basforge

Search titles, text and the “If it doesn’t work” cases.

    RU

    cloud-init: a server configured the moment it’s created

    Paste a cloud-config user-data file into the provider’s panel when you create the server, and on first boot it does what the set’s manuals do: a user with a key, key-only login, updates with reboots, swap, a firewall. You log in as example-user from the start.

    How it works

    provider panel: the "User data" / "Cloud-init" field
            │ at server creation
            ▼
    first boot: cloud-init reads the file → modules in order:
      users, ssh, write_files, swap, package_update/upgrade, packages, runcmd, reboot
            │
            ▼
    /var/log/cloud-init-output.log — output of every command

    cloud-init is in every Ubuntu cloud image. The file runs once, on first boot. Running it again takes sudo cloud-init clean --reboot, which reapplies everything in the file.

    Whether a provider supports user data, check its panel: the field is called “User data”, “Cloud-init” or “Startup script”. No field — set up by hand, following the set (A new Ubuntu server).

    The file, block by block

    user-data
    #cloud-config

    The first line is a required format marker. Without it cloud-init takes the file for a script or skips it.

    hostname: web1
    timezone: Europe/Berlin    # your own zone

    Hostname and time zone, as in First login to a server. The time zone decides when “4:00” comes for the automatic reboot.

    users:
      - name: example-user
        groups: [sudo]
        shell: /bin/bash
        lock_passwd: false
        hashed_passwd: "$6$Qm…$…"
        ssh_authorized_keys:
          - ssh-ed25519 AAAAC3Nza… laptop-2026
          - ssh-ed25519 AAAAC3Nza… desktop-2026
    Key part

    Your own sudo user with two keys from different devices. The password is given as a hash: sudo and the provider console need it, but it won’t log you in over SSH. Make the hash on your computer with openssl passwd -6, so the password itself never lands in the file. lock_passwd: false — the password isn’t locked.

    default:lock_passwd trueshell /bin/sh

    ssh_pwauth: false
    disable_root: true

    ssh_pwauth: false — cloud-init writes PasswordAuthentication no into its own 50-cloud-init.conf. disable_root: true — root’s keys are replaced with a hint to log in as example-user. No root password is set at all.

    default:ssh_pwauth as in the provider’s imagedisable_root true

    write_files:
      - path: /etc/ssh/sshd_config.d/00-hardening.conf
        content: |
          PasswordAuthentication no
          KbdInteractiveAuthentication no
          PubkeyAuthentication yes
          PermitRootLogin no
          AllowUsers example-user
          MaxAuthTries 3
          LoginGraceTime 30
          X11Forwarding no
          AllowAgentForwarding no
          ClientAliveInterval 300
          ClientAliveCountMax 2

    Your own sshd file — the same as in sshd_config. write_files runs before packages and runcmd, so the settings are in place by the first reboot.

      - path: /etc/apt/apt.conf.d/52unattended-upgrades-local
        content: |
          Unattended-Upgrade::Automatic-Reboot "true";
          Unattended-Upgrade::Automatic-Reboot-Time "04:00";
          Unattended-Upgrade::Automatic-Reboot-WithUsers "true";
          Unattended-Upgrade::Remove-Unused-Kernel-Packages "true";
          Unattended-Upgrade::Remove-New-Unused-Dependencies "true";
          Unattended-Upgrade::Remove-Unused-Dependencies "true";
          Unattended-Upgrade::AutoFixInterruptedDpkg "true";
      - path: /etc/systemd/journald.conf.d/00-size.conf
        content: |
          [Journal]
          SystemMaxUse=500M
          SystemKeepFree=1G
      - path: /etc/sysctl.d/99-swappiness.conf
        content: |
          vm.swappiness=10

    Auto-updates with reboots (unattended-upgrades), the journal limit and swappiness (Swap, journal and time) — as files too.

    swap:
      filename: /swapfile
      size: 2G

    Swap via the built-in module: it creates the file, sets permissions and writes the /etc/fstab line.

    package_update: true
    package_upgrade: true
    package_reboot_if_required: true
    packages:
      - ncdu
      - htop

    Updates at creation and a reboot if the kernel changed. Packages — what comes in handy in the first hour: ncdu for disk, htop for memory.

    default:package_update falsepackage_upgrade false

    runcmd:
      - ufw default deny incoming
      - ufw default allow outgoing
      - ufw allow OpenSSH
      - ufw allow 80,443/tcp
      - ufw --force enable
      - sysctl --system
      - systemctl restart ssh

    The firewall, as in The ufw firewall: the SSH rule before enable. --force skips the y/n prompt. The last line restarts sshd so it reads our file.

      - curl -fsSL https://get.docker.com | sh

    Docker and CrowdSec could go here as runcmd commands too, but an error in them only shows in the log. While there are few steps, do them by hand after the first login (Docker on Ubuntu, CrowdSec against SSH brute force).

    What’s not here

    What Why
    a plain-text password (passwd, chpasswd) user data is stored at the provider, and any process on the server can read it from the metadata address 169.254.169.254
    API keys, tokens same reason: put secrets on the server after creation (Secrets in Docker stacks)
    settings that will change the file runs once; live configs belong in git (Server configs in git)

    If it doesn’t work

    The server was created, but nothing from the file was applied

    Cause
    The first line isn’t #cloud-config — without it the file isn’t treated as a config. Or a YAML error: cloud-init skips the whole file.
    Fix
    On the server: sudo cloud-init schema --system --annotate shows the error with its line; the log is /var/log/cloud-init.log.

    Logging in as example-user works, but sudo asks for a password that doesn’t exist

    Cause
    The user has lock_passwd: true and no hashed_passwd: there’s no password, and sudo requires one.
    Fix
    Set hashed_passwd (a hash from openssl passwd -6) and lock_passwd: false — or run passwd example-user as root through the provider console.

    One runcmd command failed, the rest worked

    Cause
    runcmd doesn’t stop on errors: each line runs on its own.
    Fix
    Look at /var/log/cloud-init-output.log — it has every command’s output in order; run the fixed command by hand.

    A minute after creation SSH lets nobody in

    Cause
    cloud-init is still working: installing updates and rebooting the server (package_reboot_if_required).
    Fix
    Wait 3–5 minutes. Check the status with cloud-init status --wait through the provider console.

    How to check

    3–5 minutes after creation, from your computer:

    ssh [email protected] 'cloud-init status --long'
    # status: done
    # extended_status: done
    # boot_status_code: enabled-by-generator
    # detail: DataSourceNoCloud …
    # errors: []

    errors: [] means every module finished. Then check from outside, also from your computer:

    # open — only 22; 80 and 443 are closed: ufw lets them through, but nothing listens on them yet;
    # every other port is filtered
    nmap -Pn --top-ports 1000 203.0.113.10
    # root is refused
    ssh [email protected]
    # [email protected]: Permission denied (publickey).
    # swap is on, the journal has a limit, the clock is synced
    ssh [email protected] 'swapon --show; journalctl --disk-usage; timedatectl show -p NTPSynchronized --value'