Prepare your system
Before you run mailcow: dockerized, there are a few requirements that you should check:
Do not try to install mailcow on a Synology/QNAP device (any NAS), OpenVZ, LXC or other container platforms. KVM, ESX, Hyper-V and other full virtualization platforms are supported.
- mailcow: dockerized requires some ports to be open for incoming connections, so make sure that your firewall is not blocking these.
- Make sure that no other application is interfering with mailcow's configuration, such as another mail service
- A correct DNS setup is crucial to every good mailserver setup, so please make sure you got at least the basics covered before you begin!
- Make sure that your system has a correct date and time setup. This is crucial for various components like two factor TOTP authentication.
Minimum System Resources¶
OpenVZ, Virtuozzo and LXC
Please make sure that your system has at least the following resources:
|RAM||Minimum 6 GiB + 1 GiB swap (default config)|
|Disk||20 GiB (without emails)|
ClamAV and Solr can be greedy with RAM. You may disable them in
mailcow.conf by settings
We are aware that a pure MTA can run on 128 MiB RAM. mailcow is a full-grown and ready-to-use groupware with many extras making life easier. mailcow comes with a webserver, webmailer, ActiveSync (MS), antivirus, antispam, indexing (Solr), document scanner (Oletools), SQL (MariaDB), Cache (Redis), MDA, MTA, various web services etc.
A single SOGo worker can acquire ~350 MiB RAM before it gets purged. The more ActiveSync connections you plan to use, the more RAM you will need. A default configuration spawns 20 workers.
RAM usage examples¶
A company with 15 phones (EAS enabled) and about 50 concurrent IMAP connections should plan 16 GiB RAM.
6 GiB RAM + 1 GiB swap are fine for most private installations while 8 GiB RAM are recommended for ~5 to 10 users.
We can help to correctly plan your setup as part of our support.
Basically, mailcow can be used on any distribution that is supported by Docker CE (see https://docs.docker.com/install/). However, in some cases there may be incompatibilities between the operating systems and the mailcow components.
The following table contains all operating systems officially supported and tested by us (as of June 2023):
|Alpine 3.17+ and older||⚠️|
|Debian 10, 11, 12||✅|
|Ubuntu 18.04, 20.04, 22.04||✅|
|Alma Linux 8||✅|
|Rocky Linux 9||✅|
✅ = Works out of the box using the instructions.
⚠️ = Requires some manual adjustments otherwise usable.
❌ = In general NOT Compatible.
❔ = Pending.
Note: All other operating systems (not mentioned) may also work, but have not been officially tested.
Firewall & Ports¶
Please check if any of mailcow's standard ports are open and not in use by other applications:
ss -tlpn | grep -E -w '25|80|110|143|443|465|587|993|995|4190' # or: netstat -tulpn | grep -E -w '25|80|110|143|443|465|587|993|995|4190'
There are several problems with running mailcow on a firewalld/ufw enabled system.
You should disable it (if possible) and move your ruleset to the DOCKER-USER chain, which is not cleared by a Docker service restart, instead.
See this (blog.donnex.net) or this (unrouted.io) guide for information about how to use iptables-persistent with the DOCKER-USER chain.
As mailcow runs dockerized, INPUT rules have no effect on restricting access to mailcow.
Use the FORWARD chain instead.
If this command returns any results please remove or stop the application running on that port. You may also adjust mailcows ports via the
mailcow.conf configuration file.
If you have a firewall in front of mailcow, please make sure that these ports are open for incoming connections:
To bind a service to an IP address, you can prepend the IP like this:
Important: You cannot use IP:PORT bindings in HTTP_PORT and HTTPS_PORT. Please use
Important for Hetzner firewalls¶
Quoting https://github.com/chermsen via https://github.com/mailcow/mailcow-dockerized/issues/497#issuecomment-469847380 (THANK YOU!):
For all who are struggling with the Hetzner firewall:
Port 53 unimportant for the firewall configuration in this case. According to the documentation unbound uses the port range 1024-65535 for outgoing requests. Since the Hetzner Robot Firewall is a static firewall (each incoming packet is checked isolated) - the following rules must be applied:
SRC-IP: --- DST IP: --- SRC Port: --- DST Port: 1024-65535 Protocol: tcp TCP flags: ack Action: Accept
SRC-IP: --- DST IP: --- SRC Port: --- DST Port: 1024-65535 Protocol: udp Action: Accept
If you want to apply a more restrictive port range you have to change the config of unbound first (after installation):
Now the firewall rules can be adjusted as follows:
[...] DST Port: 32768-65535 [...]
Date and Time¶
To ensure that you have the correct date and time setup on your system, please check the output of
$ timedatectl status Local time: Sat 2017-05-06 02:12:33 CEST Universal time: Sat 2017-05-06 00:12:33 UTC RTC time: Sat 2017-05-06 00:12:32 Time zone: Europe/Berlin (CEST, +0200) NTP enabled: yes NTP synchronized: yes RTC in local TZ: no DST active: yes Last DST change: DST began at Sun 2017-03-26 01:59:59 CET Sun 2017-03-26 03:00:00 CEST Next DST change: DST ends (the clock jumps one hour backwards) at Sun 2017-10-29 02:59:59 CEST Sun 2017-10-29 02:00:00 CET
NTP enabled: yes and
NTP synchronized: yes indicate whether you have NTP enabled and if it's synchronized.
To enable NTP you need to run the command
timedatectl set-ntp true. You also need to edit your
# vim /etc/systemd/timesyncd.conf [Time] NTP=0.pool.ntp.org 1.pool.ntp.org 2.pool.ntp.org 3.pool.ntp.org
Hetzner Cloud (and probably others)¶
/etc/network/interfaces.d/50-cloud-init.cfg and change the IPv6 interface from eth0:0 to eth0:
# Wrong: auto eth0:0 iface eth0:0 inet6 static # Right: auto eth0 iface eth0 inet6 static
Reboot or restart the interface. You may want to disable cloud-init network changes.
Especially relevant for OpenStack users: Check your MTU and set it accordingly in docker-compose.yml. See Troubleshooting in our Installation guide.