    Use this address instead. .10 is safe because VirtualBox's DHCP only hands out .101 and above.

    Save with Ctrl+O, Enter, then exit with Ctrl+X.

  5. Protect the file, apply the settings, and look at the result.

    sudo chmod 600 /etc/netplan/60-hostonly.yaml
    sudo netplan apply
    ip -br addr
    What does this mean?
    chmod 600
    "Change mode": set who may read the file. 600 means only its owner (the administrator) can read and write it. Netplan prints a warning if its files are readable by everyone.
    netplan apply
    Read all the netplan files and make the network match them, right now, without rebooting.
    lo        UNKNOWN   127.0.0.1/8
    enp0s3    UP        10.0.2.15/24
    enp0s8    UP        192.168.56.10/24
    An error from netplan apply almost always means wrong indentation. Open the file again and compare it with the box above, space by space.

Your turn

The boss wants the office PC to have a fixed address too. Give win-client the permanent address 192.168.56.30 on its host-only adapter, and prove that the server can still ping it.

You already gave Windows a fixed address once today, in Part D. The same command works, with a different address.
Windows now has two adapters. Run Get-NetAdapter and ipconfig together to work out which name belongs to the 192.168.56.x one. It is often called "Ethernet 2".
In an administrator PowerShell: New-NetIPAddress -InterfaceAlias "Ethernet 2" -IPAddress 192.168.56.30 -PrefixLength 24. Then on the server: ping -c 3 192.168.56.30.

Challenge

Draw your finished network on paper: the host, both VMs, all four adapters with their addresses, the two networks, and the internet. Draw the path of a ping from the client to the server, and the path of sudo apt update from the server to the internet. A good drawing here will save you hours later.

Checkpoint

  • The server shows 192.168.56.10 on its second adapter.
  • ping -c 3 8.8.8.8 works on the server.
  • The client can ping 192.168.56.10.
  • My host can ping 192.168.56.10.

Hand in: your completed table, a screenshot of the checkpoint pings, your network drawing, and two or three sentences answering: why did both VMs have the same IP address under NAT, and why was that not a conflict?

How this looks in a real cloud

NAT is a private subnet behind a "NAT gateway": servers can fetch updates, but nobody outside can start a conversation with them. Bridged is a server with a public address. Host-only is a private subnet you reach through a management connection. Internal network is an isolated subnet, the kind used for databases. Your two-adapter design in Part E is the standard cloud pattern.

Lab 2: Make the server useful

A server exists to serve. Today yours gets three jobs: letting you log in from another computer, hosting the company website, and sharing files with the office PC.

Your job today. The boss wants to manage the server from her desk, see a company web page, and save documents to a shared folder. You need the two-adapter network from Lab 1, Part E.

Before you type anything: four ideas

Read these first: service, package, port, SSH.

GuidedPart A: Log in from far away with SSH

  1. Install the SSH server program and check that it is running.

    sudo apt update
    sudo apt install -y openssh-server
    sudo systemctl status ssh
    What does this mean?
    apt install
    Download and install a program.
    -y
    Answer "yes" automatically when apt asks "Do you want to continue?"
    openssh-server
    The name of the package that lets other computers log in to this one.
    systemctl status ssh
    systemctl manages services. status asks "is it running?" Look for the green words active (running). Press q to leave the status screen.
  2. Connect from the Windows client. Windows 10 and 11 already include an SSH program. The first time, it asks whether you trust the server's fingerprint: type yes. Then enter your Linux password.

    ssh youruser@192.168.56.10
    What does this mean?
    ssh
    Open a secure remote terminal.
    youruser@192.168.56.10
    "Log in as this user, on the computer at this address." Same shape as an email address: who @ where.

    Look at the prompt. It now says srv-linux, although you are sitting at Windows. Everything you type is happening on the server.

  3. Passwords can be guessed. A key pair cannot. Type exit to come back to Windows, then create your key pair. Press Enter at every question to accept the defaults.

    exit
    ssh-keygen -t ed25519
    What does this mean?
    exit
    Leave the server and return to the Windows prompt.
    ssh-keygen
    "Key generator". Creates two matching files: a private key (your secret, never share it) and a public key (safe to hand out, ends in .pub).
    -t ed25519
    The type of key. Ed25519 is a modern, strong and short kind.
  4. Install your public key on the server. This is one long line; use the Copy button. It asks for your Linux password one last time.

    type $env:USERPROFILE\.ssh\id_ed25519.pub | ssh youruser@192.168.56.10 "mkdir -p ~/.ssh && cat >> ~/.ssh/authorized_keys && chmod 700 ~/.ssh && chmod 600 ~/.ssh/authorized_keys"
    What does this mean?
    type ...id_ed25519.pub
    Windows command that prints the contents of your public key file.
    |
    The pipe. It sends the output of the command on the left into the command on the right.
    ssh ... "..."
    Log in to the server, run the commands inside the quotes, and leave.
    mkdir -p ~/.ssh
    Make a folder called .ssh in your home folder (~ means "my home"). -p means "no error if it already exists".
    &&
    "And then, if that worked, do the next thing."
    cat >> ...authorized_keys
    Add the incoming key to the end of the list of keys allowed to log in. >> means "append to this file".
    chmod 700 / 600
    Make the folder and file private to you. SSH refuses keys kept in files that other users could change.
  5. Connect again. This time there is no password question. Stay connected: from now on, do all server work through this SSH window, not the black VirtualBox screen. You can finally copy and paste, and it is how real cloud servers are managed, because they have no screen at all.

    ssh youruser@192.168.56.10

GuidedPart B: A website, and a locked door

  1. Install the nginx web server and replace its welcome page with your own. Put your real name in it.

    sudo apt install -y nginx
    echo "<h1>Served by $(hostname) - YOUR NAME</h1>" | sudo tee /var/www/html/index.html
    curl http://localhost
    What does this mean?
    echo "..."
    Print the text between the quotes.
    $(hostname)
    Run the hostname command and paste its answer right here, so the page shows the server's name.
    | sudo tee file
    tee writes what it receives into a file. We use it because the web folder belongs to the administrator, and sudo echo ... > file would not work.
    /var/www/html/index.html
    The file nginx shows when somebody visits the site.
    curl http://localhost
    curl fetches a web page and prints it as text: a browser without a screen. localhost means "this same machine".
  2. On the Windows client, open a browser and go to http://192.168.56.10. You should see your page.

  3. Turn on the server's firewall, ufw. Allow SSH first, or the firewall will cut off the very connection you are typing in.

    sudo ufw allow OpenSSH
    sudo ufw enable
    sudo ufw status
    What does this mean?
    ufw allow OpenSSH
    Add a rule: let SSH connections in.
    ufw enable
    Switch the firewall on, now and at every boot. From this moment everything not explicitly allowed is blocked.
    ufw status
    Show the list of rules.
  4. Refresh the browser on the client. The page no longer loads. nginx is running perfectly, but the firewall only lets SSH through. Open the web door and refresh again.

    sudo ufw allow 80/tcp
    What does this mean?
    80/tcp
    Port 80 is the door number that websites use. TCP is the kind of connection web browsers make.
  5. Now reach the site from your host, through the NAT adapter, using port forwarding. In VirtualBox open srv-linux Settings > Network > Adapter 1 > Advanced > Port Forwarding. Add a rule: Protocol TCP, Host Port 8080, Guest Port 80, leave the other boxes empty. On your host, browse to http://localhost:8080.

Your turnPart C: A shared folder

Windows shares folders using a language called SMB. The Linux program that speaks it is Samba. Your goal: the folder /srv/share on the server appears on the Windows client as \\192.168.56.10\labshare, protected by a username and password. A file created from Windows must show up on the server.

Try with as few hints as you can. Note how many you needed.

The plan. (1) Install the package samba. (2) Create the folder and make yourself its owner. (3) Describe the share at the end of Samba's settings file, /etc/samba/smb.conf. (4) Give your user a Samba password. (5) Restart the service. (6) Open the firewall. Which earlier commands can you reuse?
Folder and settings. Create the folder with sudo mkdir -p /srv/share and take ownership with sudo chown $USER:$USER /srv/share (chown = change owner). A share is described by a name in square brackets followed by settings such as path, read only and valid users.
Password, restart, firewall. Samba keeps its own password list: sudo smbpasswd -a youruser. The service is called smbd. ufw knows an application profile named Samba.
sudo apt install -y samba
sudo mkdir -p /srv/share
sudo chown $USER:$USER /srv/share
sudo nano /etc/samba/smb.conf

Add at the very end of the file, then save and exit:

[labshare]
   path = /srv/share
   browseable = yes
   read only = no
   valid users = youruser
sudo smbpasswd -a youruser
testparm -s
sudo systemctl restart smbd
sudo ufw allow Samba
What does this mean?
smbpasswd -a
Add a user to Samba's password list and set their sharing password.
testparm -s
Check the settings file for mistakes before you restart.
systemctl restart smbd
Restart the Samba service so it reads the new settings.

On the client, type \\192.168.56.10\labshare into the File Explorer address bar, or map it as drive Z:

net use Z: \\192.168.56.10\labshare /user:youruser

Create a text file there, then check on the server with ls -l /srv/share.

The share asks for a password on purpose. Recent Windows versions refuse anonymous "guest" shares, and so should you.

Challenge

The boss wants a second website, for staff only, on the same server. Make nginx serve a different page on port 8080. It must open from the Windows client at http://192.168.56.10:8080 and from your host at http://localhost:8081.

Three doors must open: nginx needs a second server { } block that says listen 8080; with its own root folder (look in /etc/nginx/sites-available/default for the pattern, and run sudo nginx -t to test your settings before sudo systemctl reload nginx). Then the firewall. Then VirtualBox.

Checkpoint

  • I can SSH from the client to the server without typing a password.
  • My web page opens from the client at http://192.168.56.10.
  • My web page opens from the host at http://localhost:8080.
  • A file I created in Windows appears in /srv/share on the server.

Hand in: screenshots of the checkpoint items, the output of sudo ufw status, how many Samba hints you used, and a short answer: in Part B step 4 the web server was running perfectly, so why could the client not see the page?

How this looks in a real cloud

ufw rules are what clouds call security groups, and "allow SSH before you enable" is a lesson many engineers learn the hard way. Port forwarding is how a load balancer exposes one door of a private machine. And key pairs are the normal way to log in to cloud Linux servers; most providers will not even offer a password.

Lab 3: Save, copy and move servers

On real hardware, one bad command can cost a day of reinstalling. A virtual machine is only files on a disk, and files can be saved, copied and carried. That changes everything.

Your job today. Protect the server you built, destroy it on purpose to prove the protection works, then copy it to make a second server in minutes.

Before you type anything: three ideas

Snapshot, clone, appliance (.ova).

GuidedPart A: Snapshot, break, restore

  1. In the VirtualBox window select srv-linux, click the small menu icon beside its name and choose Snapshots. Click Take and name the snapshot services-working. You can do this while the VM runs.

  2. Now break the server on purpose. Yes, really.

    sudo rm -rf /etc/nginx
    sudo systemctl restart nginx
    What does this mean?
    rm
    "Remove". Deletes files. There is no recycle bin on a Linux server: deleted means gone.
    -rf
    r = the folder and everything inside it. f = force, do not ask questions. Together they are the most dangerous letters in Linux. Always read the path twice before pressing Enter.
    /etc/nginx
    The folder holding all of nginx's settings.

    The restart fails and the website is dead. On a physical server you would now be reinstalling.

  3. Shut the VM down (sudo poweroff). In the Snapshots view select services-working and click Restore. Untick "Create a snapshot of the current machine state": you do not want to keep the broken version.

  4. Start the VM and reload the web page from the client. It is back, as if nothing had happened.

  5. Take a second snapshot called after-restore. Then, on your host, open the VM's folder (right-click the VM > Show in Explorer / Finder) and look inside its Snapshots sub-folder.

    You will find small disk files with long names. A snapshot does not copy the whole disk. It freezes the current disk and writes every later change into a new differencing disk. That is why snapshots are instant, and also why a snapshot is not a backup: if the original disk file is lost, every snapshot is useless.
  6. Delete the after-restore snapshot (select it, click Delete). VirtualBox merges its changes back. The running state of your VM does not change.

GuidedPart B: Clone the server

  1. Shut down srv-linux. Right-click it and choose Clone. Name the clone srv-linux-2. For MAC Address Policy choose Generate new MAC addresses for all network adapters (what is a MAC address?). Choose Full clone.

  2. Leave srv-linux powered off and start only srv-linux-2. Log in at its black console window and look around with hostname and ip -br addr. Same name, same address, same users, same website. It is a perfect twin, and that is a problem: two machines with the same identity cannot share a network.

  3. Give the twin its own name and address. In the netplan file, change 192.168.56.10 to 192.168.56.11.

    sudo hostnamectl set-hostname srv-linux-2
    sudo nano /etc/netplan/60-hostonly.yaml
    sudo netplan apply
    What does this mean?
    hostnamectl set-hostname
    Change the computer's name permanently.
  4. Two hidden identities were copied too. Replace them, change the web page so you can tell the twins apart, and reboot.

    sudo rm /etc/machine-id
    sudo systemd-machine-id-setup
    sudo rm /etc/ssh/ssh_host_*
    sudo dpkg-reconfigure openssh-server
    echo "<h1>Served by srv-linux-2 - YOUR NAME</h1>" | sudo tee /var/www/html/index.html
    sudo reboot
    What does this mean?
    /etc/machine-id
    A unique serial number Linux creates for itself at installation. See machine ID. We delete it and let systemd-machine-id-setup create a fresh one.
    /etc/ssh/ssh_host_*
    The server's own SSH host keys, the source of the fingerprint you accepted in Lab 2. The * means "every file whose name starts like this".
    dpkg-reconfigure openssh-server
    Re-run the SSH package's setup, which creates new host keys.
  5. Start srv-linux as well. From the client, browse to http://192.168.56.10 and http://192.168.56.11. Two web servers, and the second took minutes instead of hours.

Your turn

Make another clone of srv-linux, but this time a linked clone. Compare the size on disk of three folders on your host: the original VM, the full clone, the linked clone. Then delete the linked clone completely.

The choice between Full and Linked appears on the last page of the same Clone wizard. The VM must be powered off.
To find the folders, right-click a VM and choose Show in Explorer (or Finder). Check the folder's size in its Properties.
The linked clone is tiny because it contains only a differencing disk that points back to the original's disk. To delete: right-click the clone > Remove > Delete all files. Think: what would happen to a linked clone if you deleted the original VM?

Challenge

Package srv-linux-2 into a single .ova file that could be carried on a USB drive. Prove the file is complete: delete srv-linux-2 with all its files, then bring it back to life from the .ova. Better still, swap files with a classmate and boot each other's server. Record the size of the .ova and compare it with the VM's folder.

Everything you need is in VirtualBox's File menu: one item exports, one imports. When exporting, pick the MAC address policy that strips the addresses, for the same reason you generated new ones when cloning.

Checkpoint

  • After the restore, srv-linux serves its web page again.
  • Both servers answer at the same time on .10 and .11.
  • My imported appliance boots.

Hand in: a screenshot of the snapshot tree, both web pages side by side, the sizes you recorded (original, full clone, linked clone, .ova), and a short answer: why is a linked clone so much smaller, and what happens to it if the original VM is deleted?

When you finish, keep srv-linux-2 powered off to save memory.

How this looks in a real cloud

A snapshot is a "volume snapshot". A clone or .ova is a machine image, the template that every new cloud server is launched from. The identity clean-up you did by hand in Part B is done automatically by a tool called cloud-init the first time a Linux image boots in the cloud.

Lab 4: Control and resize without the mouse

Cloud engineers manage thousands of servers and never click on any of them. Today you stop using the VirtualBox window and start giving orders by command, then you change how powerful your server is.

Your job today. The company is growing. Run the server with no window, make it smaller and bigger, measure the difference, and give it a second disk for data.

Before you type anything: three ideas

VBoxManage, headless, vCPU.

Most commands today are labelled "Host computer's terminal". Type those on the host, the physical computer you are sitting at, not inside a VM.

GuidedPart A: Meet VBoxManage

  1. Open a terminal on your host. On macOS and Linux, VBoxManage works immediately. On a Windows host, first move into VirtualBox's folder:

    cd "C:\Program Files\Oracle\VirtualBox"
    .\VBoxManage.exe --version
    What does this mean?
    cd "folder"
    "Change directory": walk into that folder. The quotes are needed because the path contains a space.
    .\VBoxManage.exe
    .\ means "the program in this folder". On Windows, type .\VBoxManage.exe everywhere this page says VBoxManage.
  2. Look around.

    VBoxManage list vms
    VBoxManage list runningvms
    VBoxManage showvminfo "srv-linux"
    What does this mean?
    list vms
    Every VM VirtualBox knows about.
    list runningvms
    Only those switched on right now.
    showvminfo "name"
    Everything about one VM: memory, CPUs, disks, adapters. It is long; scroll up to read it.
  3. Shut down srv-linux if it is running. Now start it headless: no window appears at all. Give it half a minute to boot, then connect the way you would reach a cloud server.

    VBoxManage startvm "srv-linux" --type headless
    ssh youruser@192.168.56.10
    What does this mean?
    startvm
    Power on a VM.
    --type headless
    Run it invisibly in the background.
    ssh ...
    Your host can use SSH too, thanks to the host-only network. It will ask for your password, because your key lives on the Windows client, not on the host.
  4. Type exit to leave the server, then switch it off politely from the host.

    VBoxManage controlvm "srv-linux" acpipowerbutton
    What does this mean?
    controlvm
    Send an order to a VM that is running.
    acpipowerbutton
    Pretend somebody pressed the physical power button. Linux notices and shuts down cleanly. (The other option, poweroff, is like pulling the plug.)

GuidedPart B: Smaller, then bigger

  1. With the server powered off, shrink it and start it.

    VBoxManage modifyvm "srv-linux" --memory 1024 --cpus 1
    VBoxManage startvm "srv-linux" --type headless
    What does this mean?
    modifyvm
    Change the settings of a VM that is switched off.
    --memory 1024
    Give it 1024 MB (1 GB) of RAM.
    --cpus 1
    Give it one virtual CPU.
  2. Check from inside that the server really sees the new size.

    free -h
    nproc
    What does this mean?
    free -h
    Show memory, human-readable. Read the total column of the Mem line. It is a little less than what you gave, because the kernel keeps some for itself.
    nproc
    "Number of processors" the system can see.
  3. Power off again, grow it with --memory 2048 --cpus 2, start it and re-check. Making one machine bigger is called vertical scaling. Notice you had to switch the machine off to do it; most clouds have the same rule.

  4. Put the CPUs to work for one minute, and watch your host's Task Manager (or Activity Monitor) at the same time. Write down the bogo ops/s number at the end.

    sudo apt install -y stress-ng
    stress-ng --cpu 2 --timeout 60s --metrics-brief
    What does this mean?
    stress-ng
    A tool that keeps the computer busy on purpose, for testing.
    --cpu 2
    Start two workers, one per vCPU.
    --timeout 60s
    Stop after 60 seconds.
    --metrics-brief
    Print a score at the end. "Bogo ops" are made-up units; only compare them with other runs on the same machine.

Your turn

Task 1. Cloud providers sell cheap "shared CPU" servers. Find out how: limit srv-linux to 50% of each CPU it owns, run the same stress test, and compare the scores. Then remove the limit.

VirtualBox calls this the CPU execution cap. It can be changed while the VM runs, so which VBoxManage sub-command do you need: modifyvm or controlvm?
VBoxManage controlvm "srv-linux" cpuexecutioncap 50, run the test, then the same command with 100. The score should roughly halve.

Task 2. Are you overcommitting? Find how many CPU threads and how much RAM the host computer really has. Add up the vCPUs and memory given to all your VMs. Is the total bigger than what really exists?

showvminfo shows the numbers for each VM. On a Windows host, Task Manager > Performance shows "Logical processors" and total memory.
Promising more vCPUs than exist is fine: CPUs are shared by taking turns, and most servers are idle most of the time. See overcommitment. Memory is different. RAM given to a running VM is taken from the host, so never start more VMs than your RAM can hold, or everything crawls.

GuidedPart C: A second disk for data

  1. Shut down srv-linux. Open Settings > Storage, select the controller that holds the existing disk, click the Add hard disk icon, then Create. Choose type VDI, leave Pre-allocate Full Size unticked, size 2 GB, name data. Select it, attach it, start the VM.

  2. Find the new disk. Look for one of about 2G with nothing listed under it.

    lsblk
    What does this mean?
    lsblk
    "List block devices", meaning disks. sda is the first disk, sdb the second. Entries such as sda1, sda2 are partitions, the sections inside a disk.
    NAME   SIZE TYPE MOUNTPOINTS
    sda     25G disk
    ├─sda1   1M part
    └─sda2  25G part /
    sdb      2G disk
    Stop and check. In the next step, make sure the name you type is the empty 2G disk. Formatting sda would erase your server. (If that ever happens, you know about snapshots now.)
  3. Prepare the disk and connect it to a folder.

    sudo mkfs.ext4 /dev/sdb
    sudo mkdir /mnt/data
    sudo mount /dev/sdb /mnt/data
    df -h /mnt/data
    What does this mean?
    mkfs.ext4 /dev/sdb
    "Make filesystem": format the disk with ext4, the standard Linux format. /dev/ is where Linux lists its hardware.
    mkdir /mnt/data
    Create an empty folder that will become the door to the disk.
    mount
    Linux has no D: drive. Instead you mount a disk onto a folder; whatever you save in that folder lands on that disk.
    df -h /mnt/data
    Show the size and free space of the disk behind that folder.
  4. On the host, look at the size of the file data.vdi. It is a few megabytes, not 2 GB. The file only grows as data is written. Cloud providers call this thin provisioning.

Challenge

Without touching the VirtualBox window: take a snapshot of srv-linux named after-lab4, list its snapshots, create a file on the server, then go back to the snapshot and prove the file has vanished.

Every VBoxManage sub-command explains itself: try VBoxManage snapshot with nothing after it and read the usage text. Remember from Lab 3 what state the VM must be in before a restore.

Checkpoint

  • I can start, reach and stop srv-linux without opening the VirtualBox window.
  • nproc and free -h match the sizes I set.
  • /mnt/data shows a 2 GB disk.

Hand in: your two stress-test scores with one sentence explaining the difference, your overcommitment numbers, and the output of df -h /mnt/data and of your snapshot list.

How this looks in a real cloud

modifyvm is choosing a server size from a price list. A CPU cap is how "shared" servers are sold cheaply. Overcommitment is the business model of the cloud: providers sell more virtual CPUs than they own, because customers rarely use them all at once. Your data.vdi is a "block storage volume". And VirtualBox is a type 2 hypervisor; data centres use type 1, but the ideas you practised are the same.

Lab 5: Servers from a recipe

It took you four labs to build one server by hand. Nobody runs a cloud that way. Today you write the server down as a recipe and let a tool cook it, as many times as you like.

Your job today. Build a web server with a single command. Destroy it. Rebuild it, identical. Then serve two at once. Your final challenge is to recreate your entire Lab 2 server without typing a single command inside it.

Before you type anything: five ideas

Vagrant, Vagrantfile, box, provisioning, infrastructure as code.

You need a good internet connection: the first run downloads a box of several hundred megabytes. Shut down win-client, srv-linux and srv-linux-2 first to free memory. Today you test from your host's browser.

GuidedPart A: One server from one file

  1. First check whether Vagrant is already installed on the host by running the command below; on a computer-lab PC your instructor may have installed it for you. If you get command not found, install the latest Vagrant from HashiCorp's official download page (search for "Vagrant install"). Installing needs administrator rights on the host, so on a lab PC ask your instructor. With VirtualBox 7.2 you need Vagrant 2.4.9 or newer. Close and reopen your terminal, then check:

    vagrant --version
  2. Make a folder for the project and go into it. Every Vagrant project lives in its own folder.

    mkdir vagrant-lab
    cd vagrant-lab
  3. Inside that folder, create a text file named exactly Vagrantfile, capital V, no extension. Use a plain text editor such as Notepad or VS Code, not Word. (In Notepad, choose "All files" as the type when saving, or you will get Vagrantfile.txt.)

    Vagrant.configure("2") do |config|
      config.vm.box = "bento/ubuntu-24.04"
      config.vm.hostname = "web-auto"
      config.vm.network "private_network", ip: "192.168.56.20"
    
      config.vm.provider "virtualbox" do |vb|
        vb.name = "web-auto"
        vb.memory = 1024
        vb.cpus = 1
      end
    
      config.vm.provision "shell", inline: <<-SHELL
        apt-get update
        apt-get install -y nginx
        echo "<h1>Built automatically - YOUR NAME</h1>" > /var/www/html/index.html
      SHELL
    end
    What does this mean?
    Vagrant.configure("2") do ... end
    The wrapper every Vagrantfile needs. Everything between do and the final end describes your machine.
    config.vm.box
    The ready-made disk to start from: Ubuntu 24.04, already installed. This replaces the whole OS installation you did by hand.
    config.vm.hostname
    The computer's name. You did this with hostnamectl in Lab 3.
    private_network, ip:
    Add a host-only adapter with a static address. This one line replaces all of Lab 1, Part E.
    vb.memory / vb.cpus
    The size of the machine, as in Lab 4.
    provision "shell"
    Commands to run inside the new server the first time it starts. They run as administrator, so no sudo is needed. Everything between <<-SHELL and SHELL is the script.
    apt-get
    The older, script-friendly brother of apt. Same job.
  4. Build it, and time how long it takes. Read the messages scrolling by: you will recognise every stage.

    vagrant up

    When it finishes, browse to http://192.168.56.20 on your host. Open the VirtualBox window too: web-auto is there, running headless, a VM like any other.

  5. Log in. You do not need to know any password or address; Vagrant set up a key pair for you.

    vagrant ssh
    exit
  6. Destroy the server completely. Check that the web page is gone. Then rebuild it.

    vagrant destroy -f
    vagrant up
    What does this mean?
    vagrant destroy -f
    Switch off the VM and delete it and its disk. -f = force, do not ask "are you sure?". Your Vagrantfile is untouched.
    The second build is much faster because the box is already downloaded, and the result is identical. You no longer need to protect this server, because you can always make another. Engineers call this treating servers as cattle, not pets.
  7. Other commands worth knowing: vagrant status (is it running?), vagrant halt (shut down but keep), vagrant provision (run the script again).

Your turnPart B: Two servers

The website is popular and one server is not enough. Change your Vagrantfile so that a single vagrant up creates two servers, web1 at 192.168.56.21 and web2 at 192.168.56.22, each showing its own name on its page. Give each 768 MB of memory. Run vagrant destroy -f before you start editing.

Vagrant can define several machines in one file with config.vm.define "name" do |node| ... end. Inside the block, write node.vm... where you wrote config.vm... before.
Do not copy and paste the block twice. A Vagrantfile is written in the Ruby language, and Ruby can repeat: (1..2).each do |i| ... end. Inside a text in double quotes, #{i} becomes the number, so "web#{i}" gives web1, then web2.
Vagrant.configure("2") do |config|
  config.vm.box = "bento/ubuntu-24.04"

  (1..2).each do |i|
    config.vm.define "web#{i}" do |node|
      node.vm.hostname = "web#{i}"
      node.vm.network "private_network", ip: "192.168.56.2#{i}"

      node.vm.provider "virtualbox" do |vb|
        vb.memory = 768
        vb.cpus = 1
      end

      node.vm.provision "shell", inline: <<-SHELL
        apt-get update
        apt-get install -y nginx
        echo "<h1>Hello from web#{i} - YOUR NAME</h1>" > /var/www/html/index.html
      SHELL
    end
  end
end

Run vagrant up, then vagrant status. To log in to one of them: vagrant ssh web1.

How much work would ten servers be? Change one character. Adding more machines, instead of a bigger machine, is horizontal scaling: the opposite of Lab 4. Clean up with vagrant destroy -f.

GuidedPart C: A VM versus a container

  1. Start srv-linux again. Run one more web server on it, this time in a container, and time it.

    sudo apt install -y docker.io
    time sudo docker run -d --name web-c -p 8081:80 nginx
    What does this mean?
    docker.io
    Ubuntu's package name for Docker, the most common container tool.
    time
    Measure how long the rest of the line takes.
    docker run nginx
    Download the ready-made nginx image, if needed, and start a container from it.
    -d
    "Detached": run in the background.
    --name web-c
    A name of your choice for the container.
    -p 8081:80
    Port forwarding again: the server's port 8081 leads to the container's port 80.
  2. Browse to http://192.168.56.10:8081. Now start a second one with sudo docker run -d --name web-c2 -p 8082:80 nginx and notice: once the image is downloaded, a new container starts in about a second. Your VM needed minutes.

  3. Find out why.

    uname -r
    sudo docker exec web-c uname -r
    What does this mean?
    uname -r
    Print the version of the kernel this system runs on.
    docker exec web-c ...
    Run a command inside the container named web-c.
    The two answers are identical. A VM carries a complete operating system of its own on pretend hardware. A container is only a fenced-off program borrowing the kernel of the machine it runs on. Lighter and faster, but less separated, and it is why you cannot run a Windows container on this Linux server.

ChallengeThe capstone

Prove you understood every lab. Write one Vagrantfile that rebuilds your complete Lab 2 server with no manual steps. After vagrant destroy -f and vagrant up, all of this must be true:

  • The server is called office-srv and has the static address 192.168.56.40.
  • Your web page, with your name, opens at http://192.168.56.40.
  • A password-protected share called labshare can be opened from win-client.
  • The firewall is on and allows only SSH, the website and Samba.
  • vagrant ssh still works afterwards.
A provisioning script cannot answer questions, so every command must run without asking anything. ufw enable asks for confirmation unless you write ufw --force enable. And remember Lab 2's lesson about which rule must exist before the firewall switches on, or Vagrant locks itself out.
smbpasswd normally asks you to type a password twice. With -s it reads from a pipe instead: (echo "secret"; echo "secret") | smbpasswd -s -a vagrant. The box's built-in user is named vagrant. To add lines to the end of smb.conf from a script, use cat >> /etc/samba/smb.conf <<EOF ... EOF.

Checkpoint

  • vagrant up gives me a working web page with no manual steps.
  • vagrant status showed web1 and web2 running together.
  • I can explain the difference between a VM and a container in two sentences.

Hand in: your capstone Vagrantfile, a screenshot of vagrant status with both web pages open, your build times (first vagrant up, second vagrant up, container), and a paragraph: what are two advantages of describing a server in a file instead of building it by hand?

How this looks in a real cloud

A box is a machine image. The Vagrantfile plays the part that tools like Terraform play in production. The provisioning script is the "user data" you hand to a new cloud server. Your loop in Part B is a miniature auto-scaling group: one definition, stamped out as many times as demand requires.

What you built, and what the cloud calls it

Everything you did on one ordinary computer has a twin on a real cloud platform. The names differ between providers; the ideas do not.

You did this in VirtualBoxLabThe cloud calls it
Created a VMbefore 1Launching an instance
NAT adapter1Private subnet with a NAT gateway
Host-only / internal network1Private or isolated subnet in a virtual network (VPC)
ufw rules2Security groups, firewall rules
Port forwarding2Load balancer, public endpoint
SSH key pair2Instance key pair
Snapshot3Volume snapshot
Clone, .ova export3Machine image, template
Fixing the clone's identity3cloud-init (Linux), Sysprep (Windows)
modifyvm --memory --cpus4Changing instance type (vertical scaling)
CPU execution cap4Shared and burstable instances
Extra .vdi disk4Block storage volume
VBoxManage4Cloud command-line tools and APIs
Vagrantfile5Infrastructure as code (Terraform and similar)
Two servers from a loop5Auto-scaling group (horizontal scaling)
Docker container5Container services, Kubernetes

Final clean-up. Delete srv-linux-2 and leftover clones, run vagrant destroy -f in your project folder, and remove old .ova files. Keep srv-linux and win-client with their snapshots.

When something goes wrong

First, read the error message slowly. Second, check where you typed the command. Third, look here.

ProblemLikely cause and fix
I type my password after sudo and nothing appearsNormal. Linux hides passwords completely. Type it and press Enter.
command not foundA typo, or the program is not installed yet. Check the spelling; check you did the install step.
Permission deniedYou forgot sudo. Press ↑, go to the start of the line, add it.
I cannot leave nanoCtrl+X. If it asks to save, press Y then Enter.
The command never finishesCtrl+C stops it. A status screen closes with q.
My mouse is trapped in the VM windowPress the host key: Right Ctrl (left ⌘ on a Mac).
VM will not start: "VT-x is not available" or "AMD-V is disabled"Hardware virtualization is switched off in the host's BIOS/UEFI settings. On your own laptop, ask your instructor to help you enable it. On a computer-lab PC, do not change it yourself: tell your instructor or the lab technician.
VM is extremely slow and shows a green turtle iconA Windows host running Hyper-V features (WSL2, Memory Integrity). VirtualBox works, but slowly. Ask your instructor before disabling them.
Server pings the client and gets no replyWindows Firewall. Add the rule from Lab 1, Part C.
netplan apply shows an errorIndentation. Spaces only, two per level, exactly as shown.
SSH says "REMOTE HOST IDENTIFICATION HAS CHANGED"The server's host keys changed, or another machine now uses that address. On the client run ssh-keygen -R 192.168.56.10 and connect again.
Locked out after ufw enableSSH was not allowed first. Log in on the VirtualBox console and run sudo ufw allow OpenSSH.
Windows rejects my password for the shareUse the Samba password set with smbpasswd. Check sudo systemctl status smbd.
Both servers unreachable after cloningTwo machines share 192.168.56.10. Power one off and finish Lab 3, Part B.
vagrant up: "VirtualBox version not supported"Your Vagrant is older than your VirtualBox. Install the latest Vagrant.
vagrant up does nothing useful; "Vagrantfile not found"You are in the wrong folder, or the file is named Vagrantfile.txt.
The host computer freezes or becomes very slowToo many VMs running. Power off those you are not using. Headless VMs still use RAM: check VBoxManage list runningvms.