I want to start this by saying, I feel really dumb that I cannot understand how custom Nginx configuration works and I feel like I have searched the whole internet to try and understand them but I am just not able to comprehend what I am doing wrong. I am aware there is a page dedicated to the configuration of Mailcow behind an Nginx reverse proxy (https://docs.mailcow.email/post_installation/reverse-proxy/r_p-nginx/) but my issue is where do I actually put this configuration? When I customise this configuration to what I believe is my correct details and add it to the Advanced section on the Nginx proxy host for my mail server it goes instantly offline when I push save. I have added it to a .conf file and added it as an additional volume on my Nginx docker container then specified it as a custom location for Mailcow on Nginx but I get the same issue. restarting both Mailcow and Nginx or removing and recomposing the containers still yields the exact same issues of being offline. I am using Mailcow in docker using this guide https://github.com/christianlempa/videos/tree/main/mailcow-tutorial for installing Mailcow. I changed Mailcow HTTP to 85 and HTTPS to 445 and have 80 and 443 ported forward to my Nginx reverse proxy. I have swapped the listening ports of the config below to both 80 and 85 and 443 and 445 for testing but still cannot get it to work. server { listen 80 default_server; listen [::]:80 default_server; server_name mail.mydomain.com autodiscover.mydomain.com autoconfig.mydomain.com; return 301 https://$host$request_uri; } server { listen 443 ssl http2; listen [::]:443 ssl http2; server_name mail.mydomain.com autodiscover.mydomain.com autoconfig.mydomain.com; ssl_certificate docker_volumes/mailcow-dockerized/data/assets/ssl/cert.pem; ssl_certificate_key docker_volumes/mailcow-dockerized/data/assets/ssl/key.pem; ssl_session_timeout 1d; ssl_session_cache shared:SSL:50m; ssl_session_tickets off; # See https://ssl-config.mozilla.org/#server=nginx for the latest ssl settings recommendations # An example config is given below ssl_protocols TLSv1.2; ssl_ciphers HIGH:!aNULL:!MD5:!SHA1:!kRSA; ssl_prefer_server_ciphers off; location /Microsoft-Server-ActiveSync { proxy_pass http://192.168.50.250:85/Microsoft-Server-ActiveSync; proxy_set_header Host $http_host; proxy_set_header X-Real-IP $remote_addr; proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; proxy_set_header X-Forwarded-Proto $scheme; proxy_connect_timeout 75; proxy_send_timeout 3650; proxy_read_timeout 3650; proxy_buffers 64 512k; # Needed since the 2022-04 Update for SOGo client_body_buffer_size 512k; client_max_body_size 0; } location / { proxy_pass http://192.168.50.250:85/; proxy_set_header Host $http_host; proxy_set_header X-Real-IP $remote_addr; proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; proxy_set_header X-Forwarded-Proto $scheme; client_max_body_size 0; # The following Proxy Buffers has to be set if you want to use SOGo after the 2022-04 (April 2022) Update # Otherwise a Login will fail like this: https://github.com/mailcow/mailcow-dockerized/issues/4537 proxy_buffer_size 128k; proxy_buffers 64 512k; proxy_busy_buffers_size 512k; } } Thank you in advance and any help in understanding how custom configurations work would be greatly appreciated!

Rightio, so here is everything I have learned! This is everything you need to configure in the Advanced tab (adjust the IP and port according to your setup for both locations.) Everything else above this from the Mailcow website is already configured elsewhere in Nginx Reverse Proxy so you are doubling up and causing issues if you try to use it. Well done to @"DocFraggle"#1599 for giving me the complete answer without me realising. location /Microsoft-Server-ActiveSync { proxy_pass http://127.0.0.1:8080/Microsoft-Server-ActiveSync; proxy_set_header Host $http_host; proxy_set_header X-Real-IP $remote_addr; proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; proxy_set_header X-Forwarded-Proto $scheme; proxy_connect_timeout 75; proxy_send_timeout 3650; proxy_read_timeout 3650; proxy_buffers 64 512k; # Needed since the 2022-04 Update for SOGo client_body_buffer_size 512k; client_max_body_size 0; } location / { proxy_pass http://127.0.0.1:8080/; proxy_set_header Host $http_host; proxy_set_header X-Real-IP $remote_addr; proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; proxy_set_header X-Forwarded-Proto $scheme; client_max_body_size 0; # The following Proxy Buffers has to be set if you want to use SOGo after the 2022-04 (April 2022) Update # Otherwise a Login will fail like this: https://github.com/mailcow/mailcow-dockerized/issues/4537 proxy_buffer_size 128k; proxy_buffers 64 512k; proxy_busy_buffers_size 512k; } Port any other ports you need for mail server 25, 465, 587, 993, 4190, etc to the IP of docker host. make sure you add all 3 hostnames (mail.domain.com, autodiscover.domain.com and autoconfig.domain.com) to the domains names section on your Nginx reverse proxy host as well as on the SSL certificate tab. You will need to download the SSL certificate, copy it to YOUR_MAILCOW_PATH//data/assets/ssl/ and rename the cert and key file to cert.pem and key.pem then restart your container. You will need to copy the new certificate every 3 months to this location manually, I have not figured out a way do get the ACME request to work through proxy. I hope this helps someone! Huge shoutout to @"DocFraggle"#1599 especially for his persistence and patience! If anyone has any ideas how to maybe get the certificate request to work, let me know and we might be able to work it out together.

Please, PLEASE use the "Insert code" button to wrap your config, nobody can read this... What exactly did you configure in your mailcow.conf, especially this part as stated in the overview page (example below): ``` HTTP_BIND=127.0.0.1 HTTP_PORT=8080 HTTPS_BIND=127.0.0.1 HTTPS_PORT=8443 ``` You set the proxy_pass part to an upstream IP address (192.168.50.250), that's why I'm asking. If you used 127.0.0.1 in your mailcow.conf this can't work

Thank you for replying, I really appreciate it! This is what I currently have configured in that section. I have tried with the loopback address and the actual IP in the BIND section but still no good. HTTP_PORT=85 HTTP_BIND= HTTPS_PORT=445 HTTPS_BIND=

@"EverythingBySam"#p15648 Set this in mailcow.conf: ``` HTTP_BIND=127.0.0.1 HTTP_PORT=8080 HTTPS_BIND=127.0.0.1 HTTPS_PORT=8443 ``` ### **_RESTART mailcow afterwards!!!_** ``` docker compose down docker compose up -d ``` And this in your prox_pass lines: ``` location /Microsoft-Server-ActiveSync { proxy_pass http://127.0.0.1:8080/Microsoft-Server-ActiveSync; proxy_set_header Host $http_host; proxy_set_header X-Real-IP $remote_addr; proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; proxy_set_header X-Forwarded-Proto $scheme; proxy_connect_timeout 75; proxy_send_timeout 3650; proxy_read_timeout 3650; proxy_buffers 64 512k; # Needed since the 2022-04 Update for SOGo client_body_buffer_size 512k; client_max_body_size 0; } location / { proxy_pass http://127.0.0.1:8080/; proxy_set_header Host $http_host; proxy_set_header X-Real-IP $remote_addr; proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; proxy_set_header X-Forwarded-Proto $scheme; client_max_body_size 0; # The following Proxy Buffers has to be set if you want to use SOGo after the 2022-04 (April 2022) Update # Otherwise a Login will fail like this: https://github.com/mailcow/mailcow-dockerized/issues/4537 proxy_buffer_size 128k; proxy_buffers 64 512k; proxy_busy_buffers_size 512k; } ``` Just like in the docs,

That seems to of worked and did not knock it offline so thank you but I would like Autodiscover to work as well if possible.

Autodiscover should work as well then as long as you set your DNS config according to the docs, and the certificate was enrolled correctly

@"DocFraggle"#p15651 I have a CNAME autodiscover record for autodiscover.domain.com to go to mail.autodiscover.com. I do not have anything configured on my Nginx itself to recognise autodiscover.domain.com When I add those extra lines from the recommend config it goes offline again.

> @"EverythingBySam"#p15652 When I add those extra lines from the recommend config What do you mean? These lines? `server_name CHANGE_TO_MAILCOW_HOSTNAME autodiscover.* autoconfig.*;`

@"DocFraggle"#p15653 nah this whole section: server { listen 80 default_server; listen [::]:80 default_server; server_name CHANGE_TO_MAILCOW_HOSTNAME autodiscover.* autoconfig.*; return 301 https://$host$request_uri; } server { listen 443 ssl http2; listen [::]:443 ssl http2; server_name CHANGE_TO_MAILCOW_HOSTNAME autodiscover.* autoconfig.*; ssl_certificate MAILCOW_PATH/data/assets/ssl/cert.pem; ssl_certificate_key MAILCOW_PATH/data/assets/ssl/key.pem; ssl_session_timeout 1d; ssl_session_cache shared:SSL:50m; ssl_session_tickets off; # See https://ssl-config.mozilla.org/#server=nginx for the latest ssl settings recommendations # An example config is given below ssl_protocols TLSv1.2; ssl_ciphers HIGH:!aNULL:!MD5:!SHA1:!kRSA; ssl_prefer_server_ciphers off;

Yes, just use the config from here: https://docs.mailcow.email/post_installation/reverse-proxy/r_p-nginx/ And of course replace CHANGE_TO_MAILCOW_HOSTNAME with your mailserver's FQDN

@"DocFraggle"#p15655 I did use that but if I add the top section I put in my previous comment it takes the proxy offline and no longer works.

Nginx Reverse proxy configuration

esackbauer

EverythingBySam So Ansible does it automatically or does it require manual intervention?

It depends 😉 you could totally automate it with Ansible. However you need to learn it.
https://www.youtube.com/watch?v=uR1_hlHxvhc

EverythingBySam

esackbauer I am always up for learning something useful which why I didn’t give up on this and ended up learning how the whole Nginx proxy worked.

Looks like I have a busy weekend ahead

Thanks for the suggestion and link.

mailer

Hello everyone,
What a great project mailcow, thank you very much for it. But I’m still failing with the setup.

I have also successfully installed mailcow in a VM. All containers are running without errors. But I also have an NPM Nginx Proxy Manager running in another VM like here.
I have now proceeded as described here in the article. Then I get a “502 Bad Gateway” when I want accessing the mailcow login. The access is then not visible in the Mailcow log.
If I leave the ports in mailcow.conf at default and do not bind it to 127.0.0.1, and delete the entries on the advanced tab in NPM I at least get a “403 forbidden” and the access can be seen in mailcow.conf with the IP of the NPM with “permission denied”
Does anyone have an idea what I am doing wrong?
How should the forwarding in the NPM look like? I currently have https set to IP:8443.
Oh and I have deactivated certificates in mailcow, I would like to do this via the NPM.

Thank for any help…..

esackbauer

mailer If I leave the ports in mailcow.conf at default and do not bind it to 127.0.0.1,

mailer I currently have https set to IP:8443

These 2 are contradicting. If you leave the mailcow ports at default (which is OK if you run NPM and mailcow on different VMs), then you need to forward from NPM to Mailcow VMs IP address on port 443 of course.

Regarding certificates, you need to copy the fullchain.pem and privkey.key to mailcow VM as described here:
https://docs.mailcow.email/post_installation/firststeps-ssl/#how-to-use-your-own-certificate

mailer

Thank you for your fast answer.
Ok I have now do the following:

mailcow.conf standard, Port 80/443 and no IP bind
NPM: delete the entries on advanced tab as in the thread above describes and forward to VM-IP:443

I recieve now “403 Forbidden” and “Permission denied” in the log of mailcow. The permission denied is about the IP from the NPM…..

Something doesn’t fit and I don’t know what or where. Mailcow does not allow my NPM to access it.

I understood about the certificate, thank you. But I can do that in the end, right? Not until the 403 problem is solved?

esackbauer

Just an idea, but maybe your tries led to a blocking of the NPM host on Netfilter/Fail2ban.
Please check:
https://docs.mailcow.email/manual-guides/mailcow-UI/u_e-mailcow_ui-netfilter/

And also make sure that you use the proper nginx configuration on your NPM:
https://docs.mailcow.email/post_installation/reverse-proxy/r_p-nginx/

mailer

I don’t understand correctly. Do I have to log in for Fail2Ban? I can’t log in because the page says 403 Forbidden. Can I just stop the Netfilter container? I also have Fail2Ban in the operating system but that stopped.

esackbauer

mailer I also have Fail2Ban in the operating system

Please don’t do this in the mailcow VM, you are adding error sources. Mailcow has already Netfilter/fail2ban function built in. Remove the fail2ban altogether and leave that job for mailcow itself or your NPM.
In the Netfilter page in your mailcow UI you should see who is blocked. I have a similar setup as you have, I have whitelisted the IP address of my reverse proxy in netfilter.
Try first to publish mailcow without the NPM, directly with SNAT rule, and then check Netfilter settings

mailer

I removed fail2ban completely with apt purge. But it was also stopped beforehand. It didn’t change anything. Still 403 forbidden.
I don’t know how to configure access without NPM, can you describe it in more detail? Do you mean access directly with https://IP in the browser? Sorry I am not really an expert at the moment.

esackbauer

mailer Do you mean access directly with https://IP in the browser?

Yes.

mailer Sorry I am not really an expert at the moment.

Then you should start running mailcow in its basic configuration. You can switch to NPM later also.

mailer

Ok then I understood it correctly and I also tested access via the IP beforehand. There is always the 403 forbidden error and a “Permission denied” entry in the mailcow nginx container. If I want direct access, do I have to change anything in mailcow.conf?

esackbauer

mailer do I have to change anything in mailcow.conf?

No, it should work out of the box unless you fiddled with some settings.
Also important to know is any changes to config files in mailcow directory need docker compose up -d to make them active. Rebooting alone does not apply changes!

mailer

Ok then it’s weird. Yes, I run docker compose down and docker compose up -d every time.
The mailcow.conf is standard except for Let’s Encrypt skip yes and an address for watchtower notification is entered.

esackbauer

If you are logged in in your mailcow VM, what output does curl http://local-IP give? Local-IP being the IP address of your mailcow VM.
Is it just a short error message or does it show the html code of the login page?

mailer

This here:
[root@mail /opt/mailcow-dockerized]# curl http://192.168.1.13 <html> <head><title>403 Forbidden</title></head> <body> <center><h1>403 Forbidden</h1></center> <hr><center>nginx</center> </body> </html>

And in mailcow nginx container:
`2024/05/13 11:11:15 [error] 20#20: *20 “/web/index.php” is forbidden (13: Permission denied), client: 192.168.1.13, server: urlname, request: “GET / HTTP/1.1”, host: “192.168.1.13”

192.168.1.13 - - [13/May/2024:11:11:15 +0200] “GET / HTTP/1.1” 403 146 “-” “curl/8.5.0”`

esackbauer

Is there another nginx installation running in that mailcow VM itself or in another container?

mailer

No I think not:

[root@mail /opt/mailcow-dockerized]# ps -efwww | grep nginx root 4518 4498 0 10:29 ? 00:00:00 nginx: master process nginx -g daemon off; systemd+ 4633 4518 0 10:29 ? 00:00:00 nginx: worker process systemd+ 4634 4518 0 10:29 ? 00:00:00 nginx: worker process systemd+ 4635 4518 0 10:29 ? 00:00:00 nginx: worker process systemd+ 4636 4518 0 10:29 ? 00:00:00 nginx: worker process systemd+ 4637 4518 0 10:29 ? 00:00:00 nginx: cache manager process
If this were the case, no log entries should arrive in mailcow nginx or?
The only docker containers are in the root context is portainer, watchtower and mailcow.

esackbauer

mailer If this were the case, no log entries should arrive in mailcow nginx or?

correct.
Then I can only assume that your mailcow installation is somehow broken, and I do not know where to search further. Maybe somebody else has an idea, lets wait.

mailer

Thank you for your effort, I had actually already deleted everything and reinstalled it, but maybe I’ll try this again, maybe on a different VM.

EverythingBySam

mailer I have had this error when I was messing around with everything. Make sure your HTTP and HTTPS bind are set to the IP of your docker container or VM with Mailcow not the loopback address (127.0.0.1) and the correct HTTP and HTTPS port that you have exposed for Mailcow access.

Make sure NPM is also set to use the IP of the Docker machine or VM you have Mailcow installed on, not the loop address as well as the correct HTTP or HTTPS port you set it to.

Finally, make sure the advanced configure is set to use the actual IP address of the VM or Docket machine not the loop back address.

Like previously stated, don’t forget to download and copy the SSL certificate into correct location in your Mailcow folder.

All this has to match!

« Previous Page Next Page »