Installing AdminBolt ​
AdminBolt is installed with a single command that runs a non-interactive installer on a fresh server. This page describes the requirements, the install command, its options, what the installer does, and how to reach the panel for the first time.
Requirements ​
- Operating system: AlmaLinux 9 or Rocky Linux 9.
- Access: root (the installer must be run as root, for example with
sudo). - Port: TCP 8443 must be free. If it is already used by an existing bolt-nginx, the installer proceeds as a reinstall; if it is used by any other service, the installer stops.
- Commands:
curlandssmust be available. Both are present by default on AlmaLinux 9 and Rocky Linux 9. - RAM: 2 GB minimum, 4 GB recommended.
- CPU: 2 cores.
- Disk: 20 GB of available space.
Quick install ​
Run the following as root on a supported, freshly provisioned server:
curl -fsSL https://get.adminbolt.com/install.sh | bashThis command is non-interactive: it does not prompt you for a domain, email, nameservers, or service selection. It downloads the installer, installs the current AdminBolt release, and tees all output to /home/adminbolt-install-log.txt so you can review the full run afterward.
Since panel 1.9.5, the downloaded copy of the installer is removed once the install has succeeded. When the install fails, the copy stays in root's home directory as adminbolt-installer.sh, and the output and the install log end with a line that starts with [FAILED] and gives the command that runs it again with the same arguments. See When the install does not finish.
Options ​
The following options apply when you invoke the installer script directly rather than through the quick-install command:
--email=<address>- activate a trial license for that address during the install. See Licensing at install time.--version=<PANEL_VERSION>- install a specific bolt-panel version. When omitted, the latest version from the repository is installed.--mariadb-version=<10.11|11.4>- the MariaDB series to install. When omitted, a new server gets MariaDB 11.4. See MariaDB version.--ssh-key=<public key|https URL>- add an SSH public key for root and turn SSH password login for root off once a sign-in with that key has been seen. See SSH key for root.--without=<mail,dns,ftp>- since panel 1.9.5, leave mail, DNS, or FTP out of the server, for a server with little memory. See Minimal install.--admin-username=<name|random>- since panel 1.9.5, the username of the first administrator. When omitted, it isadmin. See First administrator name.--no-telemetry- send no report about the install to AdminBolt, and start the server with AdminBolt usage statistics off.--help,-h- print usage information.
Licensing at install time ​
AdminBolt needs a license to run. You decide at install time whether the installer asks for a trial on your behalf, or whether you activate a license yourself afterwards.
Pass your e-mail address to start a trial as part of the install:
sudo ./install.sh [email protected]The trial is activated immediately and runs for 24 hours. AdminBolt also sends a confirmation message to that address; confirming your e-mail address through it extends the trial to the full period. Until the address is confirmed, the trial stays at 24 hours. The panel's License page (/admin/settings/license) shows the state of the trial and can resend the message.
Without --email, the install still completes, and the panel starts unlicensed. Activate a license whenever you are ready on the License page, either by requesting a trial there or by entering a license key you already hold.
Note: The installer checks the address before anything is installed and stops if it is not a well-formed e-mail address of the form
[email protected]. Since panel 1.9.4 the error names what it does not accept: spaces, quotes ("),$, the backtick,;, and control characters. An address it accepts is passed to the trial request exactly as you typed it. If a license is already configured on the server, the trial request is skipped.
MariaDB version ​
Since panel 1.9.4 a new server gets MariaDB 11.4. To install the previous long-term series instead, pass --mariadb-version:
sudo ./install.sh --mariadb-version=10.11| Series | Supported until |
|---|---|
| MariaDB 11.4 (the default) | 2029-05 |
| MariaDB 10.11 | 2028-02 |
Any other value stops the installer before anything is installed. Without the option, a server that already runs MariaDB, for example when you run the installer again, keeps the series it has. Servers installed with an earlier version keep their series when they are updated; the panel moves a server to a newer series only when you choose it under MariaDB server on MySQL Settings.
SSH key for root ​
Since panel 1.9.4, --ssh-key adds your SSH public key for root and, once it is proven to work, turns SSH password login for root off, so the server is not open to password guessing against root from its first minutes.
sudo ./install.sh --ssh-key="ssh-ed25519 AAAA... you@laptop"The value is one public key in quotes, or an https:// address that returns public keys, one per line. The installer checks the key before anything is installed and stops when it is not a plain SSH public key. Keys root already has stay as they are.
Password login for root is turned off at the very end of the install, and only when every check passes:
- The key is still among root's authorized keys, and the server's SSH settings let root sign in with it.
- The key is not restricted, for example to one command, so that a sign-in with it opens a shell.
- A root sign-in with that key is seen after the installer has set up SSH. When the installer runs in a terminal, it asks you to sign in with the key from a new terminal and waits up to 10 minutes; press Enter to skip. Without a terminal, for example when the install runs from cloud-init, no such sign-in can be seen.
When a check does not pass, password login for root stays as it was, and the installer prints the reason and the command that turns it off once you have signed in with the key. Only root is affected: other users, such as hosting accounts, keep their password login, and whether root may sign in over SSH at all stays under SSH Access. When password login for root is turned off, the installer prints how to allow it again.
While root can still sign in with a password, Health shows a warning.
Minimal install ​
Since panel 1.9.5, --without leaves whole components out of a new server, for example on a server with little memory. Name one or more of mail, dns, and ftp, separated by commas:
sudo ./install.sh --without=mail,dns,ftp| Component | Left out | What the server does without it |
|---|---|---|
mail | Postfix, Dovecot, Rspamd, mailing lists, Roundcube webmail, and the system e-mail account | No mailboxes, webmail, or mailing lists. The panel's own e-mail, such as alerts and password resets, is not sent either, because it goes through the server's mail service. |
dns | PowerDNS | The server answers no DNS queries, so the domains need their DNS hosted elsewhere. The panel keeps each domain's records. |
ftp | vsftpd | No FTP accounts. Files can still be managed in the file manager. |
Any other value stops the installer before anything is installed. The end of the install lists what was left out and the commands that add each component later.
The panel treats a left-out component as left out by choice, not as a fault:
- Health raises no warning about it and lists its services in grey as Not installed, "Not installed by choice: left out when this server was installed." The System status card of the dashboard shows the Mail and DNS rows as Not installed by choice.
- The client panel hides the pages of a left-out component: email accounts, forwarders, autoresponders, email routing, mail delivery, catch-all, and mailing lists for mail, the zone editor for DNS, and FTP accounts for FTP. Feature Lists do not offer them. Accounts whose mail is handled by a mail server node keep their mail pages.
- The admin menu leaves out the Email area and Email Settings while mail is left out, and FTP under Settings > Tools while FTP is left out. The DNS pages stay, because the panel keeps each domain's records. A bookmark to a hidden page opens a notice instead, for example "Email is not installed on this server.", with a link to Health.
- New domains get no mail records and no
mail,webmail,autoconfig, orautodiscovernames while mail is left out, and noftpname while FTP is left out. - Over the API, a client request for a left-out component is refused with "This feature is not installed on this server", and so is creating a mailbox or an FTP account over the admin API.
To add a component later, since panel 1.9.6, click Install on any of its rows under Manage services on Health. The panel then sets up the whole component in the background, profiles included, and for mail also mailing lists, the system e-mail account, and Roundcube webmail, so the result is the same as on a server installed with it. An install that stops part way names the step on every row of the component, and Run install again finishes it. The commands the end of the install printed do the same from a root shell, in that order. The pages, menus, and API come back on their own within a minute once the panel sees the service installed, and Health at its next check. Nothing else has to be changed. When DNS is added, the zones of the existing domains are written. Domains created while mail was left out get their mail records when you apply the DNS template to them again; see DNS Templates.
Remove on the Postfix, PowerDNS, or vsftpd row of Health takes a component out again later. Since panel 1.9.6 it lists what stops working before it goes ahead, and the server is then treated as installed without that component. See Removing mail, DNS, or FTP.
First administrator name ​
Since panel 1.9.5, --admin-username names the first administrator of the new server. Without it, the first administrator is admin.
sudo ./install.sh --admin-username=jane- A name is 3 to 32 characters: lowercase letters, numbers, dots, dashes, or underscores, starting with a letter. It cannot be a name reserved for the system, a hosting account, or a user that already exists on the server.
--admin-username=randompicksadmin-followed by 6 random characters, so the name is not one that password guessing tries first.
The installer checks the name before anything is installed. The end of the install prints the name to sign in with. You can still change the name in the setup wizard.
On a server installed without the installer's option, bolt-cli admin-sso-generate --first-admin does the same when it creates the first administrator.
What the installer does ​
The installer runs in three stages.
Stage 1 - pre-flight checks ​
Confirms that the script is running as root, that the operating system is AlmaLinux 9 or Rocky Linux 9, that the required commands are present, and that TCP port 8443 is free (or already held by an existing bolt-nginx).
Stage 2 - base setup and packages ​
- Disables SELinux.
- Enables the EPEL and CRB repositories, installs prerequisite packages, and installs acme.sh.
- Adds the Bolt package repository and installs
bolt-agent,bolt-nginx,bolt-php, andbolt-panel, then enables and startsbolt-nginxandbolt-php.
Stage 3 - post-install via bolt-cli ​
Requests a trial license when you passed --email, connects bolt-agent to the panel, and installs and configures the service stack: the nftables firewall, SSH, PowerDNS, MariaDB (11.4 unless --mariadb-version names another series), Postfix, Dovecot, Redis, Rspamd, mlmmj, PHP 8.4, Apache, vsftpd, and Fail2Ban. It also installs the integrated apps (Local API, File Manager, phpMyAdmin, Roundcube, Git, Metrics, and Adminer), their default profiles, the firewall, the default DNS-records template, and cron jobs. Finally, it generates the first-login SSO URL.
The default hostname, nameservers, and Contact Email ​
During the install, AdminBolt gives the server a free hostname under myboltip.com, built from its public IPv4 address with dashes in place of the dots. A server at 203.0.113.10, for example, gets 203-0-113-10.myboltip.com. The panel, mail, and FTP answer on that name before you have set up a domain of your own. A publicly trusted certificate for it is requested automatically during the install; until it is issued, or if the request fails, they use a temporary self-signed certificate, and your browser warns about it.
The installer sets three things from that name:
| Setting | Starts as | Where you change it |
|---|---|---|
| Hostname | 203-0-113-10.myboltip.com | Settings > General > General Settings, Hostname & SSL tab |
| Nameservers | ns1.203-0-113-10.myboltip.com and ns2.203-0-113-10.myboltip.com | Settings > DNS > DNS Settings, see Change Nameservers |
| Contact Email | [email protected] | Contact Email on the General tab of General Settings |
A server that already has a name of its own ​
Since panel 1.9.6, a first install uses the server's own fully qualified name as the panel hostname instead of the myboltip.com name, when both of these hold:
- the name is a public name, not a local one and not the default name a hosting provider gave the server,
- its DNS A record points at a public address of this server, and so does its AAAA record when it has one.
For a server named server1.example.com whose A record points at it, the install then starts like this:
| Setting | Starts as |
|---|---|
| Hostname | server1.example.com |
| Nameservers | ns1.203-0-113-10.myboltip.com and ns2.203-0-113-10.myboltip.com |
| Contact Email | [email protected] |
The nameservers stay on the free myboltip.com name until you change them. In every other case the install uses the free myboltip.com name, as before. To have your own name picked up, set the server's hostname and its A record before you run the installer.
Only a first install chooses the hostname this way. A server that already has a panel hostname keeps it through updates and a rerun of the installer; change it on the Hostname & SSL tab.
When you move the server to a hostname of your own, change all three. A hostname change does not change the nameservers or Contact Email, and it switches off the admin@ mailbox the default Contact Email points to. See What keeps the myboltip.com name.
How myboltip.com names work ​
- A name of the form
<ip-with-dashes>.myboltip.comresolves to the IPv4 address it is built from. - The certificate for the default hostname is issued with a DNS check. While it is issued, a temporary TXT record,
_acme-challenge.<ip-with-dashes>.myboltip.com, is published through AdminBolt's service so that the certificate authority can check it, and it is removed again shortly afterwards.
First login ​
When the installer finishes, it prints a one-time SSO URL. Open that URL to reach the admin panel at https://<server>:8443 without entering a password.
Since panel 1.9.5, the end of the install also says how to sign in later and how to get back in:
- Signing in later - the admin username,
adminunless you chose another with--admin-username, or the e-mail address you enter in the setup wizard, with the password you set there. - Root sign-in -
rootcan also sign in to the admin panel with the server's root password. The setup wizard recommends turning that off and lets you keep it on; see Root sign-in. - If you get locked out - the commands that print a new one-time sign-in link, reset the admin password, and lift a Fail2Ban ban on your address. The admin sign-in page lists the same under Cannot sign in?; see Cannot sign in?.
Since panel 1.9.5, the server's address opened in a browser over plain HTTP, for example http://203.0.113.10, shows a Control panel page with Open the control panel, and a domain that points to the server but has no website yet shows No website here yet. See Server address page.
Since panel 1.9.4, the panel opened by the server's IP address, for example https://203.0.113.10:8443, shows a trusted certificate issued for that address. The panel orders it on its own within about an hour of the install. Until then, or when it cannot be issued, the address shows the hostname certificate and your browser warns about it. See Panel certificate for the server IP address.
To generate a fresh SSO URL later, run bolt-cli admin-sso-generate.
The panel asks two questions the first time you sign in and then hands you your logins. See First run for what to expect and what to keep.
Unattended installation ​
First access normally uses the one-time SSO URL described above. For fully unattended provisioning, where no operator reads the installer output, you can set the administrator password directly instead of relying on the SSO URL.
Run bolt-cli set-admin-password to set the password for the administrator the installer created, admin unless you named it with --admin-username.
bolt-cli set-admin-passwordCalled without options, the command generates a strong 20-character password and prints it to stdout. The password is temporary: the panel requires you to change it at first sign-in.
Since panel 1.9.5, --password-stdin reads the password you choose from the first line of standard input, so it never appears on the command line. For example, with the password in a file only root can read:
bolt-cli set-admin-password --password-stdin < /root/admin-password.txtSince panel 1.9.5, --username=<name> sets the password of that administrator instead. See bolt-cli set-admin-password for the details.
Troubleshooting ​
If the installer stops before completing, check for one of these conditions and review the install log at /home/adminbolt-install-log.txt:
- Not run as root - the installer exits immediately. Re-run it as root, for example with
sudo. - Unsupported operating system - the installer supports AlmaLinux 9 and Rocky Linux 9 only. On any other distribution or version, it stops during the pre-flight checks.
- Missing
curl- the installer requirescurland exits if it is not found. Installcurland run the installer again. - Port 8443 already in use - if another service (anything other than an existing bolt-nginx) is listening on TCP 8443, the installer stops and prints the current listener. Free the port and run the installer again.
- Repository download failure - if the Bolt package repository cannot be downloaded, the installer stops. Confirm that the server has outbound internet access and that no firewall is blocking the download, then run the installer again.
When a package repository cannot be reached ​
Since panel 1.9.6, an install finishes when the package repository of PHP or MariaDB cannot be reached while it runs. PHP that waits for its repository installs on its own afterwards: the dashboard says which PHP versions are waiting and when the next try is, and Try now tries at once. Until a PHP version is installed, a website cannot be created on it, and the message says that the version installs on its own once the repository answers.
The installer reports a failed install of Postfix, Dovecot, MariaDB, or PHP as a failure, so the end of the install shows what did not install.
When the install does not finish ​
Since panel 1.9.5, when an install cannot finish, the installer says what is missing, for example that the disk filled up while bolt-agent was being set up. Fix the cause it names, then run the installer again with the same arguments. The rerun sets up again what did not finish, including bolt-agent and its connection to the panel, and completes the install.
After a failed install started with the quick-install command, the last line of the output gives the command that runs the kept copy of the installer again with the same arguments and appends to the same install log. The copy is removed once that run succeeds.