Connecting to docker bridged container on CentOS 7 gives: Connection reset by peer

As the title says I have a docker image which runs fine on my development machine in docker using network bridged to an another port (-p 8080:5000). But as soon as I am deploying this on my server connections fail.

-> Ruling out not hosting the application on 0.0.0.0:5000 IP’s.

When using --network host the images do run fine.

I would like to use the Plesk docker extension. The only supported behaviour in here is using the docker bridge network.

This server is running

• Plesk => docker extension
• CentOS 7

I did have some issues with the firewall settings (There are 2 active firewalls: firewalld and the Plesk Firewall). But the issue persists when both are disabled.

-> Ruling out firewall zone trust issues.

The last answer which came up while searching is a collision in network interface IP ranges..? But I have a simple setup with a single interface eth0 (public IP), local loopback and docker0.

What could be wrong here? Any ideas how to troubleshoot this further? I am out after many hours spent researching this issue.

Thanks!

> docker ps

CONTAINER ID        IMAGE                        COMMAND                  CREATED             STATUS              PORTS                    NAMES
8c3d3f32b8ef        savahdevelopment/savah_api   "dotnet savah_api.dll"   11 minutes ago      Up 11 minutes       0.0.0.0:8888->5000/tcp   sharp_lamarr

b355e8fef0ec        savahdevelopment/savah_api   "dotnet savah_api.dll"   7 hours ago         Up 14 minutes                                savah_api_prod

e38e1b01b039        savahdevelopment/savah_api   "dotnet savah_api.dll"   7 hours ago         Up 14 minutes                                savah_api_dev

> curl http://something.hostbeter.nl:5000/admin/test
YUUUUUUUUUUUUUUP!!

> curl http://something.hostbeter.nl:5100/admin/test
YUUUUUUUUUUUUUUP!!

> curl http://something.hostbeter.nl:8888/admin/test
curl: (56) Recv failure: Connection reset by peer

Some extra info:

netstat -tulp tells me the working containers did bind to the IPv6 addresses only. But connecting externally with the IPv4 works fine? So it seems to be something network related?