Email server setup script

This script installs an email server with all the features required in the modern web.

I’ve linked this file on Github to a shorter, more memorable address on my website so you can get it on your machine with this short command:

curl -LO lukesmith.xyz/emailwiz.sh

When prompted by a dialog menu at the beginning, select “Internet Site”, then give your full domain without any subdomain, e.g. lukesmith.xyz.

I’m glad to say that dozens, hundreds of people have now used it and there is a sizeable network of people with email servers thanks to this script.

This script installs

• Postfix to send and receive mail.
• Dovecot to get mail to your email client (mutt, Thunderbird, etc.).
• Config files that link the two above securely with native PAM log-ins.
• Spamassassin to prevent spam and allow you to make custom filters.
• OpenDKIM to validate you so you can send to Gmail and other big sites.
• Certbot SSL certificates, if not already present.
• fail2ban to increase server security, with enabled modules for the above programs.
• (optionally) a self-signed certificate instead of OpenDKIM and Certbot. This allows to quickly set up an isolated mail server that collects email notifications from devices in the same local network(s) or serves as secure/private messaging system over VPN.

This script does not...

• use a SQL database or anything like that. We keep it simple and use normal Unix system users for accounts and passwords.
• set up a graphical web interface for mail like Roundcube or Squirrel Mail. You are expected to use a normal mail client like Thunderbird or K-9 for Android or good old mutt with mutt-wizard. Note that there is a guide for Rainloop on LandChad.net for those that want such a web interface.

Prerequisites for Installation

1. Debian or Ubuntu server.
2. DNS records that point at least your domain’s mail. subdomain to your server’s IP (IPv4 and IPv6). This is required on initial run for certbot to get an SSL certificate for your mail. subdomain.

Mandatory Finishing Touches

Unblock your ports

While the script enables your mail ports on your server, it is common practice for all VPS providers to block mail ports on their end by default. Open a help ticket with your VPS provider asking them to open your mail ports and they will do it in short order.

DNS records

At the end of the script, you will be given some DNS records to add to your DNS server/registrar’s website. These are mostly for authenticating your emails as non-spam. The 4 records are:

1. An MX record directing to mail.yourdomain.tld.
2. A TXT record for SPF (to reduce mail spoofing).
3. A TXT record for DMARC policies.
4. A TXT record with your public DKIM key. This record is long and uniquely generated while running emailwiz.sh and thus must be added after installation.

They will look something like this:

@	MX	10	mail.example.org
mail._domainkey.example.org    TXT     v=DKIM1; k=rsa; p=anextremelylongsequenceoflettersandnumbersgeneratedbyopendkim
_dmarc.example.org     TXT     v=DMARC1; p=reject; rua=mailto:dmarc@example.org; fo=1
example.org    TXT     v=spf1 mx a: -all

The script will create a file, ~/dns_emailwiz that will list our the records for your convenience, and also prints them at the end of the script.

Add a rDNS/PTR record as well!

Set a reverse DNS or PTR record to avoid getting spammed. You can do this at your VPS provider, and should set it to mail.yourdomain.tld. Note that you should set this for both IPv4 and IPv6.

Making new users/mail accounts

Let’s say we want to add a user Billy and let him receive mail, run this:

useradd -m -G mail billy
passwd billy

Any user added to the mail group will be able to receive mail. Suppose a user Cassie already exists and we want to let her receive mail too. Just run:

usermod -a -G mail cassie

A user’s mail will appear in ~/Mail/. If you want to see your mail while ssh’d in the server, you could just install mutt, add set spoolfile="+Inbox" to your ~/.muttrc and use mutt to view and reply to mail. You’ll probably want to log in remotely though:

Installing with self-signed certificate, in “isolated” mode

This mode skips the setup of OpenDKIM and Certbot, and will instead create a self-signed cert that lasts 100 years. It also allows to customize the logic country name, state/province name and organization name to generate the certificate automatically. An example usecase is for an isolated server that collects notifications from devices in the same local network(s) or serves as secure/private messaging system over VPN (wireguard or whatever). This server with self-signed certificate as configured will NOT be able to send anything to public mail servers (Gmail, Outlook and so on), at least not directly.

open the script and change the following line

selfsigned="no" # yes no

to become

selfsigned="yes" # yes no

it’s also possible to customize and automate the self-signed certificate creation by changing the following lines in the script

use_cert_config="no"

to

use_cert_config="yes"

and then write country name, state/province name and organization name in the following lines

country_name="" # IT US UK IN etc etc
state_or_province_name=""
organization_name=""

Logging in from email clients (Thunderbird/mutt/etc)

Let’s say you want to access your mail with Thunderbird or mutt or another email program. For my domain, the server information will be as follows:

• SMTP server: mail.lukesmith.xyz
• SMTP port: 465
• IMAP server: mail.lukesmith.xyz
• IMAP port: 993

MTA-STS and DANE for improved security

MTA-STS

By its very nature SMTP does not offer built-in security against man-in-the-middle attacks. To mitigate this risk, you can implement the MTA-STS policy, which instructs compatible senders to employ verified TLS encryption when communicating with your server.

To put this into practice, create a file named mta-sts.txt with the specified content and host it at https://mta-sts.example.org/.well-known/:

version: STSv1
mode: enforce
max_age: 604800
mx: mail.example.org

After that you need to add the following DNS records:

_mta-sts.example.org.   TXT    "v=STSv1; id=<id>"
_smtp._tls.example.org. TXT    "v=TLSRPTv1;rua=mailto:postmaster@example.org"

<id> can be an arbitrary number but it’s recommended to use the current unix timestamp (date +%s)

DANE

It’s also recommended to set up a TLSA (DNSSEC/DANE) record for further security enhancement. Go here to generate a TLSA record. Set the port to 25, Transport Protocol to “tcp”, and specify the MX hostname as the Domain Name.

After adding the TLSA DNS record you need to enable opportunistic DANE in postfix by doing the following:

postconf -e 'smtpd_use_tls = yes'
postconf -e 'smtp_dns_support_level = dnssec'
postconf -e 'smtp_tls_security_level = dane'

echo "dane       unix  -       -       n       -       -       smtp
  -o smtp_dns_support_level=dnssec
  -o smtp_tls_security_level=dane" >> /etc/postfix/master.cf

Benefited from this?

I am always glad to hear this script is still making life easy for people. If this script or documentation has saved you some frustration, donate here:

• btc: bc1qzw6mk80t3vrp2cugmgfjqgtgzhldrqac5axfh4
• xmr: 8A5v4Ci11Lz7BDoE2z2oPqMoNHzr5Zj8B3Q2N2qzqrUKhAKgNQYGSSaZDnBUWg6iXCiZyvC9mVCyGj5kGMJTi1zGKGM4Trm

Sites for Troubleshooting

Can’t send or receive mail? Getting marked as spam? There are tools to double-check your DNS records and more:

• Always check journalctl -xe first for specific errors.
• Check your DNS
• Test your TXT records via mail
• Is your IP blacklisted?
• mxtoolbox
• Check overall mail/website
• Another great mail checker
• Check DANE