
Most failed ISPConfig installs have nothing to do with ISPConfig. They fail because the server was half-prepared. The hostname doesn’t resolve, a stray web server is already bound to port 80, or the installer was run over a flaky SSH session that dropped halfway through. The panel is the easy part. Preparation decides whether the install works.
Ubuntu 26.04 LTS (“Resolute Raccoon”) shipped in April 2026 with Linux kernel 7.0, PHP 8.5 in the default repositories and cgroup v2 only. ISPConfig added official support for it in release 3.3.2, including the Dovecot 2.4 configuration syntax that this Ubuntu release uses. The project now lists Ubuntu 22.04 to 26.04 LTS as recommended platforms, next to Debian 11 to 13. Anyone who hit a wall trying to install ISPConfig on Ubuntu 26.04 before September 2026 was working with a release that didn’t know the platform.
This guide walks through a clean install of the ISPConfig hosting control panel on a fresh Ubuntu 26.04 server. It covers the reasoning behind each step, the choices that matter (Apache or Nginx, local or external DNS, which services to enable), and the problems that show up on real production boxes. Updating, hardening, tuning and troubleshooting are included, because a panel that installs fine and then falls over at the first traffic spike isn’t much use.
What ISPConfig Actually Installs
ISPConfig is a web hosting control panel for Linux. It manages websites, mail, DNS, FTP, databases, shell users and cron jobs from one interface. It does this by configuring the underlying services (Apache or Nginx, Postfix, Dovecot, PowerDNS or BIND, MariaDB or MySQL, Pure-FTPd, Rspamd) and writing their config files itself.
That last point matters. ISPConfig takes ownership of those configs. If you hand-edit /etc/postfix/main.cf and a later update rewrites it, your change is gone. Keep this in mind from day one, because it shapes how you customize things later.
The official FAQ describes the Autoinstaller as the easiest route. It supports Debian 11 to 13 and Ubuntu 22.04 to 26.04, handles Apache or Nginx, multiple PHP versions and ARM servers. Unless you need an unusual layout, use it. The old manual “Perfect Server” tutorials are long, and a typo in one of fifty steps is how a weekend disappears.
Requirements and Planning
Hardware and network
A realistic starting point for a small multi-site server:
- 2 vCPU and 4 GB RAM minimum if mail with Rspamd and ClamAV is enabled. ClamAV alone can eat 1 GB or more.
- 40 GB SSD or NVMe. Mail and website data grow faster than expected.
- A static public IPv4 address, plus IPv6 if the provider offers it.
- A fresh, minimal Ubuntu 26.04 LTS install with nothing else on it.
A server with 1 GB RAM can run the panel, but turn off ClamAV and be ready for swap activity. Don’t pretend it’s comfortable.
Decisions to make before touching the keyboard
Apache or Nginx? Apache is the more forgiving choice for mixed customer sites because .htaccess just works. Nginx uses less memory under high concurrency and suits servers that mostly run a few busy PHP applications. Either works with ISPConfig, and the Autoinstaller lets you choose.
Will this server also be the DNS authority? If yes, you need glue records at your registrar and open port 53. If your DNS lives at Cloudflare or a registrar, skip the DNS service and save yourself one more attack surface.
Will it send mail? Port 25 outbound is blocked by default at many cloud providers. Check before building a mail server you can’t use.
Preparing the Server
Update the system
Log in as root, or as a sudo user, and bring everything current:
sudo apt update
sudo apt full-upgrade -y
sudo reboot
Reboot even if it feels unnecessary. A new kernel pending behind a running old one produces confusing results later, particularly with firewall modules.
Set a proper hostname
ISPConfig wants a fully qualified domain name (FQDN) as the server hostname, for example server1.example.com. It uses this for the panel certificate, mail banner and a few internal checks.
sudo hostnamectl set-hostname server1.example.com
Then edit /etc/hosts so the name resolves locally:
127.0.0.1 localhost
203.0.113.10 server1.example.com server1
Verify:
hostname -f
hostname -i
hostname -f should return the full name. If it returns just server1 or an error, fix that now. Also create a public DNS A record (and AAAA if applicable) pointing server1.example.com to the server IP. Let’s Encrypt validation for the panel depends on it.
Set timezone and time sync
For a server in Indonesia, for example:
sudo timedatectl set-timezone Asia/Jakarta
timedatectl status
Correct time matters for TLS, DKIM, log correlation and cron. Ubuntu runs systemd-timesyncd by default and that’s adequate for most servers.
Remove conflicting software
On a truly fresh image there’s nothing to remove. On a provider image that came with Apache, Nginx, Postfix or a control panel pre-baked, start over with a clean install instead of untangling it. Check what already listens on ports:
sudo ss -tulpn
Expect to see only SSH and systemd-resolved on a stub address. Anything else on 80, 443, 25 or 3306 is a red flag.
Mind the local DNS stub
Ubuntu’s systemd-resolved listens on 127.0.0.53:53. That’s usually fine, but if you plan to run a local DNS authority (PowerDNS or BIND) on the public address, you may see port conflicts. The Autoinstaller generally manages this when you opt into DNS, but if you see “address already in use” on port 53, this is where to look.
Running the ISPConfig Autoinstaller
The Autoinstaller is a script that prepares the whole stack and then installs ISPConfig. The download page confirms it’s the recommended method and that it prepares a fresh Debian or Ubuntu server with all services.
Run it from a root shell inside tmux or screen, so a dropped connection won’t kill it midway:
sudo apt install -y tmux wget
tmux new -s ispc
sudo -i
wget -O autoinstall.sh https://get.ispconfig.org
sh autoinstall.sh
The script is fetched from the project’s own domain. Read it first if you’re cautious (less autoinstall.sh). That habit costs thirty seconds and is worth it on any script you run as root.
Choosing options
Running the script without arguments gives interactive prompts. For a repeatable build, flags are better. A common pattern:
sh autoinstall.sh --help
Read the list for your version, because options evolve between releases. Typical choices include selecting --use-nginx or Apache (the default), skipping --no-mail, --no-dns, or --no-roundcube where you don’t need those services, and --use-ftp-ports to set a passive range. Confirm exact flag names in --help on your release instead of trusting a blog post, this one included.
The practical rule: install only what the server will actually use. Every extra daemon is memory, patching and attack surface.
What the installer does
Roughly, it will:
- Check the OS and refuse to continue on unsupported systems.
- Install packages: web server, PHP versions, MariaDB, Postfix, Dovecot, Rspamd, Pure-FTPd and so on.
- Download and run the ISPConfig installer with generated answers.
- Create a panel admin account and print credentials at the end.
Save that final output securely. The admin password is shown once.
Verify after install
The panel normally answers on port 8080:
https://server1.example.com:8080
Your browser will warn about a self-signed certificate until you issue a real one. Then check services:
sudo systemctl --failed
sudo systemctl status apache2 postfix dovecot mariadb --no-pager
sudo ss -tulpn | grep -E ':(80|443|8080|25|587|993|21)\s'
systemctl --failed should come back empty. If Dovecot is failed, see the troubleshooting section, because Ubuntu 26.04’s Dovecot 2.4 syntax is exactly what the 3.3.2 release was patched for.
Make sure you’re on 3.3.2 or newer. Check the version in the panel footer or run:
grep ISPC_APP_VERSION /usr/local/ispconfig/server/lib/config.inc.php
Firewall Configuration
ISPConfig can install its own firewall helper, but many admins prefer to control this directly. With UFW:
sudo ufw default deny incoming
sudo ufw default allow outgoing
sudo ufw allow 22/tcp
sudo ufw allow 80,443/tcp
sudo ufw allow 8080/tcp
sudo ufw allow 25,465,587/tcp
sudo ufw allow 110,143,993,995/tcp
sudo ufw allow 21/tcp
sudo ufw allow 40110:40210/tcp
sudo ufw allow 53
sudo ufw enable
Trim this to your actual services. If you don’t run POP3, don’t open 110 and 995. If DNS is external, skip 53. Always allow SSH before enabling, or you’ll lock yourself out.
Better still, restrict port 8080 to your own IP range:
sudo ufw delete allow 8080/tcp
sudo ufw allow from 198.51.100.25 to any port 8080 proto tcp
The admin panel is the highest-value target on the box. There’s no reason for the whole internet to see its login page.
Note that Ubuntu 26.04 may use nftables as the backend behind ufw and iptables compatibility layers. If you run Fail2Ban, make sure its ban action matches (nftables or ufw) or bans will log but never block.
Securing the Panel and the Server
SSL for the panel and hostname
Replace the self-signed certificate. In the ISPConfig panel, create the website for the server hostname or use the CLI. Many admins use acme.sh or Certbot. Let’s Encrypt support is built into ISPConfig for customer sites. For the panel itself, ISPConfig can reuse the certificate if you set it up via System > Server Config > Web options, or by running the installer’s SSL step again. Check the official docs for the exact option on your version.
Harden SSH
sudo nano /etc/ssh/sshd_config.d/99-hardening.conf
PermitRootLogin no
PasswordAuthentication no
PubkeyAuthentication yes
MaxAuthTries 3
Create a sudo user and test key login in a second session before applying:
sudo sshd -t && sudo systemctl reload ssh
Keep your existing session open until you confirm the new one works.
Fail2Ban and brute force
ISPConfig’s Autoinstaller usually sets up Fail2Ban jails for SSH, mail and web logins. Confirm:
sudo fail2ban-client status
Add a longer ban time for repeat offenders:
[DEFAULT]
bantime = 1h
findtime = 10m
maxretry = 5
bantime.increment = true
Put overrides in /etc/fail2ban/jail.local, not the packaged files.
Security awareness: PHP patches
Ubuntu published fixes for a critical PHP vulnerability dubbed php2shell, and for 26.04 the fixed php8.5 package is version 8.5.4-0ubuntu1.5. That’s a good reminder: a hosting server running customer PHP code is exactly the target for this kind of bug. Run apt list --upgradable regularly, and enable unattended security updates:
sudo apt install -y unattended-upgrades
sudo dpkg-reconfigure -plow unattended-upgrades
Reboot-needed checks matter too. Install needrestart and let it flag services using old libraries.
Isolation per site
ISPConfig creates a separate system user and group for each website, and runs PHP-FPM pools per site. Keep that. Don’t disable PHP-FPM isolation to “make things work” with some plugin. If one customer’s WordPress gets compromised, per-site users and open_basedir limits are what keep the damage contained.
Performance Tuning
Defaults are safe, not fast. A few adjustments make a real difference.
PHP-FPM
Each site gets its own pool. The key settings in ISPConfig’s Website > Options > PHP-FPM Settings are:
pm = ondemandfor low-traffic sites. Idle sites then consume almost no memory.pm = dynamicfor busy sites, withpm.max_childrensized by RAM.
A rule of thumb: measure average PHP worker memory with
ps --no-headers -o rss -C php-fpm8.5 | awk '{sum+=$1} END {print sum/NR/1024 " MB avg"}'
Then max_children is roughly the RAM you can spare for PHP divided by that number. Guessing here is how servers hit the OOM killer during a traffic spike.
Enable OPcache with sane values:
opcache.enable=1
opcache.memory_consumption=192
opcache.max_accelerated_files=20000
opcache.validate_timestamps=1
opcache.revalidate_freq=10
PHP 8.5 is the default on this Ubuntu release, and ISPConfig lets you run alternate versions per site, which helps for legacy applications that choke on new PHP.
MariaDB
Check buffer usage rather than copying someone’s my.cnf:
[mysqld]
innodb_buffer_pool_size = 1G
innodb_log_file_size = 256M
max_connections = 150
slow_query_log = 1
long_query_time = 2
On a 4 GB server shared with web and mail, 1 GB for the buffer pool is plenty. Turn on the slow query log early, because it tells you which customer’s plugin is hammering the database.
Web server
For Apache, event MPM with PHP-FPM is the right pairing. Confirm:
apachectl -V | grep -i mpm
Enable HTTP/2 and compression, and set sensible keepalive:
KeepAlive On
KeepAliveTimeout 3
For Nginx, set worker_processes auto; and raise worker_connections to 4096 if you expect heavy concurrency.
Disk and memory
- Mount data volumes with
noatime. - Use
iostat -xz 2andiotop -oto find noisy neighbours on disk. - Add a modest swap file (1 to 2 GB) as a buffer, not as a performance plan.
- With cgroup v2 enforced on Ubuntu 26.04, per-service limits via systemd work well. Cap ClamAV, for instance:
sudo systemctl edit clamav-daemon
[Service]
MemoryMax=1200M
Network
Check the kernel’s defaults before blindly applying “tuning” sysctls. A sane addition for busy servers:
net.core.somaxconn = 4096
net.ipv4.tcp_max_syn_backlog = 4096
net.ipv4.tcp_syncookies = 1
Save in /etc/sysctl.d/99-hosting.conf and apply with sudo sysctl --system.
Mail, DNS and the Parts People Forget
Mail deliverability
Installing Postfix and Dovecot is the easy half. Getting mail into inboxes takes more:
- A reverse DNS (PTR) record matching the server hostname. Set it with your hosting provider.
- SPF, DKIM and DMARC records for every sending domain. ISPConfig generates DKIM keys when enabled per domain.
- Check your IP on common blocklists before launching.
Version 3.3.2 also added an option to enforce inbound DMARC rules and a separate sender address for system mail. Both are useful, and worth reading about before you turn them on, because strict DMARC enforcement can reject legitimate forwarded mail.
DNS
If you host DNS, run at least two nameservers on different networks. A single nameserver is a single point of failure. If that sounds like too much, use external DNS and keep ISPConfig for web and mail only.
Backups
ISPConfig has built-in backup per site and per database, but storing them on the same disk is not a backup. Push copies off the server:
sudo apt install -y restic
restic -r sftp:backup@backuphost:/srv/restic init
restic -r sftp:backup@backuphost:/srv/restic backup /var/vmail /var/www /etc /root
Dump databases first or use ISPConfig’s DB backup. Test a restore at least once. An untested backup is a hope.
Updating ISPConfig
Run updates as root. The standard sequence looks like this:
cd /tmp
wget https://www.ispconfig.org/downloads/ISPConfig-3.3.2.tar.gz
tar xvfz ISPConfig-3.3.2.tar.gz
cd ispconfig3_install/install
php -q update.php
Newer versions follow the same pattern, and a later 3.3.3 release added a security check page for administrators. Check the downloads page for the current version number before running the commands above, and use the matching filename. Take a snapshot or backup first. The updater asks whether to reconfigure services, and the safe answer is usually to accept the defaults unless you’ve deliberately customized something.
Troubleshooting Common Problems
Installer exits with “unsupported operating system”
You’re on an older ISPConfig script or release. Support for 26.04 arrived in 3.3.2. Re-download the Autoinstaller from get.ispconfig.org and confirm lsb_release -a reports Ubuntu 26.04.
Dovecot fails to start
Check the log:
sudo journalctl -u dovecot -b --no-pager | tail -50
sudo doveconf -n
Errors about unknown settings usually mean a config written for Dovecot 2.3 is being parsed by 2.4. Make sure ISPConfig is 3.3.2 or newer, and re-run the updater with the reconfigure option so it rewrites Dovecot’s config.
Panel on port 8080 won’t load
Work through it in order:
sudo ss -tlnp | grep 8080
sudo ufw status
sudo journalctl -u apache2 -n 50 --no-pager
Listening but unreachable means firewall (local or provider-level security group). Not listening means the panel’s virtual host failed to load, and the Apache or Nginx log will say why.
Websites show errors after PHP changes
An Ask Ubuntu thread from October 2026 describes Apache and PHP breaking after an upgrade to 26.04, with sites failing to load properly. That matches a common pattern: an in-place upgrade leaves old PHP modules enabled alongside the new version. On ISPConfig servers, a fresh install is cleaner than upgrading from 24.04. If you must upgrade in place, check:
sudo apache2ctl -M | grep php
sudo a2query -m
sudo systemctl status php8.5-fpm
Disable stale modules and confirm each site’s PHP mode in the panel.
Let’s Encrypt fails
Check three things: the domain’s DNS points at this server, port 80 is open, and the site’s .well-known/acme-challenge path isn’t blocked by a rewrite rule or a redirect to a login page. A “Too many failed authorizations” message means wait an hour and fix the cause first.
Mail goes to spam or is rejected
Run:
postqueue -p
sudo tail -n 100 /var/log/mail.log
Look for “blocked using” (blocklist), “Relay access denied” (account or auth issue) or “Connection timed out” to port 25 (provider blocks outbound SMTP). The last one is more common than you’d think.
Server runs out of memory
free -h
sudo dmesg -T | grep -i 'killed process'
Usual culprits are ClamAV, too many PHP-FPM children, or MariaDB’s buffer pool set too high. Fix them in that order.