Ubuntu

What it is

Ubuntu is a Linux distribution published by Canonical: the Linux
kernel and the thousands of programs around it, built, tested and
patched by one publisher on a calendar it announces in advance. It
is derived from Debian, so it uses the same `.deb` packages and the
same `apt` tool to install them, and it adds its own release rhythm,
its own defaults and a company that sells support. The same archive
feeds desktops, cloud images and bare-metal servers; this entry is
about the server.

Releases are named after the year and month they shipped, so 24.04
came out in April 2024. A new one appears every six months, and
every second year the April release is a long-term support release,
marked LTS. The difference is the length of the promise. An interim
release gets nine months of fixes. An LTS release gets five years of
standard security maintenance, which an Ubuntu Pro subscription
extends to ten through Expanded Security Maintenance, and a Legacy
add-on to fifteen.

On servers Ubuntu is usually weighed against two neighbours. Debian
is its parent: the same packaging, with a community project rather
than a company behind it. Red Hat Enterprise Linux and its rebuilds,
such as AlmaLinux and Rocky Linux, form the other family, with their
own package format and their own tools. Choosing among the three is
mostly choosing a calendar, a support contract and the habits the
team already has.

Why it matters

Once a team has picked Ubuntu, the decision that remains is which
Ubuntu, and that decision sets the maintenance budget for years.

Servers belong on LTS releases. An interim release expires nine
months after it ships, so a server that tracks them has to go
through a release upgrade twice a year to stay patched, which is a
lot of change for the privilege of slightly newer packages. Interim
releases are where you try what the next LTS will bring.

An LTS release has a firm upgrade path, and it is narrower than
people expect. A server moves from one LTS to the next one only,
without skipping any: from 20.04 the way to 24.04 runs through
22.04. The release tool, `do-release-upgrade`, offers the jump only
after the new release's first point release, so 24.04 became the
default upgrade target once 24.04.1 was out. A fleet that falls two
releases behind therefore needs two full upgrades per machine, each
with its own testing.

Ubuntu Pro is the other half of the decision. It adds the five extra
years of security fixes, Livepatch for applying kernel fixes without
a reboot, Landscape for managing machines as a fleet, and FIPS 140-3
and CIS hardening tooling for regulated work. It is free on up to
five machines, which covers a small business or a lab; beyond that
it is a paid subscription. The honest comparison is the subscription
against the cost of upgrading on schedule. Five years is enough for
a server that gets rebuilt from code; ten years suits the appliance
nobody wants to touch, where paying for patches beats migrating in a
hurry.

Finally, Ubuntu has its own way of doing ordinary things, and a team
arriving from another distribution should budget time to learn them.
Network interfaces are configured in YAML through Netplan. The
firewall is usually managed through `ufw`, a front end that ships
disabled. Security updates install themselves every day through
`unattended-upgrades`, which is on by default. Each of these is
pleasant once known and surprising before.

What it looks like in practice

A fleet of Ubuntu Linux VPS instances that had grown one machine at
a time is the clearest picture of what an unmanaged estate turns
into. The configurations had drifted apart, and nobody had designed
how the machines would be recovered or how they were networked. The
work brought all thirty instances under management as one fleet:
backups went in with restore procedures that were actually tested,
and every instance's network configuration was moved onto the same
hardened baseline. Recovery paths were now tested rather than
assumed, and the downtime risk dropped.

Databases ran on the same footing elsewhere. Ten servers running
PostgreSQL 14 and MS SQL Server 2019 on Ubuntu 20.04 LTS VPS
instances were designed, deployed and maintained as one estate.
Backups ran through `pg_dump` and SQL Server Agent jobs, with
retention policies and offsite storage. Prometheus and Grafana
tracked CPU, memory, disk I/O and query metrics. Both the databases
and the Ubuntu OS were patched on a regular schedule, the firewalls
were configured with UFW, and access was enforced with role-based
access control. Disaster recovery was documented down to
point-in-time restores and failover. The ten servers ran at

99.9% uptime, query response times improved by 30%, automation cut
manual maintenance by 50% and freed more than ten hours a month, and
the estate absorbed a 40% rise in user traffic without degrading.

Web hosting is the plainest case. Forty websites ran on Ubuntu Linux
hosting servers behind a mix of Apache and Nginx, and the job was
the configuration, the performance tuning and the maintenance that
keeps sites reliable under real traffic. All forty ran with high
availability and good performance.

Where it goes wrong

The updater installs a kernel fix, and the server goes on running
the old kernel. `unattended-upgrades` installs security updates once
a day, but its `Automatic-Reboot` option defaults to false, so a new
kernel waits on disk until someone restarts the machine, and
services keep using the old copies of updated libraries until they
restart too. The file `/var/run/reboot-required` appears when a
reboot is pending. Decide on a policy instead of leaving it to
chance: alert on that file, schedule reboots in a quiet hour, or use
Livepatch for the kernel. A scheduled reboot is two lines in
`/etc/apt/apt.conf.d/50unattended-upgrades`:

```text
Unattended-Upgrade::Automatic-Reboot "true";
Unattended-Upgrade::Automatic-Reboot-Time "03:30";
```

A server on an old LTS release stops getting security fixes while
everything on it keeps working. Standard support for 20.04, for
example, ended in May

2025; the machine boots, serves and reports green exactly as before.
Record each server's release next to its end-of-support date, and
before that date either attach Ubuntu Pro, which carries 20.04 to
May 2030, or upgrade one LTS at a time.

A new `ufw` rule goes in, and traffic flows exactly as before. `ufw`
evaluates rules in order and the first match wins, and a plain `ufw
allow` or `ufw deny` lands after the rules already there. An older,
broader rule above it decides first. List the rules with their
positions and put the specific rule where it belongs:

```sh
sudo ufw status numbered
sudo ufw insert 1 deny from 203.0.113.7
```

A package from a third-party repository stays on an old, vulnerable
version while the rest of the system is current.
`unattended-upgrades` installs only from the origins listed in its
`Allowed-Origins` setting, and out of the box that list holds only
official Ubuntu repositories. A vendor repository or a PPA added
later is updated by hand or not at all. Add its origin to the list
deliberately, or put that package on a patch schedule of its own.

Further reading

  * Ubuntu Server documentation — Canonical's own guides for
installing, configuring and running Ubuntu on servers.

  * Ubuntu release cycle — the six-month cadence, the LTS calendar
and the end date of every release.

  * Automatic updates — how `unattended-upgrades` chooses origins,
and the reboot options that decide when fixes take effect.

  * Firewall — `ufw` from first enable to numbered rules, with the
netfilter layer underneath.

  * How to upgrade your Ubuntu release — `do-release-upgrade`, the
one-LTS-at-a-time rule and the point-release wait.

  * Netplan documentation — the YAML network configuration Ubuntu
servers use, including `netplan try` with its automatic rollback.

hThis page on the web	URL:https://engineer.company/glossary/ubuntu/	engineer.company	70
