Installer walkthrough

Why this matters in production

The installer asks for about ten decisions. Most are obvious. A few have long-lasting consequences. This lesson walks through them in order.

Installer flow

flowchart LR
  A[Boot from ISO] --> B[Accept EULA]
  B --> C[Target hard disk]
  C --> D[Filesystem: ext4 / xfs / zfs / btrfs]
  D --> E[Country / timezone / keyboard]
  E --> F[Password + email]
  F --> G[Network: hostname, IP, gateway, DNS]
  G --> H[Summary]
  H --> I[Install]
  I --> J[Reboot]

Step-by-step

Target hard disk

The installer shows available disks and asks which to use. Multi-disk installs offer:

• Single disk — simple, but you have no redundancy for the OS.
• ZFS mirror — recommended. Two SSDs in a ZFS mirror; the OS is redundant against one disk failure.
• ZFS RAIDZ1 — three or more disks, single-parity RAID. Better capacity, slower writes.
• Software RAID (mdraid) — supported but not recommended; ZFS is the modern choice.
• Hardware RAID — for OS only when ZFS root is not used.

Filesystem options

Network configuration

The installer asks for:

• Management interface — the NIC that will become the management network. Pick a NIC that is on the management VLAN or dedicated network.
• Hostname — fully qualified (e.g., pve-01.lab.example.com).
• IP address / CIDR — static, in the management subnet.
• Gateway — default router for management traffic.
• DNS server — typically the local DNS resolver; sometimes the corporate AD-integrated resolver.

Country, timezone, keyboard

• Country — used to choose the closest APT mirror.
• Timezone — UTC is conventional; pick a regional zone only if the team strongly prefers local time for log readability.
• Keyboard layout — affects the installer’s input handling; not the host’s runtime.

Password + email

The installer sets the root password and a contact email. The email appears in subscription-related emails and is not used as the root’s email address. Pick a real, monitored inbox.

Summary

The installer shows a summary screen. Read every line. Common mistakes caught here:

• Wrong target disk (will wipe it).
• Wrong IP address (will not be reachable).
• Wrong hostname (will not cluster later).

Post-install reboot

After install completes, the system reboots. Verify:

• Console login works with the root password.
• The management IP responds to ping.
• ssh root@<ip> works.
• The web UI is reachable at https://<ip>:8006.

If any of these fail, do not proceed. Fix now; don’t blame the cluster.

GUI walkthrough

The web UI at https://<ip>:8006 shows the new node’s dashboard. The summary panel displays:

• Uptime, load average, memory usage.
• Storage status.
• Network status.
• Subscription status (will be “nope” without a key).

CLI walkthrough

pveversion -v

cat /etc/network/interfaces && cat /etc/hosts

lsblk

Production considerations

• Take a screenshot of the install summary screen. Attach it to the cluster record.
• Save the installer’s autoinstall config if you used unattended install. This lets you rebuild a node identically later.
• Document the disk model and firmware version of the target disk. Useful when a replacement is needed.

Common mistakes

• Picking the wrong disk as the target. Always confirm before clicking Install.
• Setting a hostname that does not resolve in DNS.
• Using DHCP for the management IP. Always static for production.
• Skipping the ZFS root option for “simplicity.” You will regret it when an OS upgrade breaks the boot loader.

Key takeaways

• ZFS root is the modern default; it costs nothing and gives checksumming + snapshots.
• DNS must work at install time or cluster join fails.
• Verify the install summary before committing.

Knowledge check