MAILCOW UI NOT SHOWING

lumideez

I migrated my mailcow to a new server about 2weeksago and everything was running until today when I couldn’t see the my mail login UI (this site cannot be reached). I checked the nginx log and there was no error. Please could you help please

lumideez

esackbauer

Looks like a firewall topic then. Or reverse proxy if you have one.
Try to find out what you have changed in your setup in the last day.

lumideez

esackbauer it was working until today. The only thing I did was to setup 2FA on individual emails but yesterday i got a IMAP 101 error while trying to add a mailbox to a mail client. Other than that there’s nothing i change. I noticed the when I do docker compse down and then docker compose up -d. it starts working again then after about 30 minutes the mail UI will no longer be accessible (site not reachable error in my browser).

Also i was thinking is it due to the cold standby script on the old mail server. right now i’m confused

esackbauer

Still looks like a firewall issue. Is only mailcow running on that host? Nothing else? ufw and firewalld disabled? And you are on a officialy supported platform?
You could check if there are any connections from your test computer in nginx logs. If there are none you could also check with netstat on the host itself.

Please post the output of netstat -tulpn on the host computer.

lumideez

esackbauer Here is the screenshot

esackbauer

And what about my other questions?

lumideez

esackbauer oh yes mailcow is the only thing running on the VPS and the ufw is inactive when I checked the status. no firewall. the surprising thing is that once I run docker compose down and then docker compose up.. it starts working again until 20-30 mins when I wont be able to access the mail UI anymore and when i run docker ps while I can no longer access through my browser, it shows all containers up and running

esackbauer

Check the log files of all containers (and also from the host itself) what is happening after 20-30 minutes. What is in the logs before it fails. Maybe you can figure out a difference.
What OS are you on? Debian? Is SELInux active?

lumideez

Its running on Ubuntu 22.04 headless server.. ok I think I will wait till its down again then maybe check the nginx container see maybe thats where the problem is

esackbauer

Maybe this issue regarding AppArmor can help you:
mailcow/mailcow-dockerized5057

lumideez

esackbauer yeah I just read it but this was for Docker v23 my docker version on old server is 28.2.2 while on the new server is 28.3.3 but I dont think the issue is apparmor since all containers are up and running. I’m just waiting for it to go down again so I could check the log for each container

lumideez

esackbauer it still hasnt gone down again but the timezone shown on the dashboard is no longer correct… mails im getting now are stamped with the wrong time. is there a way I could correct that please

esackbauer

You need to make sure NTP is properly configured on your host. That could be the root cause.
https://docs.mailcow.email/getstarted/prerequisite-system/#date-and-time

lumideez

esackbauer yes NTP is properly configured and when I run curl -vk https:/xx.xx.232.xx (my server/VPS IP) inside the server itself.. it shows my webmail page on the cli…when I ping the domain from another windows computer its replying.

my VPS is hosted on Contabo.. could this issue be from them?

esackbauer

lumideez could this issue be from them?

Yes, likely.

DocFraggle

lumideez my VPS is hosted on Contabo.. could this issue be from them?

We had this problem with Contabo before

https://community.mailcow.email/d/4366-debian-12-problem-with-unbound/19

They use a custom kernel with their VPS which doesn’t has the nftables integration.
It’s very likely that the custom kernel is the problem

lumideez

DocFraggle Yeah I thought the same because I noticed this container “mailcowdockerized-unbound-mailcow-1” takes an awfully long time to start when I run docker compose down and then docker compose up

lumideez

DocFraggle And also please how were you able to solve this?

esackbauer

lumideez

As per linked thread, the only known solution is to use a different Linux image from Contabo. Rocky Linux seems to work.
Because Contabo is so cheap they need to harden their images it seems. For mail servers I recommend to pay some bucks more and get a decent provider with good track record of customers running mail servers.

DocFraggle

lumideez My Mailcow VM is running at Hetzner, there you get “uncontaminated” OS for your system 😎